FORECASTING OF RICE CROP PRODUCTION USING RECURRENT

NEURAL NETWORK

__________________________________

A Capstone Project
Presented to
The Faculty of the Graduate School
University of the Immaculate Conception
Davao City

__________________________________

In Partial Fulfilment
of the Requirements for the Degree
Master in Information Technology

__________________________________

ANTONETTE R. ALBARRACIN

November 2017
 Table of Contents

INTRODUCTION..................................................................................................3
Significance of the Study............................................................................4
General Objective.......................................................................................4
Specific Objective.......................................................................................4
Scope and Delimitations.............................................................................5
Limitation of the Project.............................................................................5
REVIEW OF RELATED LITERATURE
Review of Related Literature Map..............................................................6
Developing of Dataset for Rice Production.................................................7
Production Forecasting....................................................................7
Importance of forecast.....................................................................7
Importance of Rice production........................................................7
Rice production in Davao del Sur....................................................8
Rice Data..........................................................................................9
Climate Data.....................................................................................9
Weather factor of Rice production...............................................................9
Climatic environment and its effects on rice....................................9
Variability of Rainfall......................................................................10
Effects of Temperature....................................................................11
Low temperature effects..................................................................11
Effects of relative humidity.............................................................11
Recurrent Neural Network...........................................................................12
Deep Learning..................................................................................12
Long-Short-term Memory................................................................12
RNN Architecture............................................................................13
RNN Methods in Data Forecast...................................................................14
Back-propagation Time....................................................................14
Deep Learning..................................................................................14
Evaluation of the System..............................................................................15
Functionality.....................................................................................15
Reliability..........................................................................................15
Usability............................................................................................15
Technological Background and Conceptual Framework..........................................15
Conceptual Framework.............................................................................................16

1
METHODOLOGY.......................................................................................................17
Require Specification.......................................................................................17
Analysis............................................................................................................17
Design...............................................................................................................18
Development....................................................................................................19
Back-end Application...........................................................................19
Web Application...................................................................................19
Testing..............................................................................................................19
GANT CHART............................................................................................................21
REFERENCES.............................................................................................................22

2
 CHAPTER I

Introduction

The primary goal of the technology is to create accessible, innovative and useful
methods for analyzing environmental factors that affect the production of rice.

Rice is the most needed by people for each day, Filipino meals consumed rice
for almost three times a day, and it is one of the leading sources of income for
millions of Filipino farmers. Moreover, it sustains protein and calories [1]. The
International Rice Research Institute (IRRI 2000) studied the food problem
concerning world population, and they predict that 800 billion tons of rice will be
required in 2025 [2]. Furthermore, the department of agriculture target rice production
this year around 19.32 million metric tons. The total population in the Philippines
now is approximately 100 million Filipinos. In case this goal would be achieved, only
96 million Filipinos would eat a regular meal every day, and the other 4 million
Filipinos would starve which primary cause by overpopulation. The population of the
Philippines is rapidly growing at least 2% per year and which is equivalent to the birth
of almost 2 million Filipinos per year. In Asia, the Philippines consider as the fast-
growing country in Asia. According to the Philippine Rice Research Institute
(PhilRice), the average Filipino consumes 3 cups of rice per meal, which means 9
cups of rice a day. And the Philippines was the most prominent importer of rice in the
world. This rice shortage is more aggravated by climate change. Climate change
remains to be one of the constraints to rice production in the country [3].

During the period of natural calamity the reservation of food is adequate in the
demand of food in the long term that is significantly concern for many developing
countries worldwide. Considering that some factors that affect the production are the
environmental effect of weather, type of soil, fertilizer used by the farmers and water
management [10].

The province of Davao del Sur is considered one of the significant rice-
producing areas of Region XI. It regarded as Mindanao's top rice-yielding province
because of its municipality's high yield performance [4]. The office of the provincial
agriculture (OPAG) - rice program provide extension and services like giving training
and seminar on Climate Risk Assessment, training on technology demonstrations,
technology commercialization, , updating and retooling of the new farm equipment,
logistical support on Piloting of Irrigation system [5]. The office of the provincial
agriculture remained utilized in a manual order which hard for them to prepare reports
and relevant data on monitoring the rice production to anticipate the future production
of rice.

The province produced a high production of rice the population in Davao del
Sur is hurriedly growing, based in 2010 census of population and housing the total
population in Davao del Sur is approximately 868,690 which means it is less than the
overall population growth in the year 2000 with the total of 758.800 [6]. Furthermore,
it is insufficient to feed the population in the coming years if the total land area is still
4,556,042.79 and including the effect of the rice production [7]. Thus, this will results
a constant increase of price.

3
 A forecast is what may be the result to happen, based upon the observations of
previous experience and present conditions. It is a basis also in deciding what action
would be to secure the necessary end [8]. The concern of forecasting of rice crop
production is stable concern since the beginning of the history of agriculture. The
techniques of forecasting have evolved and seeking the data with accuracy,
granularity, comparability and timeliness [9].

