Groundnut Crop Yield Prediction Using Machine Learning Techniques

Abstract

Yield prediction is a very important agricultural problem. Any farmer is interested in knowing how much yield he is about to expect. In the past, yield prediction was performed by considering farmer's experience on particular field and crop. Based on previous data, we can predict crop yield using machine-learning technique. Crop yield prediction is an important area of research, which helps in ensuring food security all around the world. We analyzed result of multiple linear Regression, Regression Tree, K-nearest Neighbor and Artificial Neural Network on Groundnut data of previous 8 years. We have done prediction based on Soil, Environmental and Abiotic attributes. KNN algorithm gives better result compared to other algorithms for Groundnut crop yield prediction.

Full text

CSEIT1835254 | Received : 15 June 2018 | Accepted : 26 June 2018 | May-June-2018 [ (3)5 : 1093-1097 ]
International Journal of Scientific Research in Computer Science, Engineering and Information Technology
© 2018 IJSRCSEIT | Volume 3 | Issue 5 | ISSN : 2456-3307
1093
Groundnut Crop Yield Prediction Using Machine Learning
Techniques
Vinita Shah*1, Prachi Shah2
*1Assistant Professor, Information Technology, G H Patel College of Engineering and Technology, V V Nagar,
Gujarat, India
2Assistant Professor, Department of Information Technology, BVM Engineering College, V V Nagar, Gujarat,
India
ABSTRACT
Yield prediction is a very important agricultural problem. Any farmer is interested in knowing how much yield
he is about to expect. In the past, yield prediction was performed by considering farmer's experience on
particular field and crop. Based on previous data, we can predict crop yield using machine-learning technique.
Crop yield prediction is an important area of research, which helps in ensuring food security all around the
world. We analyzed result of multiple linear Regression, Regression Tree, K-nearest Neighbor and Artificial
Neural Network on Groundnut data of previous 8 years. We have done prediction based on Soil, Environmental
and Abiotic attributes. KNN algorithm gives better result compared to other algorithms for Groundnut crop
yield prediction.
Keywords : Crop analysis; Yield prediction; K-means; K-NN; Multiple Linear regression
I. INTRODUCTION
Data mining is the process of extracting useful
knowledge or information from large amount of data.
In digital generation, data mining is becoming an
increasingly important tool to transform data into
information. When the Data mining techniques are
used with agriculture data, the term is known as
precision agriculture. The main aim of the work is to
improve and substantiate the validity of yield
prediction, which is useful for the farmers.
Agricultural crop production depends on various
factors such as biology, climate, economy and
geography. Several factors have different impacts on
agriculture. So previous year‟s researchers used
appropriate statistical methodologies. A large number
of variables can affect agronomic traits such as yield.
Yield prediction is a very important agricultural
problem. Any farmer is interested in knowing how
much yield he is about to expect. In the past, yield
prediction was performed by considering farmer's
experience on particular field and crop. Consider that
data are available for some time back to the past,
where the corresponding yield predictions have been
recorded. In any of Data Mining 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. Crop yield
prediction is an important area of research, which
helps in ensuring food security all around the world.
In our research, we have considered the effects of
environmental (weather), biotic (pH, soil salinity)
and area of production as factors towards crop
production in Gujarat. Considering these factors as

Volume 3, Issue 5, May-June-2018 | http:// ijsrcseit.com
Vinita Shah et al. Int J S Res CSE & IT. 2018 May-June; 3(5) : 1093-1097
1094
datasets for various districts, we proposed a new
approach that divides the yearly dataset in to
seasonally basis. Then we applied some clustering
techniques and suitable classification techniques to
obtain crop yield predictions.
It is explained as follows: Section 1 introduces the
area. Section 2 provides some related work in Crop
yield prediction. Section 3 explains motivations for
the work. Section 4 presents the Real dataset use for
our work. Section 5 presents methodology of the
work. Section 6 gives the analysis of proposed
approach.
II. MOTIVATION
Gujarat has been described by hot semi-dry
conditions, which has 49% of the aggregate
topographical range as developed grounds (9.6
million ha). The inundated zone is just 32% of the
aggregate developed territory. Rest of the range
confronts vast water deficiency. The ground waters
are of low quality. The alluvial soils are basic/saline
and lacking in nitrogen, phosphorus and zinc. The
editing force is low at 118%. The aggregate
sustenance grains creation of the state is 5.26 million
tons. The critical yields are groundnut, cotton, pearl
millet, maize, sorghum, castor, gram and mustard.
A farmer must have a good understanding of the soil
type, the biotic factors governing the soil and a
thorough knowledge about the traditional
agricultural practices to gain maximum crop yield.
Such practice may include harrowing and ploughing
using inputs such as fertilizers, insecticides and
herbicides [7].
It would be seen from the Fig.1 that though Gujarat
stands first in area and production but productivity
wise Tamil Nadu state stands first in the country.
However, Gujarat state groundnut production
depends on September rains. In case September rains
are not there in Gujarat, the production and
productivity will go down automatically. Therefore,
it is important to find the relation between climate
variable, rainfall and yield data.
Fig. 1.Statewise Area Production and production of
Groundnut [14]
III. PREPROCESSING
A. Dataset
In this research, two types of datasets are used:
Weather data and Yield data. The Weather dataset
used in this research has been collected from
Department of Metrological Centre, Anand
Agriculture University (AAU), Anand and Yield data
are taken from 4 different Yearbooks of Directorate
of Agriculture Gujarat State, Gandhinagar for four
different district of Gujarat which are Jamnagar,
Junagath, Rajkot, Amreli. A lot of pre-processing was
required to handle missing values, noise and outliers.
We considered different 20 attributes for this
research: rainfall, maximum and minimum
temperature, Vapour Pressure, Relative humidity,
Basic Sunshine, Evaporation Pressure, Soil
Temperature at different depth, Wind Speed,
irrigated area for all districts; and cultivated area for
crop yield considered according to the districts.
After the necessary formatting and pre-processing of

