See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/338458563

A Survey of Different Approaches of Machine Learning in Healthcare

Management System

Article · January 2019

DOI: 10.35444/IJANA.2019.11032

CITATIONS READS

9 1,854

4 authors, including:

Krishan Kumar Goyal Aejaz Hassan

Raja Balwant Singh Management Technical Campus, AGRA Bhagwant University
28 PUBLICATIONS 66 CITATIONS 4 PUBLICATIONS 9 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Krishan Kumar Goyal on 16 November 2020.

The user has requested enhancement of the downloaded file.

Int. J. Advanced Networking and Applications 4270
Volume: 11 Issue: 03 Pages: 4270-4276 (2019) ISSN: 0975-0290

A Survey of Different Approaches of Machine

Learning in Healthcare Management System
Dr. Krishan Kumar Goyal
Associate professor,
Faculty of Computer Application,
Raja Balwant Singh Management Technical Campus, Agra.
E-Mail Id: kkgoyal@gmail.com
Aejaz Hassan Paray
Ph.D. Scholar Department of Computer Science, Bhagwant University Ajmer Rajasthan,
Aejaz50@gmail.com.
-------------------------------------------------------------------ABSTRACT---------------------------------------------------------------
Machine Learning has become an important tool in day to day life or we can say it’s a powerful tool in most of the
fields which we want to automate. Machine Learning is used to develop algorithms which can learn from the data,
which is either labeled, unlabeled or learn from the environment. Machine Learning is used in most of the fields
and Especially in health care sector it takes much more benefits through proper decision and prediction
techniques. Machine Learning in health care is a scientific study, so we have to store, retrieve and proper use of
information, data and provide knowledge to the problems facing in the healthcare sector and also knowledge for
the proper decision making. Due to these technologies there is a huge development in health care sectors over the
years. For analysis of medical data, medical experts use the machine learning tools and techniques to identify the
risks and to provide proper diagnosis and treatment. The paper is based on survey in terms of health care
management system using different machine learning approaches and techniques.

Keywords: Machine learning, Healthcare Management system, Artificial intelligence.

----------------------------------------------------------------------------------------------------------------------------- --------------------
Date of Submission: Oct 10, 2019 Date of Acceptance: Dec 19, 2019
----------------------------------------------------------------------------------------------------------------------------- --------------------
I. Introduction machine learning methods/approaches we identify the
patients and provide proper care management. Patients can
M achine learning is a branch of artificial intelligence be identified (eg on the basis of patient’s characteristics,
which aims to create intelligent system which do human risks, beliefs and genetic profile. Machine learning and
like jobs by learning from a lot of relevant data. According Artificial Intelligence changing healthcare by using
to Nithya and Ilango[20], a computer program is said to predictive analytics for proper treatment and decision
learn from experiences E with respect to some classes of without any risk factor. Machine learning and Artificial
task T and performance measure p, if its performance at intelligence have been recently covers all the fields of
task t, as measured by p improves with experience E. we healthcare services even though computers will
can say a general algorithm takes the data and can learn completely replace doctors, nurses and modern
automatically itself. In other words, we can say machine technology. Recently Google develop a machine learning
learning is computer ability to learn without being algorithm to identify the cancerous tumors and deep
explicitly programming. Instead of writing a program by learning is used to identify the skin cancer. Machine
hand for each specific task, we can collect lot of examples Learning used to analyze the Enormous data and
or experiences(data) belonging to the problem, based on suggested outcomes, Risk factors and allocate the precise
the data apply some classifiers do some processes and resources.
produces the output. The healthcare sector is being
transformed by the ability to record massive amounts of This paper serves many applications and techniques of
information about individual patients, the enormous machine learning in healthcare system. Rest of the paper is
volume of data being collected is impossible for human designed in following manner. In Section-II Supervised
beings to analyze. Machine learning provides a way to and Unsupervised learning techniques which are used in
automatically find patterns and reason about data, which different health care management system are described.
enables healthcare professionals to move to personalized Section-III contains Reinforcement Learning Techniques
care known as precision medicine. The purpose of the in health care management system. Section-IV covers
Classification model is to determine a label or category – some of AI techniques used in health care management
it is either one thing or another. We train the model using a system. Finally Section-V concludes the paper.
set of labeled data. Machine learning is one of the
emerging approaches in the health sector. As increasing
population of patients the machine learning and Artificial
intelligence will allow health care systems to make care
more efficiently and appropriately .With the help of
Int. J. Advanced Networking and Applications 4271
Volume: 11 Issue: 03 Pages: 4270-4276 (2019) ISSN: 0975-0290