The researcher aims to forecast rice production actual observed for the last 4
years of climatic parameter data and the annual rice production, This may help people
to better understand the possible changes in values of the data series and may even
help the government for any actions they can make for the upward moving of
production of rice.

The purpose of agricultural forecasting is the wise guidance of production that there
may continue to be a proper balance between the various lines of output and between
agriculture and other industries. Agricultural prosperity, in general, depends primarily
upon the ability of farmers to anticipate the future.

To help the farmers, the researcher came up with the study entitled FORECASTING
OF RICE CROP PRODUCTION USING RECURRENT NEURAL NETWORK.
The forecast rice production will be implemented in a web application for providing
easy access by the users.

Significance of the Project

ICTs role helps to integrate and efficiently used in agriculture development as

facilitating tools to boost its impact on the lives of farmers. Specifically, the project
shall be beneficial to the following:

Department of Agriculture (Rice Program) - The project will provide additional

tools to the department of agriculture to monitor the consumption of rice crop
production in the province. It is vital to the Department of agriculture (rice program)
in Davao del Sur to have an idea of the supply of rice if it is sufficient to the need of
people in the coming years and to prepare the estimated price of rice in the future.

Future Researchers. The project provides a reference for other researchers concern
in rice production forecast.

General Objective

The general objective of the project is to develop the application “Forecasting of rice
crop production using recurrent neural network” this will serve as a web application
for forecasting rice crop production in Davao del Sur and providing information to the
management of rice program for the anticipated supply of rice crop production in
Davao del sur.

4
Specific Objective

In order to achieve the general objective of the project, the following specific goals
must be satisfied:

1. Train a model that will predict the average production of the rice by quarter
using the datasets of weather: temperature, humidity, sunshine, rainfall and the
rice production.

2. Develop a web-based application that manages the data of rice production

which can be accessed by the End-users.

3. Predict the future production of rice from the available information and to
study the impact of weather factors affecting the rice production.

4. Evaluate the system from the IT experts and End-users concerning:

1a. Accuracy
1b. Functionality
1c. Reliability
1d. Usability

Scope and Delimitations

The researcher will try two different processes to get the best solution in training the
model in forecasting the production of rice:

Train a model using Recurrent Neural Network (RNN) to a sequential data like time
series
Training of model will be based on the secondary data accessed through the
government website. The weather data from the PAG-ASA for the period of 2014 to
2016 and the production of rice every quarter will be collected from the site of
Countrystat Philippines. The prediction will only circulate from the enumerated
condition.

Secondary data application will use API from world weather online to automatically
get the current weather and will store in a database so that it will be included for
future retraining of the model.

The project covers the following:

a. The researcher will focus on the weather data particularly the temperature,
rainfall, humidity, and sunshine to predict the production of rice.
b. The guest can view and import rice production data.
c. The administrator is the person can add, modify and edit the data of the rice
production.
d. The system will display gathered data for rice prediction in a table or graph.

5
Limitation of the project

In developing the system, the constraints of the study are as follows:

a. The researcher includes climatic parameters for the prediction due to the
extensive data's of other factors which may affect the rice production like the
type of soil, and fertilizer used, growth, nutrient, pest and rice diseases.

b. The data gathering of the rice production will be the total production in Davao
del Sur but not categorized per production of municipalities cause of the
availability of data provided by the website of CountrySTAT Philippines.

Review of Related Literature

Significant pieces of literature were reviewed to conceptualize the project. The topics
included during the review are the importance of forecasting and the weather factor
that affect the production of rice the forecasting technique used in a known facts on a
given date, and it assumes the production condition by quarter and production.

6
 RECURRENT NEURAL NETWORK: RICE CROP
PRODUCTION FORECASTING

Developing dataset gathered data Train a Model for prediction Implementation in web application
of weather and rice production of rice production for the production of rice Evaluation of the
system

Recurrent Neural Deep Learning

Production Weather Factors of Rice Model Training Network

Forecasting Production Accuracy

W. S. McCulloch and W.
Prediction of Functionality
Pitts (1943) G. E. Hinton,
production Reliability
S. Osindero, and Y.-W. Usability
RNN Architecture Teh (2006) D. Silver, A.
Developing of a dataset
Huang, C. J. Maddison, A. Campinas, SP, Brazil
Rumelhart, D., G. Recurrent Neural Network Guez, L. Sifre, G. van den 2010
Hinton et al, (1986)
Climatic environment Man Galih Salman Man
Driessche, et al.,(2016)
Jürgen Schmidhuber (2015)
and its effects on Rice Galih Salman et al, A. Lusci, G. Pollastri, and
Importance of Rumelhart, D.E., Hinton et al,
P. Baldi,(2013)
(2015) (1988)
Forecasting SEAMEO, SEARCA
Satake, T., Yoshida, S (1978)
Chen Guonan et al, 1987 Geng, S., Cady, C.W. (Eds.
Bindu Garg et al. (1991)
Ana Ramirez et al,
Deep Learning RNN Methods in
data forecast Web Application

Variability of Rainfall P. Langley, H.A. Simon Vahid Mansouri (2014)