Volume 3, Issue 5, May-June-2018 | http:// ijsrcseit.com
Vinita Shah et al. Int J S Res CSE & IT. 2018 May-June; 3(5) : 1093-1097
1095
The datasets, the finalized version of our data
contains
Total 4 districts for the time periods of 2006 to 2013.
B. Input Variables and Measures
From the vast initial dataset, we selected a limited
number of important input variables that is 20, which
have the highest contribution to agricultural
production. All the inputs were considered for the
eight-year periods of 2006 to 2013.
There are various measures to different prediction
models:
1. RSquare [Coefficient of determination] is
simply the square of the sample correlation
coefficient (i.e., r) between the outcomes and
their predicted values. The coefficient of
determination ranges from 0 to 1.
2. RMSE [Root Mean Square Error] is a
frequently used measure of the difference
between values predicted by a model and the
values actually observed from the
environment that is being modeled.
IV. METHODOLOGY
A. Proposed Flow
The existing papers involves less number of attributes
of Climate data. Instead of that, we will use more
number of attributes including soil attributes which
had not been considered in previous work. Also,
instead of yearly approach, we have done on
seasonally basis approach to get more accurate results.
For seasonally basis approach, we divided total
dataset in to two seasons 1.Kharif And 2.Summer as
groundnut production are more in this two seasons.
For that purpose, we created GUI in java which
divides monthly data sets into season wise cluster
data and also combine the associated yield dataset of
that particular season for all 4 districts. Selection is
done by season wise and generates the dataset
according to its climatic attributes average and
fetches its yield from dataset and then generates
result for each year. So basically, it combines 2
datasets and also divides them into season wise of
kharif and summer and displays the output results in
GUI of each district.
B. Individual effect of Attributes on yield
To find the individual attribute effect on yield
correlation is perform on the data. In addition, values
of „r‟ and “R square” are generated. From that, we can
see that which attribute affect more on yield.
From the values of R Square, it clearly sees that above
values of 0.5 depends more on yield and are
important attribute to grow of plant so we have
considered it.
C. Combined effect of Attributes on yield
We have taken 4 different alternatives from which
the different models accuracy measures can be found.
A. Environmental Attributes:
Rainfall, Maximum Temperature, Minimum
Temperature, Basic Sunshine, Relative humidity(2
times), Vapor Pressure(2 times), Evaporation Pressure
B. Soil Attributes:
Temperature from the different depth of Soil (6)
C. Abiotic Attributes:
Water Content (2), Density (2), Wind Speed
D. Area Central Attributes:
Total area from which the yield is produced and total
production
The following classification/regression models were
used to obtain the crop yield prediction results:

Volume 3, Issue 5, May-June-2018 | http:// ijsrcseit.com
Vinita Shah et al. Int J S Res CSE & IT. 2018 May-June; 3(5) : 1093-1097
1096
a) Linear Regression: It is a statistical measure that
can be used to determine the strength of the
relationship between one dependent variable and a
series of other changing variables known as
independent variables (regular attributes). If
independent variable contains multiple input
attributes like in our research (rainfall, sunshine
hours, humidity, pH etc), then it is termed as
multiple linear regressions. Linear regression
provides a model for the relationship between a
scalar variable and one or more explanatory variables.
This is done by fitting a linear equation to the
observed data [7].
b) k-NN: The k-nearest neighbor algorithm compares
a given test example with training examples which
are similar. Each example denotes a point in an
ndimensional space. Thus, all of the training
examples are saved in an n-dimensional pattern space.
K is a positive integer, usually small. For our purpose,
the basic k-NN algorithm was applied. It first finds
the k examples from the training set that are closest
to the unknown example. Then it takes the most
common occurring classification for the k examples
[7].
c) Neural Network: An artificial neural network
(ANN) is a mathematical model or computational
model inspired by the structure and functional
aspects of biological neural networks for instance in
our brains. In most cases an ANN is an adaptive
system that modifies its structure based on external
or internal information that flows through the
network during the learning phase. The basic neural
network model consists of three layers: the input
layer, the hidden layer and an output layer [7].
d) Regression Tree: This are commonly used in data
mining with the objective of creating a model that
predicts the value of a target (or dependent variable)
based on the values of several input (or independent
variables).In Regression Trees target variable is
continuous and tree is used to predict its value.
V. RESULT ANALYSIS
In our research, we determined prediction results for
yields of Groundnut crop for the selected districts in
Gujarat. The predictions results were obtained
according to the selected input attributes using
appropriate classification and regression models.
The tool used to provide predictive algorithms,
include powerful methods to explore, transform and
“clean” data, partition data into “training” and
“validation” sets, train one or more machine
learning models, and evaluate and compare the
performance of those models.
For prediction, four different algorithms are used.
 Multiple linear Regression
 Regression Tree
 K-nearest Neighbor
 Artificial Neural Network
Fig. 2. Comparison of prediction algorithms
RMSE values are compared of four different methods
for all four different alternatives which are discussed
in figure 2.
Fig 3.Comparative Analysis of four-prediction model
with four approaches