II. Supervised and Unsupervised Machine (non-readmitted). In the second step, features contributing
Learning techniques in health care to predictive risk from independent models were combined
management system. into a composite model using a correlation-based feature
selection (CFS) method. Data extracted from EMR: I)
As per Chuang et al. [21] there are seven well known Diagnoses codes were extracted from Mount Sinai Data
supervised learning techniques namely C4.5, C5.0, Warehouse. II) Medications prescribed during the
CART(classification and regression tree), Logistic hospitalization were extracted from Mount Sinai Data
regression, SVM(support vector machine), Random forest, warehouse. Medication name, dosage, route of
Multivariate adaptive regression splines. The C4.5 and administration was obtained. III) Laboratory
CART are the two most commonly used decision tree- measurements. IV) Vital signs Pulse, respiration rate,
based learning techniques. These techniques are used to systolic blood pressure, heartbeats and temperature were
develop a predictive model that determines the length of compiled and captured in a MySQL database. The
stay of patient in the hospital. The case was divided into research on CAD(computer aided diagnosis) and
urgent operation and non-urgent operation group to intelligent patient management system using different
develop a prolonged length of stay prediction model using Machine learning and operations research techniques is
several supervised learning techniques. Critical factors are proposed by Florin Gorunescu [11]. CAD is used for
identified using the gain ratio technique contains medical automated diagnosis of different major diseases, such as
records, lab data of 896 clinical cases involving surgeries breast, pancreatic and lung cancer, heart attacks, diabetes
performed by general physicians. Supervised learning using different supervised machine learning techniques
technique is suitable for analyzing patient medical record (neural networks, support vector machines, Bayesian
in accurately predicting a prolonged length of stay. The decision, k-nearest neighbor, etc) An intelligent patient
proposed model supports a physician in making clinical management system develops an Application to patient
decision and communicating with medical team members, grouping using different unsupervised machine learning
patients and patient family. The system may also use for techniques(k-means clustering) and Application to hospital
making decision regarding whether patient require more bed management by optimal bed allocation, in accordance
clinical care (therefore improving patient safety). Based on with the population needs and quality medical service.
length of stay related literature, 119 independent variables One can model a hospital/clinical department by using
are collected from laboratory information system, nursing both queuing models and optimization tools (M/PH/c and
information system and from physician orders then Genetic Algorithm). Nguyen et al. [17] proposed a remote
selected variables are divided into six categories. health monitoring systems based on Internet of Things
Demographic, Medical history, vital signs, laboratory data, (IoT) technology. This paper explores the use of IoT-based
operation physician data and operation & nursing data. applications in medical field and proposes an IoT Tiered
WEKA(version 36.6) an open source data mining program Architecture (IoTTA) towards an approach for
was used to execute variables. WEKA was used to transforming sensor data into real-time clinical feedback.
construct a length of stay prediction model using several The IoT Tiered Architecture is based on the sensing,
classification techniques. Some limitations of the present sending, processing, storing, mining and machine learning.
study are, First, the data used in this study were related to Data mining involves discovering useful patterns from
general surgery patients from a single medical institution. large data sets and applying algorithms to the extraction of
Proceeding with the evaluations of clinical cases from hidden information. Its functions include classification,
other hospitals is critical for confirming the validity of the clustering, association analysis, time series analysis, and
model. Second, other potentially valuable features, such as outlier analysis. Machine learning techniques are very
data collected from the clinical pathway and supervised useful in healthcare applications as they enable managing
learning techniques, can be considered for use in the huge databases, learn from data and improve through
model. Finally, an interventional study may be initiated experience (supervised as well as unsupervised). Alhamid
based on our finding in practicing an expert system that et al. [7] proposed an open software framework that
may be conducted in future studies and try to predict provides long-term monitoring of the patient’s health
which patients will require urgent surgery. Shameer et al. situation by building a health monitoring system. That
[16] developed to predict readmission rates in heart failure detecting any abnormality in the patient’s health situation
patients. Readmission rates is a quality assessment used to and particularly during their time spent at home. The other
improve the quality of life index of patient population and main objective is the development of a convenient method
the quality of healthcare delivery. A machine learning is of analyzing and presenting the physiological data
used for the generation of composite model for performing collected from a Wireless Body Sensor Network (WBSN).
predictions using feature selected from individual model. BSN consists of wearable sensors that include Bluetooth
In his study a naïve bayes model is used for machine for communication. The paper also focus on the
learning. Bayesian model were created using feature monitoring of Activity of Daily Living (ADL), which
unique to each data element and validation of machine includes fall detection, walking, sitting, standing, and
learning models was performed with logistic regression. In lying down as well as physiological monitoring such as
the study we use multistep modeling strategy using Naïve heart rate using (simple activity recognition algorithm
bayes algorithm. In the first step, an individual model is based on accelerometers signals, K-nearest neighbor
created to classify the cases (readmitted) and controls (KNN) classifier with an applied Bayesian Network. The
Int. J. Advanced Networking and Applications 4272
Volume: 11 Issue: 03 Pages: 4270-4276 (2019) ISSN: 0975-0290