(1995) Daniel Nations (2016)
Y. Bengio, A. Courville Dana Nourie (2006)
Rice production Riehl, H. 1954 et al,
in Davao del Sur Liu, J. N. K. et al, (2015)

Philippine Statistic Long Short-Term

Authority Rice Data Effects of Temperature Memory Network
Naohisa Koide1 et al,
(2011) Naohisa Koide1 et al,
Oscar M. Lopez et al, (2015) Alex Graves
(2011)
Kaneda, C. et al, (1972) F. A. Gers et al, (2001)
Chung, G. S (1979)

Importance of
Chung, G. S
Rice Production Climate Data Effects of Relative
L.T. Evans (1998)
Humidity
Moron, V., A. W.
D.J. Greenland (1997)
Robertson et Al, (2006) Gen-ichi Hirai et al, (2000)
de Datta, S.K.(1981)
BC Ministry of Agriculture,
Fisheries, and Food (1994)
Developing of Dataset for Rice Production

Production forecasting

The Past research suggested that it may be possible to forecast aspects of

Philippine rice production based on climate information alone. Such forecasts could
potentially benefit decision making from national, regional, and local governments to
local farmers [6]. The rice production in the Philip pines has been increasing for more
than five decades through the development of lands, construction of new irrigation
systems, improvement of existing irrigation systems, and adaptation of new
technologies such as modern rice varieties and improved fertilizer usage [7].

Importance of Forecasting

The prediction of regional crop productivity and yield was essential to use
natural resources sufficiently and reasonably and help to evaluate the food security
and make the local productive plan [8]. Crop production is being handled with the
field data accuracy of data, and it is always a matter of concern. It is studied that most
of the work on time series has been done for dealing with forecasting problems like
prediction in management information system, forecasting, sales forecasting,
budgetary analysis, stock market fluctuations, and business analysis, etc.
Nevertheless, there is a persistent need for precise forecasting technique to deal with
the nonlinear and involved behavior of crop production system. Evaluating and
predicting rice production is a complex field and challenging task. The forecasting
method and its results should be understood not only by administrators in agriculture
but also to those who use the results in decision making.[9]

Importance of Rice Production

Rice farming accedes as one of the world's most productive and sustainable
cropping system. Rice is the basic food for 2.7 billion people, approximately half of
the world population and more than half of the world’s farmers grow this crop. It is
estimated that rice consumption soars by nearly 60 million tonnes almost in every
eight years. Asia [10]. Alone consume and cultivate 90 percent of total rice
production.
Fundamentally, change in demand for rice is compelled by population growth, the
rank of per capita income, and changes in the price of rice concerning other substitute
crops [11]. Rice in the Philippines is typically planted by transplanting seedlings in
the puddle, bundled fields, where a constant height of the water is maintained
throughout the growing season. This way of water management provides a suitable
environment for optimal rice growth and weed control [14].

7
Rice production in Davao del Sur

Figure 1. Rice Volume Production in Davao del Sur as of 2014-2017 [12].

Fig. 1 Shows the rice production in Davao del Sur graph the red and gray label
indicated the total production of Palay in Davao del Sur as 2014-2017[12].

Figure 2. Planted rice area by province [13]

Fig. 2 shows the Planted rice area by province within the region the Davao del Sur
ranked as the 2nd big rice area planted out of 5 provinces in a Davao region.

8
Developing of a dataset

Rice Data

The Provincial data on rice production area harvested of Irrigated systems,

rain-fed system, and all ecosystems, from 2014-2017 were downloaded from the
Philippine Rice Statistics e-Handbook published as a collaborative project of the
Philippine Rice Research Institute and the Philippines Bureau of Agricultural
Statistics. The Davao del Sur has composed of 10 city/ municipality and the big area
planted to rice in Davao del Sur the municipality of Magsaysay, Davao del Sur [13].

Climate Data

The Philippine climate is widely different by region due to its complex

topography, classified into four types defined by the Philippines Atmospheric,
Geophysical and Astronomical Services Administration (PAGASA). Type I has a
distinct summer monsoonal wet season from May to October and a dry season from
November to April. Most western regions belong to this class. Type II, on the
contrary, has no definite dry season and maximum rainfall in November to December
associated with the northeast winter monsoon. Most of the north-eastern regions are
categorized into this
Type. Type III is an intermediate band of Type I and II. It has maximum rainfall from
May to October with the unclear but relatively dry season from November to April.
Most southern areas belong to Type IV which has evenly distributed rainfall
throughout the year [15].

Weather Factors of Rice Production

Climatic environment and its effects on Rice

Climate variability is subjected according to the changes in temperature,

erratic rainfall pattern, more intense and extreme climatic events, El Nino incidences,
and combined risks of all the mentioned climate variability [16].

Fig. 3 Annual losses (‘000 Mg) in Philippine rice production due to typhoons/floods, droughts, and
pests from 1970 to 1990 (adapted from PhilRice-BAS, 1994).

