'USO12348591B2 US 12,348,591 B2 *Jul. 1, 2025 2) United States Patent Malhotra et al. (10) Patent No.: (4s) Date of Patent: (64) NETWORK COMPUTER SYSTEM TO. G06K 130 (200601) ECTIVELY ENGAGE USERS BASED ON GO6F 11/34 (200501), FRICTION ANALYSIS GOBN 2000 201901) a Hod. 4308 02201) (71) Applicant: ZimeOne, Ine. Milpitas, CA (US) casey (72) Inventors: Manish Malhotra, Milpitas, CA (USy, 62). US. Ch Laer eraprneea er gets cr 1041, 67/585 (2022.05), GOGE 9/451 Milpitas, CA (US); Azhar Zeeshan, (2018.02); GO6F 11/3006 (2013.01); GO6F Militas CA (US) Siddartha Sikdae, 1173058 (2013.01); G6E 11/3438 (2013.01), ‘Mumbai (IN). Tom Liu, Milpitas, CA, GUN 2000 (2019.01); HOML 67/14 2013.01) Ws) (58) Fleld of Classiieation Seare Pc HOSL 67/14: HOSL. 67/535 (73) Assignee: Session AL, In., Milpitas, CA (US) See application file for complete search history (*) Notice: Subject to any disclaimer, the term ofthis (56) References Cited patent is extended or adjusted under 35 Pechisam soa US. PATENT DOCUMENTS. This patent is subject to a terminal dis- S1SSS14 BL S2012 Sheen aimer " SATSMT BL 62013 Konmgstein (Continved) (21) Appl. Now 17/698,867 POREIGN PATENT DOCUMENTS (22) Filed: Mar 18, 2022 ca. aos6i9 62015, 6) Prior Publication Data Primary Examiner — Kostas } Katsikis US 202210277211 AL Sep. 1, 2022 (74) Attorney, Agent, o” Firm — Mabamedi IP Law LLP Related U.S. Application Data en ABSTRACT (63) Continuation-n-part of application No. 17/087,298, A network computer system and method are provided in filed on Nox, 2, 2020, now Pat, No, 12,045.741, which each user of a group of users is monitored during & (Continued) respective online session where the wser performs. a sxjuence of M aetivites, to selectively engage users of the G0) Foreign Application Priority Data ‘group. A determination is made as to the impact of Friction Tor each user of the group of users with respect to an Now. 20, 2021 (IN) 202141053458 intention of the respective user, and an actin is performed for the user based atleast in part on the determined impact GI) Ince, of fiction, Go6N S04 (202801) GOOF 9451 (201801) Monitor Activities Of Individual Respect To A Particular Context Users Of A Group Of Users With 1140 Determine Impact OF ion For User(s) Of WebSite 4459) Detect User Input That Is Indicative OF Friction 4455 Detect Decrease In Intention Score Of Individual Users45q Selectively Perform An Action To Remediate Determined Friction For Individual Users Select Based On Intention Score 1160 1162US 12,348,591 B2 Page 2 Related U.S. Application Data hicl is « continuation of application No. 16/387, 520, fled on Apr. 17, 2019, now Put. No, 10,846 604 (60) Provisional application No, 62/729.995, filed on Sep. 11, 2018. (St) Int cL. HO. 4316 Host. 6714 Hos. 6750 66) 707001 vo2s7218 Tolser ss Totosst Tolsaeot 11062 360 Hse 201 Isso. 2ooddori62 20040181340 sononeisst 20080201733 2oowor7sr¥6 sonwan99440 201000082516 2100138368 doroon2sest sn100773653 ort 0066650 sorvoariess 20120047022 2n120166068 20120176500 aoivoxei212 ao1po290521 ov vooast0s 2o1yorssiet DoLsor Kc soos 20140130076 2n1go1ss00 anwso1so7se aoiaor21081 aoions7s9s anaouI2@ aniso002350 20150199707 2o1sn120%6 aoisonsese 2nisos10s4 an1s0332414 Bo B? B BD BD BI 2 BD i A al M aA A Al A Al A a al (202201) (202201) (202201), References Cited USS. PATENT DOCUMENTS. 42014 32019 yin nn0 112020 7303 woa0at 22024 52004 42007 2008 12008 1009 42010 62010 S200 112010 32011 S011 20012 6202 T2012 112012 12013 a0 72013 92013 122083 ‘Sania oania saul S201 2014 aol tans 72015 Fans 1oa0is ro20is Hoi Stephens Patan Datan.Cohen | Chhabra Malhotra, Donan Huutow Nowak Mathewson Smith Mayerilinane Engr Ghee Shit Base Stundor Ghosh ‘Gag 3002 7081419 14069 3002 T0646 «6069 3010260 “os'14.66 2ois03s03s4 AL aoa000s280 AL goren200ss Al 2one0063507 Al doia0g7is@ AL aoianr7isis Al At 7 AL al At a Al aoreursions Al aornoos6sto Al Sov7ioisoost AL aoignnao2gs AL 140837 AL aoi7ist7s2 Al aov7mr7oe Al 2017031636 AL goi7ny28 Al ao1w0046057 AL ao1gonsisos Al aorsoox2i9n Al So1w00s0737 AL sniwn14ios9 Al 20190150756 AL soiwor7407% Al anign1s999 Al doiw0n68s37 Al aorw0r03Ko4 AL aoiwoaogs09 AL gorwost4sss Al ao1s0ss760 Al Sowonzriat Al aorounyroas Al 2oio00st0%1 Al so1oon6616 AL sorouneises AL 20190130797 Alt aoron0ses9 AL aoroaliooe Al* soionrsesis Al 2oro.0266825 Al 20190373101 Al 20190388787 AL 20200081815 AL 20200082288 AL gnawoox2204 AL go200092322 Al Snauoopr9st AL ana00770339 Al a0200314603 AL 2na00x64s86 AL 2orL0117833 AL 20210215848 AL 20210312509 Al 20220067559 AL 20220086108 AI* goro07720 Al * cited by examiner e201 12016 12016 Sots Mote anis s2016 72018 2016 2016 2016 2016 112016 22017 S200 S200 $2017 2017 S207 112017 2017 22018 2018 Shots Mow S201 Shoe aos S018 S2ois 102018 o2018 112018 aos 12019 2010 22019 22019 ow sro 72019 73019 8 Tata logs Nunez Goi “ambak iden Sites Chita Chang Ghose Principe Zam Riland Chen Dovan-Cohen DatancCohen Bhat Wigder Muller Shamsi Maughan Yau Peanain al Sika Ninineo Hotinan Compo ir Wats Keishaamanty Athen Detvea Song Sinks ot al Fleming Rabinein Dey esore0n ~\ oeU.S. Patent Jul. 1, 2025 Sheet 5 of 12 US 12,348,591 B2 sas fae Go eee | addedrovag | orf B__breschedesofercl_timesta 2, Vectorize visits eeoferee|—timestam)_—_ 410 | cominuedchechad_ producud 0 _enteotnee | itemsngae Pie 4 eT piscedorer | oni ice - ah | F__ searched salePrice = a ¢_| sonsasrectout_| xe H__viewedeart POPCategory_ ProductiD: fl -dcatal 4 it -|-|3018386 | -| 2612037 |-|2812936|-|- iewedkatalee___departmen 3452312]-1-[2871123]-|-|-|2361779'————_ #12 1 viewedbome | ci salePrice: eyo ie # -|-[11.99]-132]-123.99]-|-17.49]-1-117.99]-1- Va vsuallovfrom 1-120] pagetier 7 : FIG. 4B sitios userPFM loginsttue 8, Barack predictors onsale FIG. 4A Y 420 reucine puivoe Feature Total pages in visit General visit measure | Average time per page Total visit duration ‘esearch pages Visit focus measure | %viewedcatalog pages Sviewedhome Variety measure ‘seunique POP History measure Early detection ies No Purchase FIG. 4CU.S. Patent Jul. 1, 2025 Sheet 6 of 12 US 12,348,591 B2 EVENT SEQUENCE DATA SPPCPPABNSPABEBSLHCP RAP Event 20171120181902053 201714701 81997699 2027132205951182 Timestamps 20171121305S17043 20174124109557822 20171121 105688515. ‘Time gaps fms) 0.35645 99652483 S586 40779 48693 235305 32558 292595 26329 Uompured) 258818 . Product is» AS6Y1 45764-43562 43254 43251,--- $0763 56748... Category 10s - 456 456-482 452 452-431 431. "Vepend for Search, Product page, Category page, Added to bag, Bag. viewed, Home, Entered store, Left store... ae issues cespeeceareeceaeercnvaeree san crue aces suNISEINOSSI ery eieees ieee S764 Men's Levi's 369 Loose Straight Fit Jeans ttkd_ detriment AS ea sirkd int techns siting sont a este — 2 eee te etn 456 GENDER: Men's + Department: Clothing ¢ Promotions: Sale 452 PROMOTIONS: Sole + Department: Accessorins 431. GENDER: Men's + Department: Shoes » Promotions: Sale FIG. 5U.S. Patent Jul. 1, 2025 Sheet 7 of 12 ML Model $16 sini6a | Network Computer 610 Training Process 614 Outcome info. 659 REP 669 Device Interface Lj every 612 Outcome rer eal EP 669 DL End User Device 620 iB FASS 650 Service nding inertce eG | Handling comp 952 L620 80 : ig eS pee, Yremssion) | 978 nt lection mas ae os | ? OL 620 = |] 0 i Processing itor 656 Determination Monitoring 656 nent = Vector. 658 (#) act 685 vrep 669 Network Computer System 600 FIG. 6 US 12,348,591 B2U.S. Patent Jul. 1, 2025 Sheet 8 of 12 US 12,348,591 B2 ‘SApp. 705 610 Decision Logic 630 Event Collection 648 Cc Event Analysis 706 Outcome info. 659 Updates 717 FIG. 7U.S. Patent Jul. 1, 2025 Sheet 9 of 12 US 12,348,591 B2 Communicate Event Library To End User Device Generate Decision Logic Using Machine Learning Provide Decision Logic To End User Device Receive Outcome Information From End User Device Train Machine Learning Process Using Outcome Information 850 FIG. 8 Receive Decision Logic From Network Computer System 10 Detect Activities That Correspond To Events Of The Event Library 220 Record Sequence Of Events Corresponding To Detected Activities 930 Implement Decision Logic To Determine A Value Representing Intent Or Interest Of The User Implement Trigger Based On The Value Representing The Inter Or Interest Of The User 950 FIG. 9U.S. Patent Jul. 1, 2025 Sheet 10 of 12 US 12,348,591 B2 User Record 1045 T yet ‘Connector Int. 102 > 130 . RT Activity . Data Store 134 ec 3020 Tpzisrepsiepe | aan Prediction Camp. 154 tfefels [scree , ~-Sequence Invariant Model 1055 Frustration Model 1057 Network Computer System 1000U.S. Patent Jul. 1, 2025 Sheet 11 of 12 US 12,348,591 B2 Monitor Activities Of Individual Users Of A Group Of Users With Respect To A Particular Context 110 Sequence Event Records To Reflect User Activity 1 Make Determination Of User Intention For Individual Users From Time Of Kth Activity Until Determination Event. b> 112 1120 Determination Event Is Detected 1130 Determination Event Based On | Determination Event Upon Intention Score Completion of Mth Activity; 1132 me 1134 FIG. 11A Monitor Activities Of Individual Users Of A Group Of Users With Respect To A Particular Context 1140 Y Determine Impact Of Friction For User(s) Of WebSite 4459 Detect User Input That Is Detect Decrease in Intention Indicative Of Friction 4459 Score Of Individual Users 154| ¥ Selectively Perform An Action To Remediate Determined Friction For Individual Users 1160 Select Based On Intention Score 1162 FIG. 11BU.S. Patent Jul. 1, 2025 Sheet 12 of 12 Processor 1210 US 12,348,591 B2 Instructions 1242 Memory Resources 1220 Storage Device 1240 Communication Link 1280 Interface 1250 8 FIG. 12US 12,348,591 B2 1 NETWORK COMPUTER SYSTEM TO SELECTIVELY ENGAGE USERS BASED ON FRICTION ANALYSIS RELATED APPLICATIONS ‘This application claims benefit of priority to Indian Patent Application No, 202141083488, filed Nov, 20, 2021; the aforementioned priority application being hereby incorpo- rated by reference in is entirety ‘This application is a continsation-in-part of U.S. patent application Ser. No, 17/087,295, fled Nov. 2, 2020; whieh ji. continuation of U.S. patent application Ser. No. 16°37, 520, filed on Apr. 17, 2019, now U.S Pat, No. 10,846,608; ‘which claims Benefit of priority to Provisional U.S, Patent Application No. 62/729,985, filed on Sep. 11, 2018; the aforementioned applications being hereby incorporated by reference in their respective emireties, TECHNICAL FIELD. Examples deseribed herein relate toa network computing system to selectively engage users based on ffietion analy- BACKGROUND ‘Machine leaming techniques have had increasing rel- ‘evance to growing technologies and markets. ‘Typically, ‘machine learning systems analyze and act on slored dato, sometimes communieated in batch during off-hours. BRIEF DESCRIPTION OF THE DRAWINGS FIG. 14 illustrates» network computing system for real-time event analysis, aeconding fo one or more examples. FIG. 1B illastates a variation in whieh « mobile device is used as an edge device for a network computer system sich fs described with an example of FIG. 1A, FIG. 24 illusiates an example method for operating a network computer system to engage end users. FIG. 2B illustates an example method for predicting an intent of a user IG, 2Cillutates an example method for predicting user intent with respect toa particular lype of activity. FIG. 3A through FIG. 3C illustrate example data str tures of an encoded sequence of evens FIG, 4A through 4C ilustrate example data structures for use with an encoded event steam, FIG. § illustrates an example data structure to map detected event for an end user Io olher types of information, IG. 6 illustrates a distributed notwork computer system for detecting real-time events, ueeording to one or more ‘examples, FIG. 7 illustrates a sequence diagram as between a network computer and end user device, communicating 10 ‘implement real-time event analysis using a distributed arch FIG, 8 and FIG. 9 illustrate methods for using a distrib- uted architecture to analyze events detected by an end user device, according to one or more examples IG. 10 illustrates an example network computer system to selectively engage end users ofa network site, according to one of more embodiments FIG. A1A illuscates an example method for selectively ‘engaging online users, according to one or more embod ments 0 o 2 FIG. 11B illustrates an example method for engaging ‘users based on thee detemnined frustration levels, according to one or more embodiments, FIG. 12 illustrates a computer system on which one or we embodiments can be implemented, DETAILED DESCRIPTION Embodiments provide for a computer system and method to employ a sequence invariant model to detennine user intentions, based on monitoring of real-time setivities of the "According to some embodiments, each user ofa group of users is monitored during a respective online session where the user performs a sequence ot M activities, to selectively engage users of the group. In monitoring each user of the 0p, an intention score o the user is determined ater each activity that the user is determined to perform over at least «portion ofthe respective online session. In examples, the ‘monitoring begins aller the user performs a coresponding fctvity that is sequenced as a Kth activity in the wser's sequence of M activities until a detemination event is detected for the user. In examples, the determination of the intention score uilizes a predictive model that is adapted for sequence of K setivites, andthe intention score of the user js based on information associated with a sequence of K precoding activities thatthe user performed, where K and M fare integers greater than 0, and K is less than