researchers Madhumita Kathuria and Sapna Gambhir[8] of smart environments. One such example is the so-called
focus on the E-health based on the Wireless Body Area Health Smart Home (HSH). They have the potential to
Network (WBAN). WABN becomes an integral provide e-health services to meet the needs of this growing
component of healthcare management system where a population. The machine learning methods and algorithms
patient needs to be monitors both inside and outside home (supervised and unsupervised) to effectively interpret
or hospital. The applications is categorized into: Retrieval, sensor data (low-level) and build new abstractions (high-
Alerting, Prediction, Suggestion and Reminders to health level) in order to understand this complex and variable
related information using machine learning techniques data on human behavior. The methods and techniques
such as Binary Decision Tree and Genetic Algorithm. have been widely used to recognize human behaviors,
Decision tree is used to classify the unseen instances from detect normal and abnormal situations and to predict
the training set of pre-labeled instance. GA is applied on future health conditions. We broadly classify techniques
training sets to generate optimized training data sets. It is into three categories: probabilistic and statistical
responsible for managing traffic flows at different levels techniques, computational intelligence techniques, and
so that both real time traffic and non-real time health knowledge-driven techniques.
related traffic get benefit. It is responsible to manage four 1. Behavior Recognition: Statistical Techniques (Hidden
kinds of works: Packet classification, queuing, scheduling Markov Models, Bayesian Network Naive Bayes,
and dropping. The researcher Lih-Jen Kau and Chih- Multiclass Logistic Regression) Computational
Sheng Chen[13] proposed a smart phone based monitoring Intelligence Techniques (Neural Networks Support
and rescue system. The smart phone should be android Vector Machine Decision tree (C4.5), Clustering)
based only. The various wearable sensors are used for Knowledge-Driven Techniques in (Rule-based ,Fuzzy
detecting patient if the accident occurs. The proposed logic)
system contains the triaxial accelerometer and gyroscope 2. Behavior Abnormality Detection: Statistical Techniques
sensors. The triaxial accelerometer and electronic compass (Gaussian Mixture Model, Hidden Markov Model)
is uses as input if the fall detection occurs on smart Computational Intelligence Techniques (Neural Networks,
phones. In order to process the input signal, system uses Support Vector Machine, Clustering ) Knowledge-Driven
the cascading classifier and support vector machine Techniques (Fuzzy logic, Ontologies )
(supervised learning) as detection tools. Once a fall 3. Prediction: Statistical Techniques (Hidden Markov
accident event is detected, the user’s position can be Models Computational Intelligence Techniques) Neural
acquired by the global positioning system (GPS) or the Networks (Support Vector Machine, Data Mining
assisted GPS (A-GPS), and sent to the rescue center via Techniques) Knowledge-Driven Techniques (Fuzzy logic).
the 3G communication network so that the user can get Mazhar et al. [18] developed a system using machine
medical help immediately. He Jian and Hu Chen [24] learning algorithms to better care and to ensure better
proposed a system that detects the fall detection of elderly. facilities to the inpatients to predict Spinal Cord Injured
The system consists of custom vast and smart phone. The (SCI) patient’s length of stay. If we can predict accurate
custom vast contains wearable motion detection sensor length of stay, patients do not have to leave in between the
integrated with tri-axial accelerometer, gyroscope and treatment without medical advice. This study proposes the
Bluetooth. The tri-axial accelerometer, gyroscope is used use of multiple linear regressions. which will help better
to measure the reluctant acceleration and angular velocity patient care in the future and help the hospital for the
that vary real-timely, and then make up stream data. better resource management. Baker et al. [19] proposed
Bluetooth is introduced to receive the stream data from the web portal to support a patient healthcare decision.
sensors, Bluetooth is used to send the data to the smart The portal is powered by semantics open source software
phone. After getting the data the individual is falling or not (SEMOSS) and end-to-end analytic tool. Multiple linear
based on the k-NN algorithm (supervised machine regression and random forest analysis is used in order to
learning algorithm). The phone can make a call or send a determine whether a patient would recommend a hospital
message with GPS position to a healthcare center or on the basis of patient experience. The portal allows
family member as soon as it detects a fall. This system can patient to search for and compare doctors and hospitals
provide a timely warning that a fall has occurred. The based on their personal needs. In this study the doctor’s
software includes the on-chip program running on the data used for analysis was obtained from the virgina board
sensor board and the fall detection program app of medicine website. The data set includes 20 attributes for
downloaded onto the mobile smart phone. Mshali et al. 40,0000 doctors attributes includes Name, Number of
[22] surveyed about health monitoring system in smart years in experience, sex etc. The distance is calculated
environment. The aim of HMS is to not only reduce costs using Google maps API. Hospital data include
but to also provide timely e-health services to individuals communicate with doctors, communication with nurses,
wishing to maintain their independence. The most responsiveness of hospital staff, pain management,
important functions and services offered by HMS for communication about medicines, discharge information,
monitoring and detecting human behavior including its cleanliness of hospital environment etc. Patient can check
concepts, approaches, and processing techniques. The the doctor’s data in their area and see the visualization of
HMS with intelligent technologies, such as sensors practitioners clustered by specialty. This type of graphic
(Personal Sensor Network, Body Sensor Network, allows the patient to see how much choices they have
Multimedia Devices) have resulted in a rapid emergence when choosing a doctor of a certain specialty. The
Int. J. Advanced Networking and Applications 4273
Volume: 11 Issue: 03 Pages: 4270-4276 (2019) ISSN: 0975-0290