9
 Weather and climate affect plant growth and development, and the
fluctuations and occurrences of climatic extremes particularly at critical crop growth
stages may reduce yield significantly [17]. Weather and climate have a direct
influence on cropping systems and plant yield. Thus, weather fluctuations and climate
variability play a significant role in crop growth and yield. The occurrence of
abnormal weather episodes during the growing season or critical development stages
may hamper growth processes resulting in yield reduction. This makes climate
variability a threat to food production leading to severe social and economic
implications [18]. The challenge for rice production is twofold: coping with
population growth while also facing climate change. Unforeseen changes associated
with global warming in temperature, carbon dioxide, and rainfall are expected to
impact rice production. Studies have shown that increased temperature, due to climate
change, adversely affect rice crop physiology ultimately decreasing crop yields and
grain quality. Because carbon dioxide is an essential component in photosynthesis,
increase the atmospheric concentration of carbon dioxide is expected to increases
plant growth and consequently rice yields [20].

Variability of Rainfall

Most of the tropical Southeast Asian countries, such as significant parts of

Burma, Kampuchea, Indonesia, Philippines, Thailand, and Vietnam, receive about
2000 mm of rainfall annually. This should be adequate for one rice crop provided
rainfall distribution is reasonably uniform. Even in areas where the annual rainfall is
1200-1500 mm, if rainfall is concentrated in the monsoonal season (as is usual), it is
adequate for a single rice crop. Variability of rainfall affects the rice crop at different
times. If the variability is
Associated with the onset of the rain, stand establishment and the growth duration of
rice wil changed. If variability is associated with an untimely cessation at the
reproductive or ripening stage of the rice crop, yield reduction is severe.

Table 1.1 Maximum Rainfall Intensity in 24 Hours

Reported from Several Tropical Stations
_____________________________________________
Station Rainfall(mm)
_________________________________________________________________________________

Digos City, Davao del Sur 1846

Rainfall variability is more critical for upland rice than for lowland rice.
Moisture stress can damage, or even kill, plants in an area that receives as much as
200 mm of precipitation in a day and then gets no rainfall for the next 20 days. An
evenly distributed precipitation of 100 mm/month is preferable to 200 mm/month that
falls in 2 or 3 days [19]. The Minimum rainfall is 115 cm. Although the regions have
average annual rainfall between 175—300 cm is the most suitable [42].

10
Effects of Temperature

The average year-round temperature measured from all the weather stations in
the Philippines, except Baguio City, is 26.6 °C (79.9 °F). Colder days are usually felt
in January with temperature averaging at 25.5 °C (77.9 °F). The warmest days are in
May with a mean temperature of 28.3 °C (82.9 °F). Elevation factors significantly in
the variation of temperature in the Philippines. In Baguio City, with an elevation of
1,500 m (4,900 ft) above sea level, the mean average temperature is 18.3 °C (64.9 °F)
or colder by about 4.3 °C (8 °F), than in other areas.

Seasons
The climate of the country is divided into two main periods:
1. The rainy season, from June to November;
2. The dry season, from December to May.
The dry season may be subdivided further into
(a)The cold dry season, from December to February; and
(b) The hot, dry season, from March to May [21].

Low-Temperature Effects

Injury of rice plants by low temperature occurs in temperate and tropical

regions [22]. In temperate areas, the cold damage is the primary constraint limiting
the rice growing area and length of growing season. In Korea, the low temperature
often causes low rice yields [23]. Rice is a tropical crop and grown where the average
temperature during the growing season is between 20°C and 27°C. Abundant
sunshine is essential during its four months of growth. The minimum temperature
should not go below 15°C as germination cannot take place below that temperature
[42].

Two Factors cause cold injury to rice-cool weather and cold irrigation water. The
common types of symptoms caused by low temperature are:

 Poor germination
 Slow Growth and discoloration of seedlings
 Stunted vegetative growth characterized by reduced height and tillering
 Delayed heading
 Incomplete panicle exsertion
 Prolonged flowering period because of irregular line
 Degeneration of spikelets
 Uneven maturity
 Sterility
 Formulation of abnormal grains

Effects of Relative Humidity

The seedlings of rice ( Oryza sativa cv. Nipponbare) at a plant age of 3 leaves,
were cultured for 10 days in a 12-h light/12-h dark cycle, and exposed to 60% (low)
or 90% (high) relative humidity during the light and dark period in all combinations.

11
Low humidity in light and high humidity in darkness, significantly increased the rate
of leaf emergence, plant height, leaf area, leaf blade length, the number of roots, total
root length and dry matter production as compared with low humidity in both light
and dark periods[24].
If the humidity is too low, plant growth is often compromised as crops take much
longer to obtain the saleable size. Also, smaller leaves often drop off, growth is hard,
and overall quality is not very good. Whether the humidity is too high or too low, the
loss of quality reduces the selling price of crops and increases production costs, both
of which reduce profits [25].

Recurrent Neural Network

In recent years, extensive artificial neural networks (including recurrent ones)

have won numerous contests in pattern recognition and machine learning. The
historical survey compactly summarizes relevant work, much of it from the previous
millennium [26].
When the network has loops, it is called an RNN. It is possible to adapt the
Backpropagation algorithm to train a recurrent network, by "unfolding" the network
through time and constraining some of the connections to always hold the same
weights [27].