oF equal to M. ‘In some examples, a network comptter system operates 0 etoct presence of tition for online users (e.g, Visitors of a website) using an event activity stream, where the event ‘activity stream reflects activities the user performs, a8 well as other events (eg., network generated event) in the Sequence in which the activities and events occur. Further, the network computer system analyzes the event activity stream to detemine the impact of fiction towards the individual user's propensity for completing a conversion event, In some examples, the network computer system Uilizes machine learning models with realtime event streams to determine changes to the user's propensity (ez. ter determines to abandon pursuit hecanse of fiction). ln ‘his way, examples utilize the real-time event stream which records sequetice of activities performed by the user to etoot the presence of fiction, ad further to objetively ‘measure the impact of the detected friction on the user's intention (eg. propensity to conver) In examples, “rietion” refers to activities and events hie reflect an impediment or frustration of the user towards thei intent. For example, fiction can refer to () events which occur that are deemed to thwart the user's propensity to perform an action, suchas events which diminish a ser experience (eg. page loading eror), and (i ‘activities or events which rellect rusiration on the part ofthe user (eg, speod clicks on the part of the usor that reflct emotion), Further, the “impact of fretion” is intended to ‘meant a reslt, prediction or measure that reflects how a user is impeded of deterred from performing a desired action or achieving a desired outcome. Examples provide for a network computing system, and ‘method for implementing a network computing system, t0 ‘analyze events accumulated over digital channels of aa enterprise, for purpose of detemining contextual andior cestomized opts that facilitate a desied objective of the enterprise. In examples, a network computing system (or method there) is implemented to generate an encoded sequence of user events, and to analyze the encoded sequence fora userUS 12,348,591 B2 3 ‘nent. The network computing system determines a trigger based at least in part onthe user intent, and farther performs an aetion based on the tigger andor user intent "According to examples, a determination of user's “intent” (or “intention” and other variants theeoot) includes a deter ‘ination that a sce is deemed to have a particular propen- sity oF disposition. For example, the user's “intent” may reflect a propensity ofthe user to complete an activity that js a conversion event (e.g, monetization generated from the user activity). Sil further in some examples, a determina- tion of usee’s “intent” may inelude a detennination of & [ikelhood that the user will have particular response to @ trigger (eg,, where the trigger is a particular communica tion, promotion or type thereof). As deseribed with ‘examples, a detemnination of user intent can be predictive of ‘an action the user may take As compared to conventional approaches, examples re ‘ognize thatthe latency that is introduced between events of relevance and the outcome ofthe machine leaning process ‘ean significantly hinder the accuracy and quality’ of the ‘corresponding outcome. Examples as described reduce the Tatency as between when ()eveats of relevance are detected, and Gi) a determination as to whether and how to influence the user actions is made. In the context of computer systems (ex, servers that implement network-based advertisements and "promotions (eg, such as on ecommerce sites), ‘examples enable user-specific communications and other Jnerventions to be provided to a tarpet population of users in real-time, so thatthe objectives of the computer system ‘are more elicently met ‘Sill further, « network computer system ean implement machine and deep learning techniques to process real-time data for porpose of end user engagement. In some examples, ‘4 network computer system can connect end user online bbchavior with their real-world activities, at least with respect ‘o specific retilers, using event streams that reflect real-time sdtocted events In some examples, a computer system records a sequence ‘of activities that an end usr is detected as performing. The ‘computer system analyzes the sequence of user activities in feonaction with # current user activity to predict @ user ‘nent, ancl the computer system initiates an event to cause the user to perform a desired user action based on the predicted user intent, TPurher, the computer system can similarly predict user ment as the end user continues to perform additional sequence of activities, computes an updated prediction and ‘analyzes the updated sequence of aetivities in relation to the previous prediction. In some examples, a computer system operates to detect series of sctvities performed by a user, where the activities include interactions as between the user and one or more User interface components, The computer system recognizes the series of activities asa sequence of events, where each ‘event of the sequence corresponds to one more activities of the series. In response 10 the computer system detecting @ current user setivity, the computer system selects a relevant portion ‘of the sequence of evens, The computer system determines at Jest one of a user intent or interest, based on an analysis ‘of the relevant portion of the soquence of events Sill further, in some examples, a computer system oper- ates to define session boundary, where the session bound fry includes one or more activities that are detected as being performed by a user during a corresponding session in which designated set of resources are ulized. Por each user of & first group of users, the computer system records a series of 0 o 4 session activities, including a fist activity that coincides ‘ith the defined Session boundary and one or more subse- quence aetivities. From the series of session activities of tach user ofthe ist group, the computer system determines ‘one oF more models to predict a Tikelihood thatthe user will perfoem a desired typeof activity during a current or futire Ta some examples, the designated set of resources are subsequently’ monitored for session activities. of multiple ters that are not of the first group. For cach of the multiple ‘users, the computer system utilizes one or more predictive ‘models to detennine a likelihood of the user perlomning a Gesired type of activity based on one or more session ‘ctvities detected fr that user, Additionally, foreach user of ‘the multiple users for which the likelihood is below a threshold, the computer system may eause that user to be subject 10 an intervention that is selected as a tigger for ‘causing the user to perlorm a desired user action, ‘Adktionally, in some examples, computer system devel- ‘ops models and generates decision logic Based on the {developed models. The decision logie is distributed to end user devices, and the end user devices are able to implement the decision logic to detect events, determine event sequences, and correlate the determined event sequences to predicted outcomes. Still futher, ia examples, a computer system operates to receive 2 dats set that defines a plurality of events, and ‘detects multiple aetvities that define a coresponding set of events from the phrality of events. The set of events are recorded in sequence 10 rellet an order in time in whieh cach of the multiple activitics that define the set of events took place. The computer system determines, using the sequence of evens, a vale representing an intent or interest fof the uses, and the computer system implements trigger based om the value representing the intent or interest of the 'As used herein, client device refers to deviees corre sponding to desktop computers, celular devices or smart- phones, wearable devices, laptop computers, tablet devices, {elevision (IP Television), e., that ean provide network connectivity and processing resources for communicating ‘with the system over a network ‘One oF more embodiments described hercin provide tt methods, techniques, and actions performed by a computing device ae performed programmatically, oF as 2 computer implemented method. Programmatically, as used herein, means through the use of cade of compiiterexecutable instructions, These instructions ean be stored in one oF more memory resources ofthe computing device. A programma cally performed step may or may not be automatic. ‘One or more embodiments described hereia can be imple ‘mented using programmatic modules, engines, or compo fenls. A programmatic module, engine, or component ca include a program, a sub-routine, a portion of a program, oF a software component ora hardware component capable of performing one or more stated tasks or functions. As ised herein, a module or component ean exist on a hardware component independently of ether modules oF components Alteratively, module or component can be shared element or process of ther modules, programs or machines. ‘Some embodiments describe herein can generally require the use of computing devices, including processing and memory resources, For example, one or more embodi- ‘ments deseribed herein may be implemented, in whole or in part, on computing devices such as servers, desktop com: puters, cellular or smartphones, tables, wearable cletronic evices, laptop computes, printers, digital picture franUS 12,348,591 B2 5 network equipment (e, routers) and tablet devices. Memory, processing, and network resources may all be used in connection with te establishment, use, or performance of any embodiment doscribod herwin (including with the per- formance of aay method oe withthe implementation of any system), Funthermore, one oF more embodiments deseribed herein may be implemented though the use of instructions that are ‘executable by one oF more processors. These instructions may be carried on a computer-readable medium. Machines shown of deseribed with figures below provide examples of processing resources and computer-eadable mediums oa ‘which insirutions for implementing embodiments of the ‘invention can be carried andor executed. In particular, the ‘numerous machines shown with embodiments ofthe inven- tion include processor(s) and various fans of memory for holding data and instructions. Examples of computer-read- able mediums include permaneal memory stonige devices, such as hard drives on personal computers or servers. Other ‘examples of computer slonige mediums include portable storage units, such as CD or DVD units, flash memory (such 4a carried on smartphones, multifunctional devices or tab- Jets) ane magnetic memory. Computers, terminals, network ‘enabled devices (ee, mobile devices, such as cell phones) ae all examples of machines and devices that utilize pro- ‘essors, memory, and instrtions stored on computerread- able mediums. Additionally, embodiments may be imple- mented in the form of computer-programs, or a computer ‘usable earier medium enpable of carrying such « program. System Deseription TIG. LA illustrates network computing system for real-time analysis of user activities and other events, accord- ing 10 one oF more examples. In particular, a nctwork ‘computing syste 100 can implement procestes to capture ‘events in real-time, across one or multiple chaonels, and Jurher (0 implement processes to analyze and aet on the tected events, With fespect to examples as described, the system 100 can be implemented on a server, on a combi nation of servers, andor on a distributed set of computing devices which communicate over a network such as the Internet. Still father, some examples provide for the net ‘work computing system 100 to be distributed using one oF more servers and/or mobile devices. In some variations, the network computing system 100 is implemented a par of, oF jn conection with a network system, where, for example ‘end users utilize applications running on mobile devices to ‘engage in either online or real-world services (e., shopping in tore or online), The system 100 can provide user context data fom real-time streams, historical datastores and enterprise sys- tems for a mange of event events and event types. la particular, the system 100 ean implement Jeaming and ‘analysis for processing event patiers using a real-time response framework. Depending on the implementation, the real-time response framework can deteet and respond t0 user ‘events in seconds, less than a second, millisecond scale oF ‘even less, In examples, the system 100 can implement a framework to generate a tea-time response to a particular trigger oF ‘event, for purpose of causing or influencing desired ‘outcome fiom the end user, While the response to the panicular trigger or event can be generated in realtime, the response ean also be tailored or otherwise configured based ‘on historical information about the end user who is the subject of the response. Accordingly, as deseribed with various examples, the system 100 can stilize a combination ‘of information that is obtained in real-time as well as fom 0 o 6 prior dime intervals (e, historical information), in order to ‘gencmite real-time response toa particular rigger or event. ‘With further reference to FIG. 4 the system 100 includes ‘one of more fypes of connectors 102, with each connector ‘implementing processes for communicating with a partiew lar type of device andlor enterprise resource (e-., website service or channel provided for users, ete.) As desribed in greater detail, the connectors 102 ean represent an aggre tion of processes that collectively