Mockup of the portal includes a home screen, search page, intervals, such as when varying treatments can be
and various methods for viewing search results and administered. Finally, more hospitals are using electronic
visualization. Methods that are used for modeling a medical record systems to gather large amounts of patient
hospital data is multiple linear regression. Random forest data. While using this data will open the door for the
analysis of hospital group. increased application of MDPs to medical treatment
problems. Hauskrecht and Fraser [1] proposed a method of
III. Reinforcement learning in health care Markov decision processes viz partially observable
management system. Markov decision processes for the treatment of ischemic
heart disease. A disease caused by an imbalance between
Niyato et al. [5] proposed A Remote patient monitoring the supply and demand of oxygen to the heart. The
system in terms of e-health service A patient-attached proposed system uses methods for the treatment of
monitoring device with a heterogeneous wireless ischemic heart disease, characterized by hidden disease
transceiver (WiMAX based WMAN and WiFi-based states, investigative and treatment procedures, and
WLAN technologies) collects bio signal data from the temporal cost and their outcomes. Khianjoom and
sensors and transmits the data through the radio access Wipawee [9] Monitoring the patients by transferring the
network (RAN) to the eHealth service provider. The diagnostic information if the patient fall. They presents an
proposed architecture along with the reinforcement efficient, adaptive, distributed routing mechanism using
machine learning techniques (stochastic programming Anycast Q-routing to route information to the nearest sink.
Problem, constrained Markov decision process (CMDP)) Anycast has been used in wired networks and mobile ad
will be useful for an eHealth service provider to minimize hoc networks, to select the nearest server, service
the service cost while maintaining the quality-of-service identification, improve system reliability and policy
(QoS) requirements for remote and mobile patient routing. Q-routing is based on Q-learning which is a well-
monitoring. known Reinforcement Learning method. RL is a machine
Due to the mobility of the patients, a stochastic learning technique which aims at finding the optimal
programming problem has been formulated to obtain the action to perform at a given state of the dynamic
optimal number of reserved connections and minimize the environment. The proposed scheme requires low
connection cost and satisfy the delay requirements. A communication overhead and fast path search time to
constrained Markov decision process formulation has been discover optimal paths to the nearest sink. Gaweda et al.
used to obtain the optimal transmission scheduling [2] proposed the system which provides an optimal
decision. treatment to the patient of anemia using reinforcement
learning approach. The learning for the optimal treatment
occurs in the form of immediate improvements in the drug
dosing policy due to the experience gained by observing
i.e. off -policy using Q-learning rather than on-policy that
is learning form processing. Q-learning is capable of
performing adequate anemia treatment in real time. To
classify the different types of patients, a Takagi-Sugeno
fuzzy model was first built on basis of available patient
data.