Deep learning

Deep learning is a machine learning approach that is based on neural networks

[28].
Recently, deep learning has become one of the most active technologies in many
research areas. As opposed to shallow learning, deep learning usually refers to
stacking multiple layers of neural network and relying on stochastic optimisation to
perform machine learning tasks. A varying number of layers can provide a different
level of abstraction to improve the learning ability and task performance [29]. The
word "deep" in deep learning indicates that such neural network (NN) contains more
layers than the "shallow" ones as used in typical machine learning models. This multi-
layer NN has gained wide research interest after the successful implementation of the
layer-wise unsupervised pre-training mechanism that is employed to solve the training
difficulties efficiently. The "deep" architecture is very significant in compared with
Shallow models as NN with deep architecture can provide higher learning ability. The
successful applications of deep learning in various domains, which have been reported
by various researchers, have motivated its use in weather representation and modeling
[30].

Prediction of Production

Long Short-Term Memory

The LSTM architecture consists of a set of recurrently connected subnets,

known as memory blocks. These blocks can be thought of as a differentiable version
of the memory chips in a digital computer. Each block contains one or more self-

12
connected memory cells and three multiplicative units the input, output and forgets
gates that provide continuous analogs of write, read and reset operations for the cells.

Figure 3. Network Architecture Sample

The shading of the nodes in the unfolded network indicates their sensitivity to
the inputs at time one (the darker the shade, the higher the sensitivity). The sensitivity
decays over time as new data overwrite the activations of the hidden layer, and the
network `forgets' the first inputs [31]. Over the past decade, LSTM has proved
successful at a range of synthetic tasks requiring long-range memory, including
learning context-free languages [32].

RNN Architecture

Recurrent neural network (RNN) is an artificial NN used for time series

prediction [33]. RNN consists of one or more hidden layer. The first layer has the
weight that is obtained from the input layer every layer will receive weight from the
previous layer. This Elman network has activation function that can be in the form of
any purpose both continue and discontinue. Delay that is happened the first hidden
layer in the previous time (t- l) can be used at the current time (t). The unique of the
recurrent Neural network is the feedback connection which conveys interference
information (noise) at the previous input that will be accommodated to the next input.
Let x (t) and yet) be input and output time series respectively; the three connection
weight matrices are WiH, WHH, and WHO [34].

Figure 4. Recurrent Neural Network Model

13
 RNN Methods in Data Forecast

Back-propagation Time

The Back-Propagation through Time (BPTT) learning algorithm is the natural

extension of standard back-propagation that performs gradient descent on a complete
unfolded network. If a network training sequence starts at time t0 and ends at time t1,
the total cost function is simply the sum over time of the standard error function
Esse/ce(t) at each time step [35].

Figure 5. Backpropagation architecture.

Developing web application for the average production of rice

Deep Learning

Deep learning is a branch of machine learning methods lying on ‘deep’ network

architectures. The concept of ‘deep learning’ has been proposed for decades with the
name ‘cybernetics' in 1943 [36]. However, it has been regarded as being more of a
fancy concept than an appropriate technology, due to three significant technical
constraints. The three technical constraints are: 1) lack of sufficient data, 2) lack of
computing resources for large network size, and 3) lack of efficient training
algorithm. Recently, the constraints are tackled by the digitalization of modern society
and the development of high-performance computing. Furthermore [37], made a
breakthrough in efficient deep neural network training via a strategy called greedy
layer-wise pre-training, which enables practical implementations of deep learning.
Deep learning has recently seen phenomenal success in various areas including 1)
Computer Vision (CV) such as Google Goggles, which uses deep learning for object
recognition; 2) expert systems such as Alpha Go designed by DeepMind [38]. And 3)
medical sciences, which employs deep learning to assist pharmaceutical companies in
new drugs design [39].
Web Application

14
The web application has been around since before the web gained mainstream
popularity, the web application is based on the client-server to enter information,
while the server stores and retrieve information[40]. Where web application is also
accessed with a web browser and are popular because of the ease using the browser as
a user client its ability is to update and maintain web application, and it can be simple
as the page that will show the current date and time[41].

Evaluation of the System

Functionality
The application does show appropriate and has available functions required for its
execution, and it shows precise in executing its tasks and accurate in its results, the
web interacts with the specified modules and has the capacity to operate with
networks, it has a secure access through passwords and an internal backup routine.

Reliability
The application has frequent failures it reacts appropriately when failures occur, the
application informs users concerning invalid data entry, and it is capable of
recovering data in the event of failure.

Usability
Easy to understand the concept and application and perform its function, easy to learn
how to use, the application facilitates the users’ data entry and retrieval of data, easy
to operate and control then it provides help in a precise manner[43].

Technological Background and Conceptual Framework

The following describes the necessary software application, add-ons, libraries

and supporting software that the researcher will use the development of the proposed
system.

The researcher will use a python programming language in training a model.

The python language already had a wide library for numerical computation which it
can be used to accept input data of weather condition and the quarter production. This
includes numpy, matplotlibs. In training the model the keras library will be utilized in
this study, it supports recurrent network and facilitates fast prototyping and
experimentation of data.