Volume 3, Issue 5, May-June-2018 | http:// ijsrcseit.com
Vinita Shah et al. Int J S Res CSE & IT. 2018 May-June; 3(5) : 1093-1097
1097
VI. CONCLUSION AND FUTURE WORK
Weather data has a major role in affecting the
Agriculture data of crop yield. For ground nut crop
yield prediction, K-nearest neighbor for prediction
model gives better Results in training as well as
validation for the data then other three techniques
which are Artificial Neural network and two
regression techniques multiple linear regression(MLR)
and Regression tree. Apart from that, if using larger
dataset of previous years can also be affect the
accuracy of algorithms and can get better accuracy or
less RMSE value in future.
VII. REFERENCES
[1] MucherinoPetraqPapajorgji, P. M. Pardalos,”A
survey of data mining techniques applied to
agriculture”, Springer, 2009.
[2] S.S.Bhaskar, L.Arokiam, V.Arul Kumar,
L.Jeyassimaan,”A Brief Survey Of Data Mining
Techniques To agriculture Applications”,
MadwellJournels, 2010.
[3] Ramesh A. Medar,Vijay. S. Rajpurohit,”A
survey on Data Mining Techniques for Crop
Yield Prediction”,IJARCSMS, 2014
[4] Hui Chen,WeiWu,Hong-Bin Liu,“Assessing the
relative importance of climate variables to rice
yield variation using support vector
machines”,Springer,2015.
[5] D Ramesh, B Vishnu Vardhan,”Analysis Of
Crop Yield Prediction Using Data Mining
Techniques”, IJRET, 2015.
[6] D Ramesh, B Vishnu Vardhan, “Crop Yield
Prediction Using Weight Based Clustering
Technique “, IJCEA, 2015.
[7] A.T.M ShakilAhamed, NavidTanzeem
Mahmood, Nazmul Hossain, Mohammad
TanzirKabir, Kallal Das, Faridur Rahman,
Rashedur M Rahman,”Applying Data Mining
Techniques to Predict Annual Yield of Major
Crops and Recommend PlantingDifferent Crops
in Different Districts in Bangladesh”,IEEE,2015
[8] D. Ramesh,B Vishnu Vardhan, O
SubhashChanderGoud,” Density Based
Clustering Technique on Crop Yield
Prediction”IJEEE,2014
[9] Mohammad MotiurRahman,NaheenaHaq,
Rashedur M Rahman,”Application of Data
Mining Tools for Rice Yield Prediction on
Clustered Regions of Bangladesh”,IEEE,2014
[10] D Ramesh, B Vishnu Vardhan,”Data Mining
Techniques and Applications to Agricultural
Yield Data”, IJARCCE, 2013
[11] José R. Romero , Pablo F. Roncallo , Pavan C.
Akkiraju , Ignacio Ponzoni , Viviana C.
Echenique, Jessica A. Carballido,”Using
classification algorithms for predicting durum
wheat yield in the province of Buenos Aires”,
ELSVIER, 2013.
[12] YunousVagh, JitianXiao, "A data mining
perspective of dual effect of Rainfall and
Temperature on Wheat Yield", ECU, 2012
[13] Parekh, F. P, Suryanarayana, T. M. V ,”Impact
of Climatological Parameters on Yield of Wheat
Using Neural Network Fitting”,International
Journal of Modern Engineering Research ,2012
[14] Ye, Nong; Data Mining: Theories, Algorithms,
and Examples, CRC Press, 2013.
[15] Data mining Concepts and Techniques; Jiawei
Han and MichelineKamber; Second Edition,
Morgan kaufmann publishers.
[16] https://dag.gujarat.gov.in/images/directorofagric
ulture/pdf/Groundnut-Book.pdf