IV. Some AI techniques used in the health care

management system.
Hurwitz et al. [10] proposed cognitive computing, which
Fig 1. A mobile patient can use different types of wireless is combination of cognitive science and computer science.
technologies (e.g., WiMAXbased WMAN and WiFi-based cognitive computing represents self-learning system that
WLAN technologies) to transfer monitored bio signal data utilizes machine learning model to mimic the way brain
to the healthcare center. works. A cognitive computing is used by combining
technologies such as machine learning, artificial
Gaweda et al. [3] proposed the reinforcement machine intelligence and natural language processing, that helps
learning methods for the treatment of various health relate healthcare professionals to learn from patterns which
problems in terms of markov decision processes. MDP is means healthcare organizations to solve some of their
used to solve the stochastic and dynamic treatment most challenging problems. For example, The University
decisions. An MDP binds previous, current, and future of Iowa Hospitals and Clinics has identified patterns in a
system decisions through the proper definition of system population of surgical patients that help to improve both
states. In order to provide the optimal decision in terms of quality and performance in surgery. The hospital has
health care MDP provides the various methods like Finite- modeled data from hospital readmission, surgical site
horizon MDPs, Infinite-horizon MDPs (These two model infections, and other hospital acquired infections. The
provides an exact description of the system), Semi- model enables physicians to predict which patients are
Markov decision processes is used In health care and other most at risk for acquiring a surgical site infection while
applications, decisions may occur over continuous time
Int. J. Advanced Networking and Applications 4274
Volume: 11 Issue: 03 Pages: 4270-4276 (2019) ISSN: 0975-0290