The tensorflow will help to provide a number of methods for constructing

Recurrent Neural Networks in sequential data for the inputs and output of data. And
also allow efficient numerical computation using data flow graphs. The Long-short
term memory will utilize in this study since the LSTM can learn and remember over
long sequences of data.

For the development of the application, the researcher will use Asp.net for the
website builder to dynamically build web pages and Django a popular web framework
for python in the deployment of the application as a web application. The Django can
read or generate XML data and works with database management system like MySql.

15
 The researcher will use Django among other python web framework since the
framework is scalable. Furthermore, it has an enormous developer community, and it
is complicated but easy to learn.

Conceptual Framework

The proposed Forecasting of rice crop production application consists two types of
application.

The first module of the web application will be developed using web technology
particularly the python programming language and the Django web framework. The
open-source will provide the web interface for the user of the office of provincial
agriculture – (rice program). The module used to generate the output of rice crop
forecasting and will be provided as an input to the neural network model.

Second, the module of the application is the backend module the application will be
developed in Python programming language with keras library in tensorflow to
execute recurrent neural network model. Figure 6. Shows the Conceptual framework
of the forecasting of rice crop production.

Fig. 6. Forecasting of rice crop production

Conceptual Framework

16
 CHAPTER II

METHODOLOGY

This part covers the section on how to improve, conceptualize the

project life cycle from the beginning to end of a project.

Require Specification

The researcher conducted an initial interview with the head of the

office of the provincial agriculture –(rice program) of Davao del Sur to
understand the problem of rice production. The researcher gathered sample
data of rice production per year and reports of rice crop production in the form
of graph and tables the data collected would serve as a reference of this study.

The researcher will gather electronic data from the Philippines Atmospheric,
Geophysical and astronomical services administration(PAG-ASA) for the
weather data which includes four climatic parameters the temperature, rainfall,
relative humidity, and sunshine. For the rice crop production will be collected
from the CountrySTAT Philippines the website integrates national food and
agricultural statistical information.

The data will be stored in the proposed web-based application and will
serve as the input data of Recurrent neural network model. The output of the
model would produce a forecasting of rice crop production in Davao del Sur.

Analysis

Many considerations will be taken account in creating the system, first

is to find appropriate language suitable for the forecasting of rice crop
production using a recurrent neural network that must support the web
application. Second is to know the accurate library of python that suits for
forecasting, The Python is the most accessible language to learn, and the
language has numerous plugins to make computing more accessible. Python
can be deployed as a web application through a different framework like
Django. Lastly, is to find application library to be used in training the model in
tensorflow.

17
 A use case analysis was also used to determine the behavior of the system.

Rice crop Production Forecast Web Application

Figure 7. Use Case Diagram

Of the Web Application

Design

The proposed study has three parts based on the specific objectives.
The Figure 8. Shows the type of inputted data for the datasets training for the
prediction of rice production.

Rice Production of Davao del Sur

Climatic Parameters Data

Figure 8. Sample Data Gathered

18
 The data are passed to the RNN, which is useful in sequential data, and
each neuron or unit can be used in its internal memory in maintaining
information about the previous input. The LSTM will be utilized also to
perform in holding long-term memories to excel in learning from sequential
data.

Development

The rice forecasting application consists of two types of application the

backend application and web application.

Backend application

In data gathering, the researcher will be gathered weather data and rice
production from the PAG-ASA and the countrystat Philippines. In model
training first is to build the model inputting of climate parameters and rice
production datasets to the model, next is the preparation of the model and test
the model of its accuracy, The backend module will be developed with python
language with keras library to execute the RNN model. Furthermore, it will
also utilize libraries in tensorflow. And lastly is optimizing and compressing
the model.

Web development

The researcher will develop user interface of the application by using

the Python language with Django web framework. The module used to
generate the output of forecasting of rice crop production in Davao del Sur.
The application has two types of interfaces, the user and administration page
for the user side will manipulate the given output of the application while the
administrator manages the technical side of the application. The developer will
do testing and debugging of web application to find and resolve the problem
within the program.

Testing

The target evaluators of the project’s testing phase were the experts of OPAG
(Office of the Provincial Agriculture – Davao del Sur), and the farmers of Davao del
Sur who are knowledgeable, at the very least form, in using website. Upon conducting
a test on the project, a number of selected respondents/evaluators were employed as
significant elements on sampling design and technique. A total of (n) evaluators
comprised the tester/evaluator group. Mock Usability Testing will be the method used
to observe and evaluate the testers/evaluators on how well they can manipulate the
website.

The data gathered from the detailed questionnaires were used as the basis for
interpretation of the overall flow and performance of the system which provided aid

19
in solving problems, recognizing several factors that may affect the structures and
features of the forecasting of rice production in the web.

