Professional Documents
Culture Documents
Use of Machine Learning: Prediction of Crop Yield in Madhya Pradesh, India
Use of Machine Learning: Prediction of Crop Yield in Madhya Pradesh, India
Use of Machine Learning: Prediction of Crop Yield in Madhya Pradesh, India
Abstract
Machine learning in agriculture is a very novel research, consists in the application of machine
learning techniques to agriculture. Machine learning in agriculture field is a relatively research field.
In this article describe an overview of machine learning techniques applied to agricultural and their
applications to agricultural related areas. Yield prediction is a very important agricultural problem.
Any farmer is interested in knowing how much yield he is about to expect. Previously yield prediction
was performed by considering farmer's experience on particular field and crop. In any of machine
learning procedures the training data is to be collected from some time back to the past and the
gathered data is used in terms of training which has to be exploited to learn how to classify future
yield predictions. Madhya Pradesh is the heart land of the country and Agriculture is the backbone of
its Economy. Majority of the farmers are not getting the expected crop yield due to several reasons.
The agricultural yield is primarily depends on weather conditions. Rainfall conditions also influences
the rice cultivation. In this context, the farmers necessarily requires a timely advice to predict the
future crop yield and an analysis is to be made in order to help the farmers to maximize the crop
production in their crops. Yield prediction is an important agricultural problem. Every farmer is
interested in knowing, how much yield he is about expect.
The State is divided in 11 Agro climatic Zones and 5 crops zones which are, (1) Chhattisgarh plains (2) Northern hill
region of Chhattisgarh (3) Kymore plateau and Satpura hills (4) Central Narmada Valley (5) Vindhya Plateau (6) Gird
region (7) Bundel Khand (8) Satpura Plateau (9) Malwa Plateau (10) Nimar Plains (11) Jhabua hills.
Agro ecologically the state falls in Zone - VII (Eastern Plateau and hills Zone), Zone VIII (Central plateau and hills Zone)
and Zone IX (Western Plateau and hills Zone).
Major perennial rivers of the State are Mahi, Narmada, Rapti, Chambal, Betwa, Sone, Wanganga, Ken and Pench. The
rivers originate in M.P. and flow towards the bordering States.
1
CAS, University Institute of Technology, Rajiv Gandhi Proudyogiki Vishwavidyalaya , Bhopal, India.
2
Deputy Director, Department of Agriculture, Government of Madhya Pradesh, Bhopal, Madhya Pradesh, India.
3
Department of Computer Science and Engineering, National Institute of Technology, Raipur, Chhattisgarh, India.
Correspondence: Mr. Rishi Rai, National Institute of Technology, Raipur, Chhattisgarh, India.
The State has about 61.41% area under rain fed farming The agricultural yield is primarily depends on weather
area where agriculture depends on monsoon. Erratic conditions In this context, the farmers necessarily
and uneven distribution of rainfall is the major requires a timely advice to predict the future crop
constraint for achieving targeted production. Due to productivity and an analysis is to be made in order to
failure of rains, drought condition is created every year help the farmers to maximize the crop production in
in one part or the other. their crops. Yield prediction is an important agricultural
problem. Every farmer is interested in knowing, how
The major Kharif crops are: - Paddy, Jowar, Maize, Bajra, much yield he is about expect.
Tur, Urad, Moong, Soybean, Groundnut, Cotton etc. The
major Rabi crops are: - Wheat, Gram, Lentil, Peas, Overview of Data
Mustard, and Linseed etc.
The data used for this article are collected from
Machine learning, the extraction of hidden predictive Agriculture economic survey for the years from 2004 to
information from large databases, is a powerful new 2016 for Madhya Pradesh. The data are taken in eight
technology with great potential to help Farmers and input variables. The variables are 'Year', 'Rainfall', 'Area
companies focus on the most important information in of Sowing', 'Fertilizers' and 'Pesticide'.
their data warehouses. Machine Learning tools predict
future trends and behaviours. The automated, The attribute 'Year' specifies the year in which the data
prospective analyses offered by Machine Learning move are available in Hectares. 'Rainfall' attribute specifies the
beyond the analyses of past events provided by actual rainfall in the specified year in millimetres. 'Area
retrospective tools typical of decision support systems. of Sowing' attribute specifies the total area sowed in the
Machine Learning tools can answer questions that specified year in Hectares. 'Fertilizers' specify in Tons in
traditionally were too time consuming to resolve. the specified year. 'Fertilizers' specify in Tons in the
specified year 'Pesticides' specify in Tons in the specified
Machine learning techniques can be implemented year
rapidly on existing software and hardware platforms to
enhance the value of existing information resources, Methodology
and can be integrated with new products and systems
as they are brought on-line. When implemented on high In this article Multiple Linear Regression technique was
performance client/server or parallel processing used for the estimation of crop yield analysis. Multiple
computers, Machine Learning tools can analyze massive linear regression denotes to model the relationship
databases to deliver answers to questions. between two or more explanatory variables and a
response variable by fitting a linear equation to
Machine learning is the process of analyzing data from observed data.
different perspectives and summarizing it into useful
information - information that can be used to increase A regression model involves more than one predictor
revenue, cuts costs, or both. Machine Learning software variable is called Multiple Regression Model. Multiple
is one of a number of analytical tools for analyzing data. Linear Regression (MLR) is the method, used to model
It allows users to analyze data from many different the linear relationship between a dependent variable
dimensions or angles, categorize it, and summarize the and one or more independent variables. The dependent
relationships identified. Technically, machine learning is variable is sometimes termed as predictant and
the process of finding correlations or patterns among independent variables are called predictors. This crop
dozens of fields in large relational databases. Machine yield prediction model is presented with the use of
learning mainly consists of five elements: multiple linear regression technique where the
predictant is the yield and there are five predictors
• Extract, transform, and load transaction data onto namely Year, Rainfall, Area of Sowing, and Fertilizers
the data warehouse system. and pesticides.
• Store and manage the data in a multidimensional
database system. Results and Discussion
• Provide data access to business analysts and
information technology professionals. In this article an effort is made in order to know the crop
• Analyze the data by application software. yield analysis and it is processed by implementing
• Present the data in a useful format, such as a graph Multiple Linear Regression technique. In this article used
or table. data was overview of Madhya Pradesh. The actual value
42
J. Adv. Res. Appl. Arti. Intel. Neural Netw. 2017; 4(1&2) Shukla P et al.
along with the corresponding estimated value using Multiple Linear Regression technique which are ranging
Multiple Linear Regression technique for 12 years between -10% and +13%.and revealed that prediction
shown in the Table 1. The estimated yield results using has 95 percent accuracy.
Table 1.Actual Production and Predicted Values Using Multiple Linear Regression Technique
Year Yield (actual) Yield (prediction) Difference ratio
2004 1200 1094.35 0.09
2005 1086 1063.50 0.02
2006 1103 1187.49 -0.08
2007 1034 1137.22 -0.10
2008 1151 1162.51 -0.01
2009 1220 1237.17 -0.01
2010 1216 1318.92 -0.08
2011 1485 1475.82 0.01
2012 1750 1521.51 0.13
2013 1671 1661.60 0.01
2014 1778 1741.75 0.02
2015 1729 1821.17 -0.05
43
Shukla P et al. J. Adv. Res. Appl. Arti. Intel. Neural Netw. 2017; 4(1&2)
44