aggregate activity nor ‘mation 101 in real-time from various user-operated devices andor enterprise resources. “According to examples, the connectors 102 ean include processes that communicate with applications running on ‘eves, as well as processes that are provided with network ‘or web resources of an enterprise, For example, the connec- tors 102 can include a mobile device application connector 102A that s implemented using. combination of a network or server-side process (or set of processes) and an applica- ‘ion program interface (APD) of a corresponding wser-oper- ated device 98 (eg. mobile device). As an addition or fltemative, the connectors 102 ean inchade a website con- rector 1028 that is implemented using a combination of a network or serverside processes) and one or more pro- grams that are implemented with the website (or other enterprise resource 99) through a software developer kit (SDK) of the website. As an addition or variation to eon- rectors which collect information from user-operated devices, the system 100 can implement connectors 102 10 ‘monitor other types of enterprise resources 99 for activities performed by users. Such connectors 102 can employ peo- cesses that, for example, execute in a computing environ- ‘meat of an enterprise resource 99, in connection with & service that is provided to users (eg, chatbot). ‘With reference to an example of FIG. 1A, the system 100 includes multiple different connectors 102 to receive and record activity information 101 from one or more kinds of resources, sch as from user-opersted devices 98 and from enterprise resources 99. In examples, the activity informa tion 101 inclides data which identifies an activity the end user performed, using a corresponding user-operated device andar enterprise resource. The activity information 101 can also include one oF more wser-identitirs for the respective ‘ser that performed the corresponding activity. By way of example, the ativity information 101 can include one or ‘mon’ identifiers that reflect any one of a user account, a useroperated device, a session identifier (eg, such as identified by cookie information stored on a user device) andor a user signature determined from multiple atibutes fof the end user's interaction wih the respective resource of the enterprise. Additionally, the activity information ean include descriptive information about the activity per formed. For example, the atvity information ean inelude dscripive information relating to an item that ste subject fof the end user activity. Still fan, the etivity information 101 can include contextual information (eg. time of day’ day of week, calendar day, ee.) elated to an activity thatthe fend user ig detected as having performed. ‘In some examples, an aetivity that a given user is detected as performing can be in connection with the end user's access or use of an enterprise resource 99 (e., website, ‘mobile device application, chavbot ete). For example, the ‘mobile device application connector 102. can comitunieste ‘witha program tha executes as pat of a mobile application fof an enterprise, to receive activity information 101 that perians to the ead user's interaction with the enterprise's ‘mobile application. Asan addition or altemative, the mobile device application connector 102A can interact with ()US 12,348,591 B2 7 third-party applications rumaing on the corresponding mobile deviee, andior (i) one or more APIs that are avail ‘able on the mobile device to obisin activity information from device resources (eg. satellite eceiver to sample for location information, accelerometer and/or gyroscope 10 sample for movement information, camera, mierophor ‘etc.) andior software resources (ex, thirt-pary applica tions) In such examples, the determined activity informa tion relates to setivities which the end user performs on & ‘corresponding mobile device, distinct or independent fom ‘any enterprise mobile device application, For example, the mobile device application connector 102A can communicate with one or more APIs on the comesponding mobile device to detemnine activity information that reflects a user's inter ction with a website and/or mobile device application, as ‘well as a location of the mobile device (eg,, as determined from the mobile device satellite receiver). As another ‘example, the mobile device appliation connector 102A can, include an API provided with 3 corresponding mobile appli ‘ction to obtain sensor information by, for example, reading from a satelite receiver or other loeation-aware resource of the mobile device, andlor sampling an socelerometer andor zyroscope of the mobile device. Sill further, the mobile ‘device application connector 102A can interface with other programmatic resources of the mobile device, such a8 with third-party applications or services which run on the mobile device, Tn variations, the connectors 102 can inchide other device and/or programmatic interfaces that execute on a user- ‘operated device. For example, the connectors 102 can include programmatic interlaces which execute to comm- nicate with processes running on diferent types of scr devices (eg., desktop computer, wearable device). In such ‘eases, the connectors 102 can communicate to receive activity infomation from multiple types of wser devices (ex, desktop computer, wearable device, et.) ‘Aaditionally, in examples, the website connector 1028. ‘can also eccive activity information 101 which relate tothe ‘teraction of individual users with respect to a website andlor designated veh resources of an enterprise oF website publisher. The activity information 101 may be collected through. for example, one or more processes that execute with the website and/or respective web resourees to detest user activity (eg. page view, search). Stil Mtber, dhe activity information 101 can be collected from programs that execute on end user deviees (e-. ‘are published as part ofthe webpage), In contrast to moni toring user devices and resources for corresponding user activity, the website connector 102B can monitor enterprise fo publisher resources for activity information generated from the interactions of a population of users. In such ‘exumples, the enterprise or publisher resources cua inlude, or example, a website, « kiosk, of @ network-enabled fiducial or encoded object distributed ata given locality. In variations, the connectors 102 can include interfaces 10 ‘communicate and receive activity information 101 from altemative instrumented venues, sch as physical stores that Uuilize aetWork-enabled resources, suc as cameras, eae ‘cons, Wi-Fi access points, JoT sensors, and devices. In examples, the connectors 102 store event records 113 in the reabtime activity store 134, where the event records 113 are based on a corresponding tivity information 101 With respect to a particular user, the event record 113 ean ‘dentfy an underiying activity or activity type which the end user is detected as performing indicated hy the respective ‘activity information 101 0 o 8 In some implementations, each event record 113 ean include of otherwise ink to one of more attributes that are identified from the corresponding activity information 101 ‘The attributes of a given event record 113 can include parameters that reflet descriptive information about the Selected event, sich as, for example, information about an item that was the subject ofthe even, as well as contextual information regarding the performance of the event, Sil further, given event record 113 can be associated with one for more identifiers of a user who is detected as having performed the underlying activity. In some examples, each Selected event of the sotvity information can be associated with a record or data set ofthe roaktime data store 134, to reflet associated information, such as attributes of the event (ext, descriptive information about the events), and ident fiers of the end user performing the respective events ‘In some examples, an event encoding component 120.can ‘encode the event records 113 ofthe real-time data store 134 In examples, the encoding component 120 assigns a cade value to event reoords 113 based on a predetermined encod- ‘ng scheme. The event eneoding component 120 may, for ‘cxample, utilize an encoding scheme which assigns code valucs to event records based on a categorization of the event record. In some implementations, the encoding scheme ean assign like eade values to event records which hue a common categorization. The encoding scheme can utilize, for example, an enterprise profile 124, to determine ‘event categories, where each event category identifies ativi- ties that are similar by nature, or deemed similar in regard to theie respoctive probative valve for making. predictive terminations about the end user. The event encoding ‘component 120 can also sequence events 113 for individual ters based on time-stamps associated with cach event record 113. The activity data store 134 may then associate an ‘encoded event steam with each user. Tn some examples, the activity data store 134 ean be implemented using cache memory to enable rap read ‘operations from the learning sub-system 150, A profiling ‘component 130 can copy the contents of the activity data store 134 ta comesponding historical data sore 132, In this ‘way, memory management resources can repeatedly fish the real-time data store 13410 remove data that may be aged, based on a predetermined threshold of its (e., data older than 5 seconds). Ta examples, the profiling component 130 develops user profiles for users ofa given user-base (eg, users Who acess ‘an enterprise website). The profiling component 130 can include a profil store 137 whieh includes information about users ofthe end userhase (e.g. registered user-base). The profile store 137 can associate user-identifiers, such as account identifier or session identifiers (eg. cookie iden- ‘fie, with information tat is provided about the end user by the enterprise and/or the end user, The profile store 137 can also associate the end useridentifers with information that is developed or leamed about the end user through ‘implementation of processes deseibed by various examples below. In some examples, profiling component 130 can also include logic to identify historical datasets from the real time data store 134. The profiling component 130 can ‘generate the historical data siore 132 from the real-time data store 134, basod on events that age beyond a threshold uration (e_2, events which are more than ane hour ol) ‘The historical data store 132 can store, fr individual users, the encoded event streams for event records that have aged beyond the threshold duration of the real-time data store 1MUS 12,348,591 B2 9 Sill further, the profiling component 130 develops event profiles 131 for individual were. The event profiles 131 can Tink event records 113 of the real-time aetivity data store 134 swith corresponding records of the historical data store 132, ‘The event profiles 131 can match user identifiers of corre sponding event records 113 and thir respective encoded ‘event sireams with corresponding user identifiers of data sets stored withthe historical data store 132. In variations, vectorization logie 138 can be implemented to generate vector representations of user activity profiles sored in the istorieal data store 182 and/or the real-time activity data stow 134. The veetorzation logic 138 ean generite vector representations for encoded data streams ‘associated with the end user inthe historical data store 132, In variations, the vectorization logic 128 can also generate the vector representations Tor encoded data steams whic may be stored with the realtime activity data store 134 Accordingly, in examples, the profling component 130 can provide ie leaming sub-system 150 with activity profiles of individual users which can include the end user's encoded 2 ‘event stream as well 3s vectored representations of theend user's encoded event stream 121, Additionally, the atvity profiles of the individual users can exist in different forms Tor different time periods, For example, the encoded event stream can include one oF more vectorized representations, to represent the portion ofthe end user's activity stream that js aged beyond a threshold time period (e.g. 1 month or 1 yea). In examples, the learning sub-system 150 can implement multiple types of machine leaming and deep leaning teh- niques to process the activity profiles of individual users. In particular, the learning sub-system 180 can implement machine and decp learning techniques to leam from end users who interact with an enterprise across one or multiple digital channels, As described with various examples the Teaming sub-system 150 can utilize the encoded event streams 121 fora given subject (e., end user) that extends ‘overa window of time, from present moment (eg. eal-time for near real-time) toa selected moment in the past. Addi tionally, the leaming sub-system 150 can utilize vectorized representations of the end user's encoded event steam 121 spanning a timeline that extends beyond the window of ime. In examples, the leaming sub-system 180 includes a stream interface component 182 which monitors input into the real-time aetivity data store 134, When data pertaining to ‘a new subject (eg, end user) is received, the stream inter- Tice component 182 triggers analysis by one or more of the Jmoligence processes of the Teaming sub-ystom_ 150. Depending on the model and technique applied, the stream Jimterface component 152 can retrieve portions of the sub- jects encoded event stream from the historical data-store 4132, so