The data gathered in the study were the following:

1. Descriptive ratings on the level of functionality of the system on its features.

2. Possible problems that were encountered.
3. The recommendation and the improvement of the system.

20
 Gantt Chart

Tasks Duration Start Finish June July August September October November
1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5 1 2 3 4 5
Initial meeting 1 day June 29,
2017
June 29, 2017

with adviser
Selection of the 28 Days July
2017
1, July 28, 2017

capstone project
Literature review 19 Days July 29,
2017
August
2017
16,

Review of 43 days July 30,

2017
September 9,
2017
Methodology
Revision of paper 8 days August
26, 2017
September 2,
2017
Research Methods 45 days August 9,
2017
September
22, 2017
planning
Submission of 14 days Septemb
er 17,
September
30, 2017
proposal draft to 2017
adviser
Collection of 12 days Septemb
er 9,
September
20, 2017
primary Data 2017
Revision of Paper 52 days Septemb
er 16,
November 7,
2017
2017
Project Proposal 1 day Novembe
r 15,
November 15,
2017
Presentation 2017

21
 REFERENCES

[1] S. Ali et al., “Climate Change and Its Impact on the Yield of Major Food
Crops: Evidence from Pakistan,”.
[2] Masayoshi Kubo and Minjmaa Purevdorj., “The Future of Rice Production
and Consumption”. (2004).
[3] Renee Juliene Karunungan et al., “The problem with rice” 2014, retrieved
from https://www.rappler.com/move-ph/issues/hunger/commentaries/73333-
problem-rice-shortage-philippines.
[4] Rosa Fe D. Hondrade., “The LGU Extension Services in a Major Rice-
Growing Area: The Case of Hagonoy, Davao del Sur”. (2007)
[5] Luyjilene V. Legaspi., “Comparative analysis of government extension
serviceson rice farmers: the case of the municipalities ofmagsaysay and
matanao, davao del sur”.
[6] Philippines Statistics Authority, Retrieved from
https://psa.gov.ph/content/davao-del-sur-had-population-870-thousand-results-
2010-census-population-and-housing.
[7] CountrySTAT Philippines, Retrieved from
http://countrystat.psa.gov.ph/selection.asp
[8] M. SDahikar and D. VRode, “Agricultural Crop Yield Prediction Using
Artificial Neural Network Approach,” Int. J. Innov. Res. Electr. Electron.
Instrum. Control Eng., vol. 2, no. 1, pp. 2321–2004, 2014.
[9] J. Miaoguang and J. Chaochong, “Forecasting Agricultural Production via
Generalized Regression Neural Network,” Proc. IEEE Symp. Adv. Manag. Inf.
Glob. Enterp. AMIGE 2008, pp. 1–3, 2008.
[10] S. Ali et al., “Climate Change and Its Impact on the Yield of Major Food
Crops: Evidence from Pakistan,”
[6] Lansigan, F. P., 2005: Coping with climate variability and change in rice
production systems in the Philippines. Rice is Life: Scientific Perspectives for
the 21stCentury: Proceedings of theWorld Rice Research Conference, K.
Toriyama, K. L. Heong, and B. Hardy, Eds., International Rice Research
Institute, 542–545. ——, W. L. de los Santos, and J. O. Coladilla, 2000:
Agronomic impacts of climate variability on rice production in the Philippines.
Agric. Ecosyst. Environ., 82, 129–137.
[7] Kikuchi, M., A. Maruyama, and Y. Hayami, 2003: Phase of irrigation
development in Asian tropics: A case study of the Philippines and Sri Lanka.
J. Dev. Stud., 39, 109–138.
[8] Chen Guonan, An preliminary study on plant productivity in China based on
Miami Model. Journal of natural resources, 1987 2(3):270-277.
[9] Bindu Garg, M.M. Sufyan Beg, A.Q. Ansari, “Fuzzy Time Series Model to
Forecast Rice Production”.
[10] L.T. Evans, Feeding the Ten Billion: plants and population growth. University
Press, Cambridge, UK, Pg: 247,(1998).
[11] D.J. Greenland, The Sustainability of Rice Farming. CAB International,
Wallingford,UK, Pg:273, (1997)
[12] Philippine Statistic Authority, retrieved from
http://countrystat.psa.gov.ph/selection.asp
[13] Naohisa Koide1,4, Andrew W. Robertson*2, Amor V. M. Ines2, Jian-Hua
Qian2 , David G. DeWitt2, Anthony Lucero3 “Prediction of Rice Production
in the Philippines using Seasonal Climate Forecasts”, 2011.

