Crop yield prediction using data mining techniques.pdf

© 2019, IJCSE All Rights Reserved 186
International Journal of Computer Sciences and Engineering Open Access
Review Paper Vol.-7, Issue-9, Sept 2019 E-ISSN: 2347-2693
Crop Yield Prediction Based on Data Mining Techniques: A Review
M. Saranya1*
, S. Sathappan2
1
Dept. of Computer Science, Erode Arts & Science College, Erode-9
2
Dept. of Computer Science, Erode Arts & Science College, Erode-9
*Corresponding Author: drmsaranyamca@gmail.com
DOI: https://doi.org/10.26438/ijcse/v7i9.186188 | Available online at: www.ijcseonline.org
Accepted: 12/Sept/2019, Published: 30/Sept/2019
Abstract—Agriculture is the main source of occupation which forms the backbone of our country. It involves the production of
crops which may be either food crops or commercial crops. The productivity of crop yield is significantly influenced by
various parameters such as rainfall, farm capacity, temperature, crop population density, humidity, irrigation, fertilizer
application, solar radiation, type of soil, depth, tillage and soil organic matter. An accurate crop yield prediction support
decision-makers in the agriculture sector to predict the yield effectively. Machine learning techniques and deep learning
techniques play a significant role in the analysis of data for crop yield prediction. However, the selection of appropriate
techniques from the pool of available techniques imposes challenges to the researchers concerning the chosen crop. In this
paper, an analysis has been performed on various deep learning and machine learning techniques. To know the limitations of
each technique, a comparative analysis is carried out in this paper. In addition to this, a suggestion is provided to further
improve the performance of crop yield prediction.
Keywords— Agriculture, crop yield prediction, productivity of crop yield, machine learning, deep learning.
I. INTRODUCTION
Agriculture is the backbone of India as well as the entire
world. The development of agricultural technologies is
occupied millions of years back and its advancement has
been encouraged and described through different
environments, civilizations, and technologies. The largest
agricultural productivity with the highest quality is achieved
by recent prediction techniques that provide the appropriate
growth environment under reproduction conveniences
protection circumstances. The modern growth of agricultural
activities is monitored by data mining techniques and
enhanced communication and information processes [1].
Generally, data mining is defined as the process of realizing
samples from large databases. The main objective of data
mining techniques is to mine the most important information
from the database and convert the information into a
reasonable formation for additional utilize.
Based on the agricultural applications, the data mining
techniques [2] in agriculture are varied. Few data mining
techniques are used to forecast environmental conditions
such as weather conditions, environmental pollutions and,
etc. Few more data mining techniques are used for
recognizing purposes such as weed detection, soil
characteristics and also used to monitor water cores. Even
though these techniques are helpful to predict crop yield,
crop yield prediction using data mining techniques is the
most significant issue in agriculture. The machine learning
technique is also used for crop yield prediction. Machine
learning is an application of artificial intelligence that
provides system ability to automatically learn and improve
from experience. In this paper, an overview of previous
researchers using supervised machine learning, unsupervised
machine learning, and deep learning is presented. The main
aim of this paper is to study detailed information about the
previous crop yield prediction using supervised machine
learning, unsupervised machine learning, and deep learning
techniques. Moreover, their limitations are addressed to
further improve the crop yield prediction.
The rest of the paper is structured as follows: Section II
presents the analysis of different machine learning and deep
learning techniques for crop yield prediction. Section III
presents the challenges in machine learning and deep
learning technique-based crop yield prediction. Section IV
presents the conclusion of this survey.
II. SURVEY ON CROP YIELD PREDICTION
TECHNIQUES
A. Crop Yield Prediction using Supervised Machine
Learning Techniques
A decision tree algorithm [3] was introduced for soybean
yield prediction. The decision tree induction technique was
adopted to develop innovative approaches to predict the
influence of climatic parameters on the predominant crop
productivity of the Bhopal district. A decision tree is a tree-