ast seamlessly combine the historical and real-time portions of the encoded event steam 121. In variations, dhe Stream interface component 182 can also retrieve the vec- torized representations of the end user's encoded event stream. The stream interface component 182 ean furrher ‘query the historical data store 132 in aocordanee with & ‘configurable window of time. Specifically, the stream inter- Tice component 182 ean determine the past time and the ‘current time to define the window of time. The configuration ‘of the window of time ean be based on, for example, settings forthe enterprise, and further optimized for eontext and use In this way, the encoded event stream 121 ean represent 8 ‘continuous collection of events forthe given user, extending between a past time and a current time. Por the respective ‘window of time, the stream interface component 182 can assimilate the encoded event steam 121 using portions 0 o 10 stored in realtime data store adr the historical data store Aditionaly de steam interface componeat 182 can incor porte the Vectorize representation of the encoded event Stream for sole ime peods, such as fr atime prio tht ented beyond that Window of time used t asimilte the ovoded event stim 121 Tn examples, the window of time can be configured 19 ‘dently the end user activities between anyone of () Bix Start and end times, (i activity fom a fxed star time uti preset moment, (i) activity over prior duration of time, unr (iv) activity daring & measured tine-intenal (ex, very 10 second). “According o examples, the Iaring sub-system 150 can proces the encoded event steams) 121 of the end user to ‘make intelligent deteminations for the user, and furter to termine an action for the given subject in realsime (or eat realtime). For example, the leaming sub-system 150 an implement an inlligent process (eg using © machine or deep learing technique) to determine # user intent, a current ser context, ad relevant past user cooten. Fin the determinations, the leaning sub-system 180 can deter ‘ine an ouput tat engages the end user in 8 manner tat Jr likely ta nBcace the ead wser action, and farther to Provide a desired outcome for both the end user and the terpise. For example, the learning sub-system 150 ean aencrte ¢notiication or other interaction with @ subject Within a Second, or even milliseconds using the encoded vent steam 121 of a given subject, where the encoded vent steam 121 inchides © coninoous sct-of events detected for to subject cross window of ine that extends {fom the present moment (e. sing real-time dats) back wands in ime toa selected moment in the past (e., tsing istvial dat) ‘examples, the intelligent processes of the ering subsystem 150 ca include (a prictive componcat 154 {© make a predictive derminaton about te end se, based fon the encoded event steam 121; (i) an intervention com- ponent 186 10 determine an intervention er engagement {eg chanel selection for sending notification, ining of oieatio, content ofnatifeation for the end user, andor (Gi) scastomization component 58 that customizes given ‘ser experince, foe # given channel (e online) oF real ‘world event t cause. promote or otherwise inlucace particular senimeat, ston or outcome from the end use. ‘The fntlligen processes ofthe leaning sub-system 150 ean {rer be configured with enterprise specific data and logic The intligent processes can genenste one or more triggers 185 forthe end user, where the iggt 188 can ident, for example, porameters fr the action that an event hander 160 Js to take with espect othe end user. The parameters can, for example, set the communication chanael for engaging the end ser (or recipient who is to engoge the end user) spel the ining, and further the content Intra. the eveat ander 160 can ake the action tsing a respective connector 102 for the selected chan. Tn examples, the predictive component 14 canbe mple- ‘mente using a mochineeaming process or mods, where the machine-feaaing process i tained to futher patcn Jar objective o outcome of the system 100, By Way of example, the predictive component 184 ean be configured to ategorize users in accordance with set of predictive atgoris of categorization schem, where each category of the categorization schema categorizes the end use ia fccondance with a prediction about the wer. The eaegor- zation schema can be made specific oa vary of facto, Including a desired outcome for a particular context (ea increase propensity of User at ecommerce site To makeUS 12,348,591 B2 u purchase). Still further, in other examples, the Feared mod- ‘els ofthe learning sub-system 180 ean model for an outcome ‘of series of user events, In such examples, the leamed ‘models can utilize one of multiple possible machine learning ‘algorithms, such as @ random forest algorilim or a neural network, to predict a particular ottcome to a sequence of ‘events dotcted from user activity. In such examples. the categorization can be determined through implementation of a machine-earing process. Ia ‘examples, the machine-leaming process can use, as input, a sequence of encoded events relating (© a panicular user ‘setivity. For a particular end user the categorization can reflet, for example, a likelihood that () the particular user has a given intent or propensity, (i) the particular user will take « particular action, and/or (ii) the particular user will respond to a particular intervention in @ particular manner. In some examples, the categorization schema can be a binary determination, such asa determination as to whethee the end user is likey to perform a particular action without 2 ‘ervention, For example, the categorization can rellest Whether a given end user is Tikely to makelnot make @ purchase, In varistions, the categorization schema can predict an ‘outcome for the end user based on a particular type of > intervention, Ths, the categorization schema can identify ‘end users who will have a threshold likelihood of having 3 ‘desired response to a panieular intervention, Alternatively, the categorization schema can identify a type of interven- tion, from multiple possible interventions, that is most ikely to result i the end useeseting in accordance witha desired ‘output [By way of illustration, examples recognize that one ofthe primary reasons end users drop out of an online checkout process is hecause the end users are not sullcintly assed ‘on purchasing 2 product, such a high-value items, using an ‘ecommerce channel. In many cases an end user may be indecisive (e.g, "sit on the fence”) for prolonged periods of time without taking any farther action to conclude the trunsoetion. Examples recognize that such eases of impend- Jing or near purchases can be influenced to completion oF the transaction (eg, "converted”) if the end user is “persuaded” (o visit the store. The Teaming sub-system 150 can selee- tively implement machine Ieaming techniques on the ‘encodes event seam of individual end users, to identify end users who can be influenced to perform or complete a teansaetion. Onee such end users ate identified, the event handler 160 can implement processes to learn elfective ‘engagement actions with respect o individual end users. For ‘example, the event handler 160 can target select individuals, ay identified by the learning sub-system 180, with person- ‘alized messages to invite them to a real outlet (e. physical store, online store, ot.) where the end user can Sample the product andor engage with salesperson (ez. an instore sales associate). Alternatively, the event handler 160 can determine a personalized incentive forthe end user to vist the soe In variations, the event handler 160 can select to generate recommendations and other messages to on-floor sales asso- ‘ites relating to end users who are in their stores, of whom they are engaging personally or through another commni- ‘ation channel online medinm). till further, examples ‘ean improve in-store user experience, by enabling faster ‘checkouts, reduced queue times, and efiient store pick-ups {or online purchases, In this way, the system 100 can be Jmplemented to miero-taret end users, by way of, for o 12 example, end user-specific messaging, communication channels, timing. ander incentives (eg, loyalty points iscounts, pricing). Tn examples, the learning sub-system 150 can implement intelligent processes to make predictive determinations for the end user Por example, the predictive determination can reflect the likelihood that the end user will complete or otherwise convert 1 transaction (eg, purchase an item) if ven a particular trigger (e.., notified that the particular stem they were looking at isin the store they are visiting) For example, the predictive determination score can reflect ‘decimal value between 0 and 1 (ex, 0.55), reflecting the Tikelihood of the determination, Still further, in other ‘examples, the predictive determinations ean personalize an particular type of produet, the customization component 158 ‘an send a aofcation or message to a representative of the enterprise t enable an agent to focus the end user's expe- ence about the product they are most interested in ‘While some examples of FIG. 14 implement the system 100 in context of online andor real-world shopping, in variations the system 100 can be implemented in context oF real-time offers, a well as other context of in-session information belp, in-session eros-sel upsell, omnichannel personalization “and transaction completion, online 1 branch, event-based interactions, watch lists and alerts, {aud detetion and setion, in light Wiggers based on evens, ‘nd cll center analties and event pattem detection. Stil further in some examples, the system 100 is implemented in context of system that intervenes in the course of a user interaction with a website, The intervention can be ia the {orm of actions which are taken to promote the user lowards a panicular outcome. For example, the intervention can Prive a promotion or offer that makes it more likely that the user Will perform a conversion event. As another example, the intervention ean provide the user with @ ‘communication channel (eg, chatbot, live agent to address problems that may he easing sce frustration. In some variations, functionality as deseribed with ‘examples above, for network computer system 100 cat be istrbuted! to the computing devices of the end users or subjects. For example, # mobile application of an end user evice can operate to implement some of the functions or ‘eatures, as deseribed with system 100. In such variations, a mobile device can exeeute an application that records the ‘occurrence of certsin, predefined user events related to the mobile deviee. To illustrat, the end user may take a pietare ofaproduet wien Walking na sore and the picture capture ‘when eross-elated tothe location of the user, can identify the event. Sill further, the end user can operate the app 10 place an item with an online retailer in shopping ear. Alternatively, « detected event ean correspond to the end ‘ser being detected ss visiting a location, such asa shopping small oF sore. FIG. IB illustrates a variation in which « mobile device is used a5 an edge device for the network computer system 100. In an example of FIG. 1B, mobile application 196 can ‘execute on a mobile device 190 to perform operations that include (i) detecting predelined evens, or series of evens, and (i) encoding the events, so as to generate a local encoded event stream 191. By way of comparison, the Tocally- encoded event stream 191 can form a truncated portion ofa corresponding encoded data stream 121 forthe same subject or end user. “The mobile aplication 196 can further communicate with the system 100 via a comesponding connector 102A,US 12,348,591 B2 13 ‘examples, the mobile application 196 can generate requests that are responsive to, Jor example, a newly detected event Iv tuen, the connector 102 ean make @ programmatic ell to the fearing subsystem 180, and more spocifially to Jmolligent processes ofthe Teaming sub-system 150, such as predictive component 154. The incligent processes can also be called by the connector 1024 to determine, for ‘example, a user intent. In some variations, the intelligent processes can futher determine a tegaer 185 o¢ action that fs to be performed on the device. The connector 102A caa respond (0 the mobile applicaon requests with responses that identify @ desired outcome ofthe intelligent processes (ea, intent, trigger, action, et.) In this way, the mobile device 190 of the end user ean be operated as an intelligent ‘edge device that utilizes machine or deep learning to imple= ‘ment real-time operations tnd aetions locally on the mobile device 190, FIG. 2A illustrates an example method for operating 3 network computer system to engage end users. FIG. 2B ‘iustrtes an example method for predicting an inte of @ twser FIG. 2C illustrates an example method for predicting user intent with respect to a particular type of activity. Examples such as deserihed with FIG, 2A through FIG, 2C may be implemented using a system such as described with Jor example, an example of FIG. 1A. Accordingly, reference may bende elements of FIG, LA for purpose of lusting suitable components for performing a step or sub-sep being described, ‘With reference to an example of FIG. 2, the system 100, may be provid with a data set that dafines. a set of ‘immediately relevant context (210). The immediately rel ‘evant context can correspond to, for example, online sbop- Ping activity of the end user (eg, what online retail site and products the end user viewed, what tem the end usee placed tm online car, etc.) as well as historical