22
[14] de Datta, S.K. 1981. Principles and practices of rice production. John Wiley
and Sons. 619 pp.
[15] Moron, V., A. W. Robertson, and M. N. Ward, 2006: Seasonal predictability
and spatial coherence of rainfall characteristics in the tropical setting of
Senegal. Mon. Wea. Rev., 134, 3468–3482.
[16] SEAMEO,SEARCA, “Impact of Climate Change on the Philippine Rice
Sector: Supply/Demand Projections and Policy” retrieved from
http://www.searca.org/project-development/29-projects/rad-projects/1654-
impact-of-climate-change-on-the-philippine-rice-sector-supply-demand-
projections-and-policy.
[17] Satake, T., Yoshida, S., 1978. High temperature-induced sterility in Indica
rices at flowering. Jpn. J. Crop Sci. 47 (1), 6–17.
[18] Geng, S., Cady, C.W. (Eds.), 1991. Climatic Variation and Change:
Implications for Agriculture in the Pacific Rim. Proceedings, U.C. Davis,
California
[19] Riehl, H. 1954. Tropical meteorology. McGraw-Hill Book Co., Inc., New
York. 392 pp.
[20] Ana Ramirez, “The impact of climate change on rice production”.
[21] Philippine Atmospheric, Geophysical and Astronomical Services
Administration (PAGASA) and the Oscar M. Lopez Center for Climate
Change Adaptation and Disaster Risk Manage-ment Foundation, Inc. (Oscar
M. Lopez Center). 2015. State of the Philippine Climate 2015. November
2015. Available at http://www.omlopezcenter.org/
[22] Kaneda, C. 1972. Terminal report on studies on the breeding for cold
resistance. International Rice Research Institute, Los Baños, Philippines 80 pp.
(unpubl. mimeo.)
[23] Chung, G. S. 1979. The rice cold tolerance program in Korea. Pages 7–19 in
International Rice Research Institute. Report of a rice cold tolerance
workshop. Los Baños, Philippines.
[24] Gen-ichi Hirai, Toshikatsu Okumura et.,al “ Studies on the Effect of the Relative
Humidity of the Atmosphere on the Growth and Physiology of Rice Plants” .
pp. 1 2000.
[25] BC Ministry of Agriculture, Fisheries and Food. 1994. "Understanding
Humidity Control in Greenhouses Floriculture" Fact Sheet file no 400-5.
[26] Jürgen Schmidhuber, “Deep learning in neural networks: An overview”,
2015.
[27] Rumelhart, D.E., Hinton, G.E., Williams, R.J.: Learning representations by
backpropagating errors. Cognitive modeling 5, 3 (1988)
[28] P. Langley, H.A. Simon, “ Applications of machine learning and rule induction”
in Communications of the ACM, 38 (11) (1995), pp. 54-64
[29] Y. Bengio, A. Courville, and P. Vincent, "Representation Learning: A Review
and New Perspectives," IEEE Transactions on Pattern Analysis and Machine
Intelligence, vol. 35, pp. 1798-1828, 2013.
[30] Liu, J. N. K., Y. Hu, Y. He, P. W. Chan, and I.. Lai. "Deep Neural Network
Modelling for Big Data Weather Forecasting." Information Granularity, Big
Data, and Computational Intelligence. Springer International Publishing, 2015.
pp. 389-408.
[31] Alex Graves, “Supervised Sequence Labelling with Recurrent Neural
Networks”, pp. 31-32.

23
[32] F. A. Gers and J. Schmidhuber. LSTM Recurrent Networks Learn Simple
Context Free and Context Sensitive Languages. IEEE Transactions on Neural
Networks, 12(6):1333{1340, 2001.
[33] Rumelhart, D., G. Hinton, and R. Williams. "Learning internal representations
by error propagation," In Parallel Dist. Proc., pp. 318- 362. MIT Press, 1986.
[34] Man Galih Salman, Bayu Kanigoro and Yaya Heryadi, “Weather Forecasting
using Deep Learning Techniques”, 2015.
[35] Vahid Mansouri and Mohammad E. Akbari, “Efficient Short-Term Electricity
Load Forecasting Using Recurrent Neural Networks”, 2014.
[36] W. S. McCulloch and W. Pitts, "A logical calculus of the ideas immanent in
nervous activity," The bulletin of mathematical biophysics, vol. 5, pp. 115-
133, 1943.
[37] G. E. Hinton, S. Osindero, and Y.-W. Teh, "A fast learning algorithm for deep
belief nets," Neural Comput., vol. 18, pp. 1527-1554, 2006.
[38] D. Silver, A. Huang, C. J. Maddison, A. Guez, L. Sifre, G. van den Driessche,
et al., "Mastering the game of Go with deep neural networks and tree search,"
Nature, vol. 529, pp. 484-489, 01/28/print 2016.
[39] A. Lusci, G. Pollastri, and P. Baldi, "Deep architectures and deep learning in
chemoinformatics: the prediction of aqueous solubility for drug-like
molecules," Journal of chemical information and modeling, vol. 53, pp. 1563-
1575, 2013.
[40] Daniel Nations, “improve your understanding of web application” 2016,
Retrieved from https://www.lifewire.com/what-is-a-web-application-3486637.
[41] Dana Nourie, “ Java Technologies for web applications” 2006, Retrieved from
http://www.oracle.com/technetwork/articles/java/webapps-1-138794.html.
[42] Smriti Chand, “Cultivation of Rice: Suitable Conditions Required for the
Cultivation of Rice”, retrieved from
http://www.yourarticlelibrary.com/cultivation/cultivation-of-rice-suitable-
conditions-required-for-the-cultivation-of-rice-6-conditions/25491/.
[43] Campinas, SP, Brazil, “The development and evaluation of software to verify
diagnostic accuracy”, 2010.

24