they are still in the operating room and corrective actions etc. 2. Income Items which contains Total budget, Basic
can be taken. Cognitive system enables healthcare salary expense, Social right expense, Medical device
organization to gain more value from data. Gaining more expense, etc. and 3. Expense Items contains Medication
value from data is a multifaceted process that requires both income, Inpatient medical implementation income,
technology and human knowledge. Getting the data right Outpatient laboratory income, Social security institution
is more important. The relevant data needs to be accurate, income, etc. is used. The provinces where hospitals are
trusted, consistent, and available for access. However, located are also classified. A genetic algorithm is
having accurate data is only the baseline for improving developed to generate fuzzy linguistic summaries. In this
health outcomes for patients. Physicians need the skill and system, the parameters of genetic algorithms were
experience to make sense out of what is often a complex established with trial and error for the presented
set of symptoms and diagnostic tests. They need to operational and financial healthcare dataset. The proposed
internalize best practices that enable them to ask the right approach is implemented in Java environment .linguistic
questions and listen for answers from the patient. A summarization is meant as a process of a comprehensive
cognitive system helps in find patterns and outliers in data description of big and complex datasets through short
that can help to fast track new treatments, improve statements in natural language .but the limitation of this
efficiencies, and treat patients more effectively. But the approach is representation and processing of imprecision
limitation of cognitive computing is that the system fails at that is characteristic for natural language. Roderick et al.
analyzing the risk which is missing in unstructured data. [15] used Data science in health care, Description of
That includes socio-economic factors, culture, political medical data, Machine learning in health care, and
environment and people. The scope of cognitive Prediction of adverse events based on the traditional
technology is limited to engagement and decision. medical data. Data science is defined as a process of
In Altintop et al. [14] healthcare management, they tried extracting knowledge from externally generated data. It is
to provide quality healthcare service and the effective combination of mathematics, statistics, computer science
utilization of limited resources, it is necessary to make and applied field of study. Description of medical data by
efficiency measurement or knowledge discoveries on raw storing and analyzing medical histories for millions of
data in the Healthcare information system. This can be patients, using thousands of variables to provide better
done using two methods. The first is the efficiency resource allocations. The stored healthcare data is provide
measurement techniques and data mining techniques. In as a input into the model in order to make predictive
this study, fuzzy linguistic summarization is used to model for rare events. In this study the health care data are
analyze operational and financial data of the healthcare in terms of : (a) Census information (such as age or
facilities. Fuzzy linguistic summarization has been shown ethnicity) and simplest biometrics (such as weight). (b)
to be simple, efficient and human consistent way of Clinical information: time stamped labels of transactions,
knowledge extraction. The linguistic summaries are kind assigned according to a medical coding system. These
of outputs of Decision support systems (DSS). Decision record instances of medical decisions, not detailed
support systems (DSS) are a class of computerized outcomes. (c) Financial information: depending on
information system that support decision-making conditions of data sharing with insurance providers, this
activities. The main idea in linguistic summarization for may include claims data, payments received and aggregate
the healthcare is to provide the statements such as “Most metrics of subscriber loyalty or risk. (d) Narratives,
hospitals with few beds have a small amount of total including unstructured text documents such as doctor’s
budget. There are two approaches for generating fuzzy notes. (e) Multiformatted measurements and test results,
linguistic summaries. One way is to generate all possible including numerical data and images. (f) Self‐reported
linguistic summaries and select them, of which truth behavior, such as questionnaires of patient satisfaction. (g)
degree is over a predefined threshold. However, this way Inferred and miscellaneous socioeconomic features, such
is very exhaustible when the number of summarizers as frequency of use of social media or likelihood to
becomes very large. The other way to generate fuzzy purchase an insurance plan. Machine learning methods
linguistic summaries is to use metaheuristics such as (supervised and unsupervised algorithms ) are used for
genetic algorithms. The idea behind the use of genetic medical data which is noisy, imbalanced, standard
algorithm is that the individuals (summaries) of which predictors and classifiers makes the data proper and clear.
truth degree are equal or greater than the predefined Machine learning take a model in cognition process in that
threshold will only survive after several generation it deals with patterns. It decomposes and synthesizes
evolutions. This makes that a desirable set of linguistic information. Decomposition leads to creation of patterns
summaries can be efficiently found instead of generating that will be recognized when observed again. Synthesis
all linguistic summaries. In this paper genetic algorithm is allows completion of partially observed data into familiar
used to generate fuzzy linguistic summaries. In this study, patterns. Supervised learning aims at achieving the best
the dataset of Operational and financial healthcare representation of output. Unsupervised learning attempts
covering the records for the year 2013, which belong to to best group subsets among different sets of data(input).
677 public hospitals located in Turkey with 227 features Lastly the paper predicts rare adverse events for general
that are categorized into 3 groups, 1. Administrative which patient population, based on nonspecific, traditional
contains Total number of medical examination, Load clinical data, identified clinical and medical insurance
factor of beds, Total number of inpatients, City population, records of 500,000 patients. This study is performed in the
Int. J. Advanced Networking and Applications 4275
Volume: 11 Issue: 03 Pages: 4270-4276 (2019) ISSN: 0975-0290