aetivity (eg. end user's loyally award program for a given retailer) and real-world activity (eg. end user walking into store, end user Walking ino store fora panicular purpose). Fora given set o individuals (eg. end users), the system 100 selects a tigger for each individual (212), By way of ‘example, the system 100 can implement machine leering techniques to determine triers that are personalized for individuals, and fora specific context or use. Moreover, the triggers can be based on multiple conditions that are spevitic to the given user, By way of example, the activity informa- tion 101 of the user ean obtain information from the user's mobile deviee, o detect that an end user is walking into @ retail store, This determination ean be made in real-time. Likewise, time interval between the recorded events ofthe user can indicate that the end user placed items from the ‘online store into a cart for purchase, and the particular retail, Jocation ofthe end user has the same item as isin the user's conline cat. “The system 100 can select an action to perform forthe end user In Variations atime when the selocted action is to be performed may also be selected based on the selected action ‘and the likelihood of a desired outcome occuring (218) Sill further, the particular channel in which the action is 0 be performed or initiated may also be selected, based on, for ‘example, the selected aetion and the likelihocd of a desired ‘outcome occurring. The actions ean correspond 10, for ‘example, sending notifications or messages to the subject of ‘end use, oF toan agent who ist interact with the end user, bused on an oxteome determination tat is predictive ‘With reference to an example of FIG. 2B, the system 100, ‘ean record sequence of events that reflect the user's ‘activities, as detected with respect to @ particular resource 0 o 14 (ex, website) or through a particular computing platform or channel (222) In some variations, fora givea user, activity information can be received and recoeing using one oF ‘more multiple connectors 102, such that the detected soquence of evens ean reflect user activities detected across ‘multiple channels (eg, activities of the user performed on diferent types of devices andior computing environments). Tn examples, the system 100 can analyze the sequence of ser setvities in connection with 2 cuent user activity 10 predict a user intent (224). The prediction component 184 fan, for example, analyze the user's encoded event steams 121 (eg., provided by the historical data store 132), in connection with an identified real-time event or event sequence (eg. as reflected by event records 113 of the realtime daa stone 134), to make predictive determinstion of the user intent In response to making the predictive determination, the Jeaming sub-system 150 can initiate an intervention that is intended to cause the user to perlonm a desired user action (226), The initiated event can correspond to, for example, ‘communicating message to the end wer, where the message includes, for example, a promotional offer or other content. In such examples, the communicated message can be deter ‘mined by type, based on a predictive determination of the Jeaming sub-system 150/35 to the intent of the user. In variations, the deteraination of the learning sub-system 150 includes an inital dotermination as to whether, for example, the user should be subject to an intervention, For example, the predictive detemination may be thatthe intent of the particular user is firm, and likely not subject to change with {he intervention In other variations, the learning sub-system 150 can select the intervention for the user, based on & predictive determination that an intervention of a particular {ype is most-likely to generate a postive response ‘With reference to at example of FIG, 2C, the system 100 ‘ean be implemented to define a session boundary, where the session boundary is to include one oF more activities that users of population can perform with respect 10 a desig- rated. set of resources (232). In an aspect, the session boundary can be defined with respect to a set of resources that are published or othervise provided fom an enterprise resource 99. To ilstrate, the designated set of resources can cormespond to a website operated by an enterprise, the mobile device application published by the enterprise, ne- work sites accessed by mobile devices, andor a iosk that is operated by or for the enterprise at a locality of the enterprise. In such eases the type of user activity ean inelude activities users of a population perform with respect to the enterprise's website, mobile device application or kiosk. In sch context, the types of user activity which can be detected ‘an incade users visiting « spovific website, users launching an application for a mobile device or users operating. @ ‘mobile deviee application to access a particulir network resouree published by the enterprise, or a wser's initial set oF interactions witha kiosk or loeality-based resource Tn such eases, the session boundary ean be defined in part by an initiating event, or sequence of events, coresponding to activities of individual users with respect tothe website oF other designated resource of an enterprise. A session bound- ary can futher be defined by a period of inactivity, atleast with respect to the designated resource, preceding the ini tinting event, By way of example, the session boundary ean be defined to correspond to user initiating an online session (ex, usor opening browser, user launching mobile d application, ec.) in which a website (e-., ecommerce is accessed (eg, user views or downloads home page or Janding page from browser or through mobile device apUS 12,348,591 B2 15 cation), As an addition or variation, the session boundary ‘can be defined by the occurence of a threshold period of inactivity with respect to the designated resource, followed by the initiating event (eg. user opening browser, user Junching mobile device application, ee). In some variation, the channel/device connector 102 can include processes that execute on the website resources of the specific website to detect users of the population access- ing the website andr performing other activities (9, page view, search, placing item in shopping cart, ee.) AS cone necior 102 can utilize processes running on the website (© detect trafic, the recorded information can include activity jnformation for both known and unknown users. Each user to aecess the website can be ideniied through a variety of ‘identifiers, such as through an account identifier (e., if user Jogs in w service or account managed through the specific website), cookie or local device data identifi, oF through machine identifier. In examples in which the designated set ‘of resources include a website, the type oF user activity that 2 ‘ean be monitored and detected can include page views, click, mouse over, site search, item selection, item pre purchase activity (place item in shopping eat), or conver sion event. The recorded activity information can furher be ssociated with one or more identifiers, For known uses, the ‘dentifier associated with the user's activity information can be correlated w historical activity information of that user such as recorded by historial data store 132 For each user ofa first group, the system 100 can further record a series of session activities, including a first activity that coincides with the defined session houndary and one oF ‘more subsequence activities (234), In examples, the frst group of users can correspond to known users, ora subset of known users (eg., known users who have accounts with the ‘enterprise ofthe designated resource). For known users, the activity information can be recorded across mnutiple S23- sions, with each session being marked by a session bound- ay Tn examples, the ecorded activity information can reflect ‘events and attributes ofthe event. For example, the activity Jnformation ean reflect an event of search, dnd an attiute ‘of the event may comespond tothe search teem, To farther the example, a subsequence event, as identified by the ‘ctvity information, can reflect a page view, with atiibutes ‘of one or more produ identifiers (¢., product name, manufacturer name, SKU, ete). Additionally, the activity information ean reflect contextual information, such as (}) ‘information about the device oF application the end user used during the particular session, (i) the timing of the activity (eg. ime of day, day of week, month of yea, et.) (ii) information about how the end user may have initiated the session (eg. through use of a landing page, bookmark ‘or subscription link), Gv) geographic information about 3 location ofthe end user (e.g ifavailable) and/or (v) whether the end user responded to a promotional event (eg. sale on site in connection witha marketing event) and whether the promotional event related to the item which isthe subject of a subsequence conversion event. ‘Additionally, the activity information of known wsers ean be processed for the occurrence of a desired aetvity. la ‘examples in which the designated resouree is an e-com- meree site, the desired sctivity can coerespond to, for ‘example, a conversion event where the end user purchases fan item (eg., product, service, etc.) In variations, the ‘conversion event can be defined as the wer placing item in the shopping cat or requesting more information about an 16 item. The desired activity, or type of activities can be predefined for a given enterprise, based on, for example the enterprise profile 124 In examples, the event encoding component 120 can process the recoded aetvity information of known users t0 ‘entity evens, and further io determine a sequence amongst the identified evens. The identified events ean be encoded, and each event may also he associated with coresponding tributes as detected through the respective connector 102 to, as described. In this way, the real-time data store 134 can store event sequences in association with identifiers of ‘known users, for corresponding sessions in which dhe knowa users utilize a website or other resource of the enterpse ‘Additionally, the sequence of events can further reflect an outcome with respect to whether the user performed a desired activity. For example, the profiling component 130 ean determine an event sequence subset for each known ‘user, where the particular event sequence subsct includes () a frsin-time event, such as an intiting event as defined by the boundary, and (i) a as-in-time event, coresponding to ‘conversion event o an event in which the session is terminated (¢., user logs of forecloses the program used 10 access the enlerprise resource). In examples, the profiling ‘component 130 sequences event records 113 for individual users to reflect timing and order in which eoreesponding user factvities and other recorded events occurred, In this Way, the profiling component 130 associates individual users with ‘corresponding ser record anda plurality of event records 113, For a given user, the fist-in-time event ean be repre= sented by an event record 113 where the user accesses a website, where no prior historical information is known about the user. In such examples, the real-time activity data store IM retains ser records, cach of which are associated ‘or include event records 113 that reflect a sequence of user fctvities and other events. The profiling component 134 can urtheraecess the activity data store 134 to categorize andlor soore the user based on, for example, their intention oF Bropensy 0 perform «| ghen ation (ex, coverion The learning sub-system 150 of system 100 may further develop ane oF more models to predict an outcome of a user session, where the models incorporate leaming using session activities of known users (236). The developed models can for example, utilize a random Jorest or neural network sethodology. In examples, the prediction component 154 can aggregate event sequences of known users during indi- vidual sessions, where the aggregated sequences can reflect the occurrence oF nonoceurrence of a conversion event, as well #8 termination andior the initiation oF session® as defined by the session boundary. Through aggregation, the prediction component 134 can determine patterns rellecting the oecurence of specific evens or event types (@. any ‘one OF muliple possible events relaing 10 a particular tribute, such as a particular product or type of product), an rede io am oer when determina tems ecu amongst the own users. In some variations, the predicted outeome is made fora partctlr ters cameat session, using such a determined pattern for that usee. In romponent 184 can utilize saregations to develop, for example, an ensemble of models, ulizing analysis of events identified from activity information of known users. ‘In examples, the aggregation of evens and/or event types can reflect graph-type data stricture, where nodes of the graph reflect the occurrence ofan event, event sequence, oF teveni(s) of a panicular type. In such examples, the node levels ofthe graphetype stricture can reflect a relativeUS 12,348,591 B2 17 jn a user session when a corresponding event, event type or st of multiple events occur. For example, an initial level ean reflect various events or event types Which are possible initiation events for a session, as defined by the boundary session, Each node ean frther be associated with one oF more possible outcomes. Ths, for example a branch fol- lowing given node can be associated with the likelihood of tn oulcome, were the outcame reflects whether a desired activity occured. Through