International Journal of Computer Sciences and Engineering Vol. 7(9), Sept 2019, E-ISSN: 2347-2693
like structure where each branch represented an outcome of
the test and node represented classes or class distributions. In
order to classify an unknown sample, the attribute value of
the samples was tested against the decision tree. The findings
of the decision tree were framed into different rules for a
better understanding of soybean productivity. The decision
tree algorithm predicts the crop yield in advance for market
dynamics.
A Nonlinear Autoregressive Neural Network (NARNN) [4]
was presented for crop yield forecasting. NARNN was a
widely used statistical forecasting model for time series. A
feed-forward neural network was used to establish NARNN
models in which the conventional function was replaced by
several neurons that worked together to implicitly
approximate the same functionality as a neural network. In
addition to this, Nonlinear Autoregressive with External
Input Neural Network (NARXNN) was introduced for crop
yield forecasting. It was a modified version of NARNN by
adding another relevant time series as an extra input to the
forecasting model. The NARNN has high accuracy in crop
yield forecasting.
A crop yield prediction approach [5] was proposed using a
parameter based adaptive neuro-fuzzy inference system
(ANFIS) model. In this approach, the crop yield was derived
by using attributes. The Neuro-fuzzy, multiple linear
regression and fuzzy logic techniques were used to predict
the wheat yield. This prediction was obtained based on
different input parameters are biomass, rain, radiation, water
and extractable soil. The pre-processing technique was
employed to remove the outliers, inconsistency, redundant
and missing values in the database. Then, Neuro-fuzzy,
multiple linear regression and fuzzy logic were applied to
predict the wheat yield. The ANFIS based crop yield
prediction approach has low RMSE value.
A smart farming system [6] was proposed to predict the crop
yield and suggested the optimal climatic factors to maximize
the crop yield. This system used support vector regression,
multivariate regression and random forest models to predict
the crop yield per acre. Initially, yield and weather data were
collected from the United States Department of Agriculture.
Then, the support vector regression, multivariate regression,
and random forest models were applied on the collected
dataset to choose the most suitable temperature and moisture
content at which the crop yield was optimal. The smart
farming system has less root mean absolute error.
B. Crop Yield Prediction using Unsupervised Machine
Learning Techniques
A crop yield prediction method [7] was introduced using the
Self-Organizing Map (SOM) and multi-layer feed-forward
neural network to predict the accurate crops suitable for
different kinds of soil. A pre-processing procedure was
applied to the large dataset to improve the exactness of the
mining process. It converted the raw data adapted for
clustering processes. Then, SOM was applied to soil data
which categorized soil data based on its geology. It helped to
predict the accurate crops suitable for different kinds of soil.
Finally, Multi-layer feed-forward neural network was applied
to determine the best match of crops for the given type of soil
characteristics after the data set was clustered. This crop
yield prediction method demonstrated tolerance to a
considerable degree over errors present in the training set.
Density-based clustering [8] technique was introduced for
crop yield prediction. This technique tried to split the crop
data into non-equal clusters based on the Euclidean distance
mathematical model. Several parameters were selected for
the number of independent factors. Then, the number of
clusters was picked to be divided upon the data. The data
points were grouped based on the Euclidean distance
between the cluster centre and the data points. The clustered
data were used to predict crop yield. This technique is
suitable for approximate predictions of crop yield.
Weighted-Self Organizing Map (W-SOM) [9] was
introduced for crop yield prediction in the Mysore region.
The W-SOM was a combination of SOM and Learning
Vector Quantization (LVQ). The prediction accuracy of W-
SOM was enhanced by minimizing the Within Class Error
(WCE) among the clusters. Hence, it provided a clear idea
about suitable crop cultivation in the Mysore region. The W-
SOM has effective performance.
A crop yield prediction model [10] was proposed using
Fuzzy C Means (FCM) clustering and neural network. The
FCM allowed one piece of data to belong to two or more
clusters that were based on the minimization of the objective
function. FCM calculated the output in terms of the degree of
membership. Fuzzy clustering dealt with assigning data
points in such a manner so that it got the close relationship of
similarity between data points as much as possible. The
clustered crop data by using FCM was given as input to the
neural network which predicted the yield of a crop at a
particular region. This model enhanced the basic leadership
in accuracy agribusiness.
C. Crop Yield Prediction using Deep Learning Techniques
The impact of the deep neural network [11] on predicting the
application rate of fertilizers for coconut trees was presented.
DNN is an artificial intelligence model which considered a
more efficient way to predict fertilizers under variety
cropping patterns. Initially, soil test reports were collected
from different places of Thrissur district, Kerala. The
collected data consisted of various parameters such as
Organic Carbon (%), Soil reaction (pH), Potassium (Kg /ha)
Total Soluble Salts (T.S.S) and Phosphorous (Kg /ha). These
parameters were processed by DNN to predict the crop yield.
The DNN based crop yield prediction has high accuracy.