framework of personalized medicine that automatically for ongoing care. Lee et al. [4] proposed a conceptual
select only the most relevant training data for each patient, classification model toward consumer’s behavior in
but at the same time don’t want to restrict the study to any choosing hospital, Besides, the back propagation network
particular population group such as in‐patients, or patients was utilized to build the classification model for finding
already recorded as having increased risk of an adverse consumers’ behavior of choosing hospitals. A neural
event. This has led us to a decision to build an individual network model is useful in identifying existing patterns of
predictive model for each patient. In Khalaf et al. [12], hospital’s consumers.
system reduces costs, provide 24 hours communication
between medical experts and patients and remote V. Conclusion:
monitoring for SCD patients. When the system detects any In this study, we provide the comprehensive survey of
critical condition from the patient, it generates an Machine learning approaches in terms of heath care
automatic message to the medical doctors based on web- management system. we focus on how the different
based platform in order to assist them with optimal methods of machine learning whether they are supervised,
decision-making using ANN with MlP algorithm is to unsupervised or Reinforcement are used for different
extract the information from medical dataset automatically healthcare problems. The healthcare problem can be: fast
and supervised machine learning algorithms for the treatment of a patient, prediction of various diseases,
prediction of the amount of medication. The proposed Remotely Monitoring the Patients using internet of things
framework is divided into two sides. The patient side is (IOT) and Monitoring the patient using sensors.
used to monitor, to store and collect data, as well as to
send notifications to the clinicians in terms of abnormal References:
conditions. The hospital side consists of a database and a
decision support system. Medjahed et al. [6] proposed an 1. Hauskrecht, Milos, and Hamish Fraser. "Planning
automatic in-home healthcare telemonitoring system for treatment of ischemic heart disease with partially
several use. The proposed telemonitoring system is a observable Markov decision processes." Artificial
multimodal platform with several sensors that can be Intelligence in Medicine 18.3 (2000): 221-244.
installed at home. A set of microphones placed in all
rooms, for allow remote monitoring. wearable device 2. Gaweda, Adam E., et al. "Individualization of
called RFpat that can measure physiological data like pharmacological anemia management using reinforcement
ambulatory pulse heart rate, detect posture (standing/ learning." Neural Networks 18.5-6 (2005): 826-834.
sitting and laying). A set of infrared sensors that detect the 3. Schaefer, Andrew J., et al. "Modeling medical treatment
person's presence in a portion of a given home also his using Markov decision processes." Operations research
movement .and a set of domotic sensors like contact and health care. Springer, Boston, MA, 2005. 593-612.
sensors, temperature sensors and several other domotic
sensors for environment conditions monitoring. With the 4. Lee, Wan-I., Bih-Yaw Shih, and Yi-Shun Chung. "The
help of these sensors is the only way to obtain the required exploration of consumers’ behavior in choosing hospital
breadth of information. But the key to produce the by the application of neural network." Expert systems with
required detection is to use the right method that properly applications 34.2 (2008): 806-816.
fuses the provided data from various sources called
multimodal data fusion. The fuzzy logic method is the 5. Niyato, Dusit, Ekram Hossain, and Sergio Camorlinga.
proposed way to meet these challenges of this multimodal "Remote patient monitoring service using heterogeneous
data fusion application. The fuzzy logic decision module wireless access networks: architecture and
reinforces the secure detection of older person distress optimization." IEEE Journal on Selected Areas in
events and his localization. Miškuf et al. [23] proposed a Communications 27.4 (2009).
cloud-based architecture for the healthcare data
6. Medjahed, Hamid, et al. "A pervasive multi-sensor data
classification allows to create one smart system to
fusion for smart home healthcare monitoring." Fuzzy
monitor, track, and store patients healthcare information
Systems (FUZZ), 2011 IEEE International Conference on.
for ongoing care. Therefore, we provide our case study in
IEEE, 2011.
which we took hospital data and used cloud services for
data storing and machine learning. Then we have 7. Alhamid, Mohammed F., et al. "Hamon: An activity
developed a web application on the top of this architecture recognition framework for health monitoring support at
and evaluated the concept in the real environment of the home." Instrumentation and Measurement Technology
hospital. The keywords used in this paper is Healthcare Conference (I2MTC), 2011 IEEE. IEEE, 2011.
IoT, Data Mining, Azure ML, R Shiny. The aim of this
paper is to present the concept definition of a platform for 8. Kathuria, Madhumita, and Sapna Gambhir. "Leveraging
HealthI IoT that connects the advantages of IoT with the machine learning for optimize predictive classification and
emerging cognitive computing. The healthcare Industrial scheduling E-Health traffic." Recent Advances and
Internet of Things (HealthIIoT) is a combination of Innovations in Engineering (ICRAIE), 2014. IEEE, 2014.
communication technologies, things, applications, and 9. Khianjoom, Siwatchaya, and Wipawee Usaha. "Anycast
people that would function together as one smart system to Q-routing in wireless sensor networks for healthcare
monitor, track, and store patients healthcare information monitoring." Electrical Engineering/Electronics,
 Int. J. Advanced Networking and Applications 4276
Volume: 11 Issue: 03 Pages: 4270-4276 (2019) ISSN: 0975-0290