aggregation, a likelihood of one ‘or more outcomes can be determined at nodes of each level Additionally, the uncertainty of a particular outcome oceur- Fing can lesen at each level, suk that a particular outcome becomes more certain asthe node level ofan event sequence ‘Depending on implementation, the likelihood of the out= ‘come can be represented as a probability, and @ particular ‘outcome can be deemed likely or suficently ikely ifthe probability excveds a threshold value (eq. if probability of 2 given output is greater than 50%, 60%, 709% ee. then the ‘outcome is: deemed likely). In other implementations, the 2 likelihood of an outcome occurring ata parilar node can be binary (eg. outcome likely to oceur/not ikely to occu). trinary (eg, outcome likely to occur in current session ‘outcome likely to oocur in a current or fature session, ‘outcome not likely to occur, oF ofa defined set. Stil futher, the likelihood ean also be associated with « confidence value ‘Accordingly, the prediction component 184 can use the sctvity information of the known users to develop model(s) to predict a likelihood of panicular outcome following an initiation event of a defined session, The models can further Indicate a confidence or cerinty with respect to the deter ‘ination of the likelihoods, with the confidence of cersinty fof the determination increasing as a length of the event sequence inereases, reflecting the user performing additional ‘activites during the session ‘Asan addition or variation, the prediction component 184 ‘ean develop models to be specifi to a categorization fend uscr. In particular, some examples provide that the prediction component 184 categorizes users in aecordance ‘with a predicted intent ofthe end user, where the predicted Jntent can he determined from historical information about the end user. In the context of an e-commerce site, the ‘categorization can reflect the end user's willingness oF propensity to perlonn a conversion event (eg, make a Purehase) when initiating an online sesso As an alternative or addition, the prediction component 154 implements one or more models to soore the User's ‘tention towards a particlar event (eg, conversion eveat, ‘where user makes a purchase). In such examples, the re diction component 184 categorizes the user based on the ‘determined score. For example, the propensity ofthe user (0 ‘convert (eg, make a purchase) can be categorized based on, possible ranges for the user's propensity score ‘While some examples such as described with steps (232)- (235) can be utilized to predict in-sssion outcomes, in Variations, the determined outcomes can be used to develop predictive models for predicting in-session outcomes for detected user-aetvities, imespective of historical informa tion or detected activity patter. To illustrate possible variations, such developed models can be used (0 predict ‘outcomes for unknown users (eg. users for whom detected ‘activity does not correlate 1o histriea oF profile informs tion). For example, the prediction model 184 can implement models on user record that reflect a sequence of activities and events, recorded in real-time, during a given session where the user accesses a website. In such examples, the 0 o 18 session initiates once the user acoess the website (when user activities are recorded), and no additional historical activity in formation is known about the user, According to examples, the system 100 can further moni tor the designated set of resources for activities of users, including users who are of an unknown class (238). For ser ofthe unknown elas, the system 100 may not be able to link a recorded identifier ofthe end wser's current session ‘with an identifier that is associated with any prior historical information about the end usee. For example, such unknown ‘users may’ correspond! 10 users Who have not signed-in to their aocount with the designated website, users who have no ‘account with the designated website, oF users Whom system 100 has not previously identified or recorded activity infor- ‘mation in connection with the designated website. Still further, users of the unknown class may include users who are not recognized by system 100 28a result of, for example, the end user accessing the website or designated resource of the enterprise using a channel or device where the user's ‘dentfer is unknown andior not persistent Tn some examples, the website connector 1028 can receive site trafic information from a process oF program that ris with the particular site. The real-time data store 134 can record activity information generated from the mioni- ‘ored site, with the stvity information reflecting an event, attributes ofthe event, ad one or more identifies forthe end ‘user who performed the activity of the event. The profiling component 130 can process the recorded activity informa tion for identifiers of the end user, to fink the end user with, {or example, prior session profiles andor historieal infor ‘mation about the end user. For users of the unknown class, the initiation ofa session (as defined by the site houndary} fsa fist encounter for the system, a a result of the end user being unknown. AddiGonaly, the system 100 can record activity information for users who are of () a known class, ‘seh as users for whom the recorded session profile can be Iinked to historical information, using persistent identifiers: andi (i) a parly unknown clas, including users associ- fated with recent session profiles (eg. session profiles which ‘ceurred in same day or same week), as a result of, for cxample, non-persistent session identifiers. For each user ofthe unknown class, te predictive models can be used to determine a likelihood of the end user performing a desired activity, based on an event sequence of the session profile, reflecting the activities ofthe end user during a current session (240). In examples, the desired activity can correspond 1 conversion events, such asthe end ‘user purehiosing. an item, subscribing to a service, or per {orming some other activity that is classified as a conversion event for the particular site. The likelihood determination can reflect a probability ander an outcome, as well as a confidence level ofthe determination. Ta examples, the prediction component 184 can repeat cedly make the determination daring the given user session {or outeomes within or outside ofthe piven use session. For ‘example, the likolihood determination ofthe end user per- forming the conversion event can be initiated after a mini- ‘mum numberof events are recorded in the end user session, ‘and the detemnination can be updated repeatedly with detee- ‘ion ofaditonal events which update the end user's session profle In some variations, the likelihood determination can be repeated afer individual events are recorded to reflect the end user's session aetvities, until the confidence value ofthe ‘determination reaches a minimam threshold. Sil farther, in ther variations, the likelihood determination canbe repented until a detemnination is made that the end user is Tikely to not perform a conversion event. TUS 12,348,591 B2 19 ‘of such likelihood can be reflected by, for example, 3 probability value, reflecting the kelibood determination of the end user performing the conversion event, that is less than dhe minimum thresld In examples, the intervention component 186 uses the likelihood determination of the end user performing the ‘conversion event to determine whether the end user should receive an intervention In examples, ifthe likelihood deter mination of the prediction component 154 is that the end ‘user will perform the conversion event, then the interventi ‘component 156 identifies the end user to not weeive any Jnterveation, On the other band, if the likelihood detemni- natin is thatthe end user will aot perform the conversion ‘event, then the intervention component 186 identifies the ‘endl user asa candidate for receiving an intervention In some implementations, the intervention component 156 identifies individual users to receive or not receive Jmerventions, For example, those users that are to receive Jnerventions may be identified to the enterprise sys based on the respective user session identifier, and logic ‘exceuting on enterprise system ean select and deliver in Jnervention to that user before the end user ends his or her session. As deseribed, however, ithe end use is determined 'o likely perfomn the conversion event, no identification of the end User session profile may be made, in order 10 preclude the enterprise system from setting that user and ‘nervention In this way, interventions are utilized on an as rnosced basis, rather than globally implemented for users of the website ‘Sill iter, in some examples, the prediction component 184 can determine the user intent, and further determine @ time interval, action, or series of actions which the user is likely to perform before the user performs a particular action associated with a particular intet. In such examples, the ‘nervention component 156 can select an intervention for the user tha is predited to reduce the time until when the user performs the desired action, Thus, for example, the prediction component 184 may determine that a user is likely to make a purchase, but the user may first place the item in a shopping cart, and then perform the action of ‘completing the purchase ata later time. In such cases, the selected intervention may be selected to cause the user 10 ‘complete the purchase shortly after the item is placed inthe ‘online shopping ear. FIG. 3A through FIG. 3C illustrate data structures that represent detected activities of an end user. FIG. 4A-4C ‘state data structures for encoding a sequence of events Tor an end user FIG. §ilustates a data siiture to map @ detected event for an end user to other types of information, In describing examples of FIG. 3A through FIG. 3C, FIG. 44 through FIG. 4C, and FIG. , ilustrate examples of an ‘encoded data stream 121, suc as deseribed with FIG. 14, and FIG. IB. Accordingly in describing examples of FIG. 3A through FIG, 3C, FIG. 48 through FIG. 4C, and FIG. § reference may be made to elements of FIG. 14 and FIG. 13 to illustrate examples, anetionality andlor to provide con- ‘With reference to FIG. 3A, the data streture 300 repre= sents an encoded sequence 310 of events (e-., encoded ‘event stream 121), where the detevied events can comelate ‘or represent detected setviies of an end user, The encoded sequence 310 ean include event identtirs 31 (et, letters) sequenced by time of occurrence. The event identifiers 311 ‘can represent an event type or category. In examples, the ‘encod sequence 310 can be generated by, for example, the system 100 utilizing connectors 102 fo interface witha user 0 o operated device 98 (ex, ee.) such asin the ease of the end user operating an app oF browser Tn examples, the encoded sequence 310 can include event identifiers 311, representing events identified across one or ‘multiple chanics, where cach channel represents a particu- Jar type of connector 192 andior interface. Accordingly, in some variation, the encoded sequence 310 fora given end user ean include events detected across multiple eiannels oF end user devices (ex, event detected on user mobile devie (ormobile app) and on user desktop or browser). To generate the encoded event sireum 310 for a given user across ‘multiple channels, some varations provide tha system 100 rmatehes user identifiers (eg, account identifiers) used by end user when performing corresponding activities) andlor ther types of user identifiers (eg, machine identities, mobile device phone number, et). ‘Funtber, the sequence of event can identify or otherwise be associated with contextual information, such as the location fof the end user when the event oceurred andor the channel from which the event was detected. The event data stream 121 can further be contextualized and personalized for the end user, so as to be representative ofa subject's enterpise- relevant activities or activity pattems. Such detected activi ties or activity patterns ean, for example, enable highl customized iaterations with customers, As deseribed With ther examples, the encoded event stream 121 canbe analyzed for patterns, using, for example, machine learning, ‘deep leaming and other learning techniques, ‘With further reference to an example of FIG. 3A, the encoded event stream 310 can be processed and analyzed (ex, such as by leaming sub-system 10) to predict, for ‘example, the user's intent with in conjunction witha curent activity ofthe user. For a panicular event, the encoded event sreaim 310 can be processed or analyzed to determine one ‘or more portions M2 which are relevaat (or most relevant) toa panicular context (e.g. curent event detected for use). Depending on implementation, a portion of the encoded event stream 310 can be determined relevant based on, for ‘example, a recency oF Window of time peecoding a current context (eg., current event detected for user), where the relevant window of time can be configured or otherwise selected based on, for example, the particular context andior predictive model PIG. 3B and FIG, 3C illustrate examples of data structures 320, 330 to represent individual events. As shown, each of ‘the event identifiers 11 ofthe encoded data stream 310 can represent a detected event (eg., viewing a product in an e-commerce site; searching for a product in an e-commerce site). Each event ideattier S11 may be associated with a set of contextual information pertaining to thit event. In some ‘examples, the contextual information may include paramet- fic information