International Journal of Computer Sciences and Engineering Vol. 7(9), Sept 2019, E-ISSN: 2347-2693
A deep learning methodology called deep convolutional neural
network [12] was introduced for crop yield prediction. Initially,
Normalized Difference Vegetation Index (NDVI) and RGB data
were collected from Unmanned Aerial Vehicles (UAVs). Nine
crop fields located in the vicinity of the city of Pori was selected
for this study. These nine fields divided into overlapping data
frames of sizes 10m, 20m and 40m. Then a dedicated holdout
test data was then constructed from 15% of shuffled data frames
and the remaining data frames were used to train the models
with k-fold cross-validation. The test data were given as input to
deep CNN which predicted the crop yield. The deep CNN has
better performance.
A deep neural network [13] was introduced for crop yield
prediction. It found the underlying representation of data
without handcrafted input of features. DNN had multiple
stacked non-linear layers that converted the raw input data
(genotype and environmental information) into a higher and
more abstract representation at each stacked layer. More
complex features were extracted when the DNN network grown
deeper. The DNN learned the non-linear and complex
relationship between genes, environmental conditions as well as
their interactions from historical data. It made a reasonably
accurate prediction of crop yield. The DNN based crop yield
prediction has superior prediction accuracy.
III. ANALYSIS OF CROP YIELD PREDICTION
TECHNIQUES
This section presents a detail about the demerits of different
crop yield prediction techniques whose functional information is
discussed in the previous section. Through the review of
different crop yield predictions using supervised machine
learning techniques, unsupervised machine learning techniques,
and deep learning techniques, the following challenges are
addressed. The decision tree algorithm requires more storage
space. The user must be careful when using either NARNN or
NARXNN for crop yield forecasting since there is any
inconsistency between the results of training and forecasting.
The ANFIS based crop yield prediction has strong
computational complexity restrictions which is the major
drawback of this method. In a smart farming system, more
features would be added to improve the crop yield prediction.
The major disadvantage of SOM based crop yield prediction is
SOM relied on pre-defined distance in feature space. The
density-based clustering is not much more efficient. The
prediction rate of W-SOM could be increased by a combination
of parallel layer regression along with a deep belief network
strategy. The proper selection of a membership function is more
difficult in FCM. The DNN based crop yield prediction
approach will be extended to monitor and handle the problems
likely to be created by the collection of advanced data from a
wider area. In deep CNN, CNN will be trained on a large set of
features along with times series data to tune the trained model
for accuracy. Although DNN captures parameter and yield
interactions, its complex model structure makes it hard to
produce testable hypotheses that could potentially provide
biological insights.
IV. CONCLUSION
In this article, a detailed review of crop yield prediction based
on various techniques was presented. According to this
analysis, it is known that the deep neural network [13] based
crop yield prediction has better performance than other
techniques. However, a major limitation of the deep neural
network [13] based crop yield prediction technique is its black
box property, which is shared by many machine learning
methods. Moreover, the complex model structure of DNN
makes it hard to produce testable hypotheses that could
potentially provide biological insights. The limitation has to
overcome this by looking for more advanced models that are not
only more accurate but also more explainable.
REFERENCES
[1] B. Milovic, V. Radojevic, “Application of data mining in agriculture”,
Bulgarian Journal of Agricultural Science, Vol.21, Issue.1, pp.26-
34,2015.
[2] D. Ramesh, B. V. Vardhan, “Data mining techniques and applications
to agricultural yield data”, International Journal of Advanced
Research in Computer and Communication Engineering, Vol.2,
Issue.9, pp.3477-80, 2013.
[3] S. Veenadhari, B. Mishra, C. D.Singh, “Soybean productivity
modelling using decision tree algorithms” International Journal of
Computer Applications, Vol.27, Issue.7, pp.11-15,2011.
[4] W. W. Guo, H. Xue, “Crop yield forecasting using artificial neural
networks: A comparison between spatial and temporal models”,
Mathematical Problems in Engineering, 2014.
[5] Shastry, H. A. Sanjay, M. Hegde, “A parameter based ANFIS model
for crop yield prediction” IEEE International Advance Computing
Conference (IACC), pp.253-257, 2015.
[6] Shah, A. Dubey, V. Hemnani, D. Gala, D. R. Kalbande, “Smart
Farming System: Crop Yield Prediction Using Regression
Techniques”. In Proceedings of International Conference on Wireless
Communication, Springer, Singapore, pp. 49-56, 2018.
[7] F. N. Ogwueleka, “Crop growth prediction using self-organizing map
and multilayer feed-forward neural network” American-Eurasian
Journal of Sustainable Agriculture, Vol.5, Issue.2, pp.168-176, 2018.
[8] B.V. Vardhan, D. Ramesh, “Density based clustering technique on
crop yield prediction”, International Journal of Electronics and
Electrical Engineering, Vol.2, Issue.1, pp.56-59, 2014.
[9] P. Mohan, K. K. Patil, “Weather and Crop Prediction Using Modified
Self Organizing Map for Mysore Region”, International Journal of
Intelligent Engineering & Systems (IJIES), Vol.11, Issue.2, pp.192-
199, 2017.
[10] A.Verma, A. Jatain, S.Bajaj, “Crop yield prediction of wheat using
Fuzzy C Means clustering and neural network”, International Journal
of Applied Engineering Research, Vol.13, Issue.11, pp.9816-9821,
2018.
[11] S. Suchithra, M. L. Pai, “Impact of Deep Neural Network on
predicting application rate of fertilizers (Focus on coconut trees of
Kerala northern coastal plain agro ecological unit)”. International
Journal of Pure and Applied Mathematics, Vol.119, Issue.10, pp.451-
466,2017.
[12] P. Nevavuori, N. Narra, T. Lipping, “Crop yield prediction with deep
convolutional neural networks” Computers and Electronics in
Agriculture, Vol.163, 2019.
[13] S. Khaki, L. Wang, “Crop yield prediction using deep neural
networks”, Frontiers in plant science, Vol.10,2019.

Crop yield prediction using data mining techniques.pdf

More Related Content

Similar to Crop yield prediction using data mining techniques.pdf (20)

Recently uploaded (20)

Crop yield prediction using data mining techniques.pdf