Computer, Telecommunications and Information 21. Chuang, Mao‐Te, Ya‐han Hu, and Chia‐Lun Lo.
Technology (ECTI-CON), 2014 11th International "Predicting the prolonged length of stay of general surgery
Conference on. IEEE, 2014. patients: a supervised learning approach." International
Transactions in Operational Research 25.1 (2018): 75-90.
10. Hurwitz, Judith S., Marcia Kaufman, and Adrian
Bowles. Cognitive computing and big data analytics. John 22. Mshali, Haider, et al. "A survey on health monitoring
Wiley & Sons, 2015. systems for health smart homes." International Journal of
Industrial Ergonomics 66 (2018): 26-56.
11. Gorunescu, Florin. "Intelligent decision systems in
Medicine—A short survey on medical diagnosis and 23. Miškuf, Martin, Iveta Zolotová, and Jozef Mocnej.
patient management." E-Health and Bioengineering "Healthcare data classification—Cloud-based architecture
Conference (EHB), 2015. IEEE, 2015. concept." Cybernetics & Informatics (K&I), 2018. IEEE,
2018.
12. Khalaf, Mohammed, et al. "Applied Difference
Techniques of Machine Learning Algorithm and Web- 24.Jian, He, and Hu Chen. "A portable fall detection and
Based Management System for Sickle Cell alerting system based on k-NN algorithm and remote
Disease." Developments of E-Systems Engineering medicine." China Communications 12.4 (2018): 23-31.
(DeSE), 2015 International Conference on. IEEE, 2015.
Profile
13. Kau, Lih-Jen, and Chih-Sheng Chen. "A smart phone-
based pocket fall accident detection, positioning, and Dr. Krishan Kumar Goyal received
rescue system." IEEE journal of biomedical and health M.Tech in Computer science from U.P.
informatics19.1 (2015): 44-56. Technical University, Lucknow. He has
14. Altintop, Tunahan, et al. "Fuzzy Linguistic received Ph.D. in cryptography from
Summarization with Genetic Algorithm: An Application Dr. B. R. Ambedkar University, Agra.
with Operational and Financial Healthcare He has also received master's degree in Computer
Data." International Journal of Uncertainty, Fuzziness and Application & Mathematics from Dr. B.R. Ambedkar
Knowledge-Based Systems 25.04 (2017): 599-620. University, Agra. Presently, he is working as an Associate
Professor and Dean, Faculty of Computer Application at
15. Roderick, Oleg, et al. "DATA ANALYSIS AND Raja Balwant Singh Management Technical Campus,
MACHINE LEARNING EFFORT IN HEALTHCARE: Agra. He has participated in several faculty development
ORGANIZATION, LIMITATIONS, AND programs, seminars and workshops. He has published
DEVELOPMENT OF AN APPROACH." Internet of several research papers in leading journals of national and
Things and Data Analytics Handbook (2017): 295-328. international repute. He has authored two books. He is
16. Shameer, Khader, et al. "Predictive modeling of Life member of several societies such as Computer
hospital readmission rates using electronic medical record- Society of India, Ramanujan Mathematical Society of
wide machine learning: a case-study using Mount Sinai India, Cryptology Research Society of India etc. He is
Heart Failure Cohort." PACIFIC SYMPOSIUM ON also a reviewer and member of editorial board of
BIOCOMPUTING 2017. different national and international journals. His area of
interest includes Cryptography, Cyber Security, Privacy
17. Nguyen, Hoa Hong, et al. "A review on IoT healthcare and Security in Online Social Media, Machine Learning,
monitoring applications and a vision for transforming Natural Language Processing etc.
sensor data into real-time clinical feedback." Computer
Supported Cooperative Work in Design (CSCWD), 2017 Aejaz Hassan Paray received
IEEE 21st International Conference on. IEEE, 2017. Master of Computer
Application from University of
18. Mazhar, Tabib Ibne, et al. "Spinal Cord Injured (SCI) Kashmir. Presently he is pursing his
patients' length of stay (LOS) prediction based on hospital Ph.D degree in Computer
admission data." Electrical Information and Application from Bhagwant
Communication Technology (EICT), 2017 3rd University of Ajmer Rajasthan. He has participated in
International Conference on. IEEE, 2017. several faculty development programs, seminars and
19. Baker, Claire, et al. "Healthcare analytics and workshops. He has published several research papers in
visualization using SEMantic Open Source Software leading journals of national and international repute. He is
(SEMOSS)." Systems and Information Engineering member of several societies such as Computer Society
Design Symposium (SIEDS), 2017. IEEE, 2017. of India etc. His area of interest includes Artificial
Intelligence, Artificial Neural Network, Machine Learning
20. Nithya, B., and V. Ilango. "Predictive analytics in etc.
health care using machine learning tools and techniques."
International Conference on Intelligent Computing and
Control Systems (ICICCS). IEEE, 2017.

View publication stats