which ean be defined or olherwise provided by an enterprise operator (e.g. publisher of website), FIG. 44-4C illustrate example data strictures to represent fn encoded “click stream’ of an end user. With reference t0 FIG. 4A, an encoded click stream 400 is an example of an encoxled event stream 310 (eg. se FIG. 3). Inexamples, the encoded click stream 400 can comprise set of multiple records, with each record representing an event ofthe user ‘click’ (eg, user operating input device on a computing dovice to selet an interactive component on a website). By ‘way of example, system 100 can encode a sequence of clck-events, with each being represented by a eorrespond- ing event record 113. Tn some examples, at least some of the contextual infor: tion cam be parametrized and associated with correspond.US 12,348,591 B2 2 ing event ecords. In some examples, the event encoding 120 ‘can collect event data 111 in real-time, and apply a code to Specie parameters associated withthe event data With reference to an example of FIG. 4B, a data structure 410 provides a voctorized dataset 412, As deseribod with an ‘example of FIG. 1A, the event encoding 120 can use the ‘enterprise data to vectorize the evens. For example, certain, ‘events can he associated with specific products that were the subject of the event, and the vectorization can identify jnformation about a particular event type or characteristic for certain events ofthe user's encoded click steam 400, Ia ‘examples, the events of a given sequence may be parsed iato subcomponents (eg, sbsequences), In an example of FIG. 4B, the vectorization can be Based on a common atibute oF ‘event characteristic For example, individual vectors may be enerated (0 represent product identifies andioe sales price ‘with respect t0 coresponding events. With reference to an example of FIG. 4C, a data structure 420 may provide pattems or other indicators ofa vectorized data set for given use. In examples, the system 100 can, process the encoded event streams (eg. using processes of the learning sub-system 180) to extract pattems or other indicators of an end user intent of outcome. The learaing sub-system 180 may, for example, develop and apply mod- ‘lst predict user intent of behavior hased at least in part on the extracted pattems. FIG. 5 illostrates an example data steucture $00 to map an ‘event (or sequence of events) in an encoded event steam (0 f contextual information of the enterprise (eg, products, ‘category of products, et.) An example data sieeture $00 ‘ean be based on a designed schema that accounts for the ‘enterprise customer base and objectives. TIG, 6 illustrates a distributed network computer system for detecting real-time events, aeeording to one oF more ‘examples. As shown, @ network computer system 600 ‘includes one or more network computers 610 that eomn- nicate cross one of more networks with end wser devices (represented in FIG. 6 by end user device 620), In an ‘cxtmple shown, the network computer 610 can correspond ‘o, for example, a server, or combination of servers, and the ‘end user device 620 can correspond to, for example, 2 desktop computer or mobile device (eg, tablet, cellu telephony and/or messaging device, wearable device, et.) ‘According to examples, the end user device 620 can implement an event analysis subsystem 680, using. for ‘eximple, a dedicated service application, or a combination ‘of applications and/or programmatic functionality. The ‘event sub-system 650 can execute on the en! user device 620 to communicate with the network computer 610, ia ‘onde to receive and implement functionality for detecting tnd responding t real-time events on the end user device 620, As described, the functionality provided tothe end user device 620 can include decision logic 630, whieh can be ‘generated from machine leaming processes of the network ‘computer 610, for deployment to end user devices (as represented by end user devices 620), to enable end user ‘devices to use machine leaning in detecting and responding to rwal-time events According 10 examples, the network computer 610 includes a device interioce 612, taining processes 614 and ‘one or more machine Teaming models 616 (including 3 sequence invariant model 616A). The device interface 612 ‘ca operate to provide functionality, inching instructions and data sos, tothe end user device 620. Additionally, the ‘device interface 612 can receive information from the end twser device 620, including feedback which the network ‘computer 610 cat use for training and updating purposes 0 o n The machine lea models 616 can include one or smuliple machine Iearaed models 616 that sre trsined to enero an output tat is ingicaive of a wsee intent or Interest, in response 10 receiving input in the form of sequenced events tht corespond to detected se activites, ‘The network computer 610 may store an event brary 60 tat includes data to define mamerovs predefined events. In capes, the evento the event library 68 can correspond {individually oF in combination with other events, o aetv tis that can be detected by the end user device 620 “The event analysis subsystem 680 can be implemented as, forexanpl, an application tit a user can lunch onthe ed ser device 620. According (0 an aspect, the event aulysis subsystem 680 can be dedicat, oF otherwise configured for use with the network computer 610. When ‘inning on the end ser device 620, componcats provided by the event analysis subsystem 680 may include a service Interface 682, an activity monitoring component 684, an vent processing component 656 anda determination com- oneal 68. The service interface 682 can communicule over fone oF more networks ith the device interface 612, t0 frchange data seis a8 describe by various examples Ts examples the network comspr 610 comicats st least a portion ofthe event library 68 to the end user deviee 1620 via the device interface 612. The end user device 620 cam store an event collection 688, coresponding to a potion fr complete version ofthe event library 608. The activity monitor 684 can iatctace with and monitor cone or more device esoures, sing one or more program: ‘atc esonrees 648 andor senso resources 647. The peo zrammatic resources 64S can inch, for example hie arty opplicatons, plugins. or services avalable trough {he eal weer device 620. Iy way of example, the program: ‘atc resource can ince a bowser interface tht enables the activity monitor 684 to detect certain pes of browser aetivity (esearch for product, product Viewing, placing product in shopping car, purchasing prev, et). The Sensor resources 647 can incl, for example, an interface tothe device's satelite roceiver, to obtain a eure location ‘ofthe end user deve 620.As an addition or alternative, the Sensor resources 647 can include an interlace to a motion Sensor of the end user device 620 (eq., accelerometer, yroscoe), an environmental seas of the em user device (Gg, barometer, temperature) andor interfaces wo a camera ‘emierophoneof ti end ser device. In some examples, the fetivity monitor 684 accesses oF here utilizes logic for processing output fom specie sensor resoures 647. Like- ‘wise the ativity monitor 6S4 ean acess the lize logic for Aetecting and dterniaing information abou diferent types ‘of programmatic activities, such s ecommerce ativitis In {his way, he activity monitor 684 can interface and monitor programmatic andor sensor resources 648, 647 10 detect Gevice usage that matches one ce more definitions for comresponding events ofthe event collction 648 Ta examples, the activity monitor 684 detects spcitic types of ser activities that are monitored through the [roarammutic resource 648 and/or the sensor resource 647 When the setivity monitor 684 detects a elevant activity the aedvity monitor 654 ean also detemune information Ut is relevant tothe stivity. Far example, when ser porchases fn item onlin, the feevant information can inclu, for capt, () the numb of page views that wore detect From the use in the network session, before the User made the purchase, (ji) the numberof proets the user viewed, Gian dente ofthe pret (€g, SKU}, (s) a price the ‘scr paid for the product andor (+) information about any promotion the user sed when making the purcse. TheUS 12,348,591 B2 23 ‘setivity monitor 684 can communicate detected activity, ‘including predefined relevant information about such events to the event processing component 656. The event processing componeat 656 can implement various tasks in connection with detected evens. Por ‘example, the event processing component 656 can structure the information about detected events, by generating @ recon! 657 of the detected activity. In processing the dected activity, the event processing component 656 can, match the activity to one of more ofthe events as defined by the event collection 648, Whe a detected activity 655 is tected as an event (oF events) a provided by the event ‘collection data 648, the event processing component 656 a, ‘associate @ code value 10 the event, and further populate a ‘current event record 687 to reflect the code value and other information determined from the detecod activity. In some ‘examples, the code value of the event can rellct a category designation of the event. Additionally, the current event record 687 can inclide stributes which elect specific Jnformation that was determined from the detected event. Additionally, the current event record 687 can include a timestamp to refleet when the corresponding user activity ‘cour The event processing component 656 may maintsin an ‘event record store 660 to receive and store the curren event rococd 657. Additionally, the event record store 660 may include a collection of historial event records 651, with ‘each historial event record 681 providing information about ‘8 proviously detected event of the respective user. Tn ‘cxamples, the records 651 can include histrial information| regarding detected events of the user. Still further, the historical records 6S1 of the record store can inclde oF reflect an encoded series of events for locally detected ‘events (corresponding to user activities that are defined for those events). Additionally, the record store 660 can rellest the detected events asa series, on sequences, based on the timestamps that are associated with the individual records “Accolng to examples, the event processing component 656 can retrieve historical event records 651 in connection swith the event processing component 656 encoding and analyzing the current event record 687. The event processing ‘component 656 can update the record store 660 swith the ‘curent event record, aad Veetorzation component 688 ca Vestorize the updated record store 660 fo generate a vector representation 669 that is based at least in part on the ‘encoded event seam of the ser, The event processing ‘component 686 can further analyze the vector representation {669 in connoction with the current event record, in order to make a predictive determination of the user's intent of Jnerest. In some examples, the event processing component 656 can, for example, parse events, as recorded! bythe record store 660, to identify sequences of events which share a charecteristc of relevance (e-., by time, by typeof activity ‘of user performed, in relation to particular product oF service, ce.) The event processing component 656 can also proup events by criteria of relevance and/or time, so as 10 ‘identify relevant events in accondance with a series of Seaquence. In this way, the record store 660 itself may record sn entire series of events according toa timeline of occur Fenoe, while separately identifying sequences oF series of ‘events in accordance wth, for example, criteria of relevance, In examples, the separately identified event sequences or seros ean he analyzed collectively, for decision making, In some examples, the event processing component 686 applies instrctions or logie received by the nctwork com- puter 610 to determine event sequences or series. The 0 o 2 determinations made on the end user device can rele, for ccxample, a patter or trend with respect to a user intent oF interest, In some examples, the vetorization component 688 can generate vector representations 669 of event sequences, based on for example a determination that a set of events are relevant to one another, andr in response toa current user activity, The record store 660 may also store vectorizd torms of individual events, as well f vectorl representation of identified event sequences or groupe, where the vectral representation applies tothe collective group. "According 10 examples, the event analysis sub-system 650 may also include a determination component 670, whieh responds to detected events by determining an pro” priate outcome for current or recently detected event. The ‘etermination of the outcome ean be based on, for example analysis of prior events in combination with a eurent event. As described by some examples the analysis of prior events fan include analysis of multiple evens, in sequence or clusters, where the events are deemed relevant to one father, sich as when the events share a particular eharae- teristic (eg., ecommerce activity) ‘As an addition or alternative, the determination compo- nent 670 can analyze prior events to detect pater or tends ‘whieh may be probative in predicting, for example, a user intent or interest a5 (0 a particular subject (ex. item of ‘commerce. In examples, functionality for analyzing events in sequence can be integrated with the event analysis ub- system 680, such a in the form ofa program that rans with the event analysis sub-system 680, In some examples, the

You might also like

• United States Patent (10) Patent No.: US 8.458,168 B2

  

  No ratings yet

  United States Patent (10) Patent No.: US 8.458,168 B2

  20 pages
• Us 9766769

  

  No ratings yet

  Us 9766769

  25 pages
• Social Footprint-Based Credit Scoring

  

  No ratings yet

  Social Footprint-Based Credit Scoring

  27 pages
• Us 11080336

  

  No ratings yet

  Us 11080336

  468 pages
• United States: Reissued Patent

  

  No ratings yet

  United States: Reissued Patent

  17 pages
• Recunoas Terea Exercit Iilor Fizice: Universitatea Babes - Bolyai, Cluj-Napoca

  

  No ratings yet

  Recunoas Terea Exercit Iilor Fizice: Universitatea Babes - Bolyai, Cluj-Napoca

  70 pages
• Systems and Methods For Generating Expanded User Segments

  

  No ratings yet

  Systems and Methods For Generating Expanded User Segments

  20 pages
• Determining User Engagement

  

  No ratings yet

  Determining User Engagement

  22 pages
• BTP Research Internship Final Report

  

  No ratings yet

  BTP Research Internship Final Report

  21 pages
• Plaintiff's Opening Brief As Filed

  

  No ratings yet

  Plaintiff's Opening Brief As Filed

  33 pages
• AI Applications & Search Strategies

  

  No ratings yet

  AI Applications & Search Strategies

  9 pages
• Neral Introduction

  

  No ratings yet

  Neral Introduction

  35 pages
• 1701 07274

  

  No ratings yet

  1701 07274

  70 pages
• United States: (12) Patent Application Publication (10) Pub. No.: US 2013/0318015 A1

  

  No ratings yet

  United States: (12) Patent Application Publication (10) Pub. No.: US 2013/0318015 A1

  26 pages
• Vaagdevi Ngin Eering Colege: Ard Nakig

  

  No ratings yet

  Vaagdevi Ngin Eering Colege: Ard Nakig

  8 pages
• Kairouz, McMahan Et Al 2019 - Advances and Open Problems in Federated Learning

  

  No ratings yet

  Kairouz, McMahan Et Al 2019 - Advances and Open Problems in Federated Learning

  121 pages
• Advances and Open Problems in Federated Learning

  

  No ratings yet

  Advances and Open Problems in Federated Learning

  105 pages
• Ulllted States Patent (10) Patent N0.: US 8,490,074 B2

  

  No ratings yet

  Ulllted States Patent (10) Patent N0.: US 8,490,074 B2

  96 pages
• A Flexible Framework For Rogue Access Point Detection: Ricardo Jorge Esteves Gonçalves

  

  No ratings yet

  A Flexible Framework For Rogue Access Point Detection: Ricardo Jorge Esteves Gonçalves

  75 pages
• Patent

  

  No ratings yet

  Patent

  16 pages
• 1.1 Motivation

  

  No ratings yet

  1.1 Motivation

  65 pages
• US11810080

  

  No ratings yet

  US11810080

  31 pages
• Prediction of Cyber Attacks Using Data Science Technique

  

  No ratings yet

  Prediction of Cyber Attacks Using Data Science Technique

  11 pages
• Batch Reinforcement Learning Approach Using Recursive Feature Elimination For Network Intrusion Detection

  

  No ratings yet

  Batch Reinforcement Learning Approach Using Recursive Feature Elimination For Network Intrusion Detection

  16 pages
• Presentation 4

  

  No ratings yet

  Presentation 4

  10 pages
• IntentEmail PROJECT REPORT

  

  100% (1)

  IntentEmail PROJECT REPORT

  58 pages
• United States Patent (10) Patent N0.2 US 8,549,154 B2

  

  No ratings yet

  United States Patent (10) Patent N0.2 US 8,549,154 B2

  28 pages
• Us 20130254838

  

  No ratings yet

  Us 20130254838

  31 pages
• Wo2020056331 Pamph 20200319 4096

  

  100% (2)

  Wo2020056331 Pamph 20200319 4096

  53 pages
• US20180357915A1

  

  No ratings yet

  US20180357915A1

  14 pages
• US9009192

  

  No ratings yet

  US9009192

  26 pages
• MSR Thesis Document

  

  No ratings yet

  MSR Thesis Document

  43 pages
• Disaggregated Memory Systems

  

  No ratings yet

  Disaggregated Memory Systems

  29 pages
• (Chapman & Hall - CRC Computer & Information Science Series) Miguel A. Labrador, Oscar D. Lara Yejas - Human Activity Recognition - Using Wearable Sensors and Smartphones-Chapman and Hall - CRC (2013)

  

  No ratings yet

  (Chapman & Hall - CRC Computer & Information Science Series) Miguel A. Labrador, Oscar D. Lara Yejas - Human Activity Recognition - Using Wearable Sensors and Smartphones-Chapman and Hall - CRC (2013)

  206 pages
• Ad Hoc Networks Jyoti Prakash Singh PDF Download

  

  No ratings yet

  Ad Hoc Networks Jyoti Prakash Singh PDF Download

  181 pages
• Generative - Artificial - Intelligence - For - Wireless - U22EC076 - Seminar - Report (1) - 3-44

  

  No ratings yet

  Generative - Artificial - Intelligence - For - Wireless - U22EC076 - Seminar - Report (1) - 3-44

  42 pages
• United States Patent: Dutta Et A1. (10) Patent N0.: (45) Date of Patent

  

  No ratings yet

  United States Patent: Dutta Et A1. (10) Patent N0.: (45) Date of Patent

  26 pages
• Clear Force Us 20170061345 A 1

  

  No ratings yet

  Clear Force Us 20170061345 A 1

  47 pages
• Detecting The Unseen: A Study of CNNs For Real Time Abnormal Event Detection

  

  No ratings yet

  Detecting The Unseen: A Study of CNNs For Real Time Abnormal Event Detection

  12 pages
• Research Paper 3

  

  No ratings yet

  Research Paper 3

  29 pages
• United States Patent (10) Patent No.: US 8,051,024 B1

  

  No ratings yet

  United States Patent (10) Patent No.: US 8,051,024 B1

  17 pages
• Thesis PDF

  

  No ratings yet

  Thesis PDF

  114 pages
• Mohak RR

  

  No ratings yet

  Mohak RR

  57 pages
• US9002889

  

  No ratings yet

  US9002889

  22 pages
• Monitoring of Resource

  

  No ratings yet

  Monitoring of Resource

  25 pages
• Sample Report77

  

  No ratings yet

  Sample Report77

  35 pages
• Information Security

  

  No ratings yet

  Information Security

  337 pages
• Reply Brief On Claim Construction As Filed

  

  No ratings yet

  Reply Brief On Claim Construction As Filed

  32 pages
• SKN Report

  

  No ratings yet

  SKN Report

  26 pages
• US20230394938A1

  

  No ratings yet

  US20230394938A1

  31 pages
• Cyberspace Monitoring Using AI and Graph Theoretic Tools

  

  No ratings yet

  Cyberspace Monitoring Using AI and Graph Theoretic Tools

  38 pages
• Mathematics

  

  No ratings yet

  Mathematics

  84 pages
• Business Contact Management System

  

  No ratings yet

  Business Contact Management System

  30 pages
• US20230376390A1

  

  No ratings yet

  US20230376390A1

  43 pages
• AI & Deep Learning Seminar Report

  

  100% (1)

  AI & Deep Learning Seminar Report

  42 pages
• MSC Chudasma PH 2020

  

  No ratings yet

  MSC Chudasma PH 2020

  74 pages
• Hsi1501 Finals Notes Merged

  

  No ratings yet

  Hsi1501 Finals Notes Merged

  145 pages