International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056

Volume: 07 Issue: 02 | Feb 2020 www.irjet.net p-ISSN: 2395-0072

Crop Prediction System using Machine Learning Algorithms

Pavan Patil1, Virendra Panpatil2, Prof. Shrikant Kokate3
1Pavan Patil SPPU, (Pimpri Chinchwad College of Engineering)
2Virendra Panpatil SPPU, (Pimpri Chinchwad College of Engineering)
3Professor Shrikant Kokate, Dept. of Computer Engineering, PCCOE, Maharastra, India

---------------------------------------------------------------------***----------------------------------------------------------------------
Abstract - As we are aware of the fact that, most of Indians life. However, within the current time, folks tend to require a
have agriculture as their occupation. Farmers usually have the lot of like a shot appreciated jobs. There are fewer, and fewer
mindset of planting the same crop, using more fertilizers and folks concerned in crop cultivation. additionally, the
following the public choice. By looking at the past few years, continual increase of human population makes the
there have been significant developments in how machine cultivation of the crops at the proper time and right place
learning can be used in various industries and research. So we even a lot of vital, because the climate is dynamic and
have planned to create a system where machine learning can therefore the shifts from traditional weather pattern are a lot
be used in agriculture for the betterment of farmers. The of frequent than before manufacture. Food insecurity may be
surveyed research papers have given a rough idea about using a drawback that can't be avoided, and humans should build
ML with only one attribute. We have the aim of adding more use of latest innovative technologies to create use of existing
attributes to our system and ameliorate the results, which can soil, water and air conditions to get larger crops. The
improve the yields and we can recognize several patterns for information gap between ancient ways that of cultivating
predictions. This system will be useful to justify which crop can and new agricultural technologies may be overcome if the
be grown in a particular region computer code may be designed to model the interactive
impact of climate factors, particularly the impact of
Key Words: Machine Learning in Agriculture, maximum events (e.g. heat, rainfalls and excess water)
Classification Algorithms, Decision Tree, KNN occurring at completely different growing phases of crops.
The temperature change undoubtedly affects the native and
1. INTRODUCTION world food production, therefore planning computer code to
model crop predictions needs new methodology for
Crop production may be a complicated development that's
temperature change studies, situations for temperature
influenced by soil and environmental condition input
change adaptation, and policymakers which will limit the
parameters. Agriculture input parameters vary from field to
devastating effects of weather on food provide. Experimental
field and farmer to farmer. Collection such info on a bigger
proof is employed to form environmental condition zones
space may be a discouraging task. However, the
that have seen changes in weather and water, the 2 most
environmental condition info collected in Republic of India
significant factors in guaranteeing a in crop. The soil sort will
at each 1sq.m space in numerous components of the district
modification over time because of weather and pests,
is tabulated by Indian meteoric Department. The massive
therefore crop management must manage a fancy quantity of
such knowledge sets may be used for predicting their
information, directly or indirectly associated with one
influence on major crops of that individual district or place.
another. It will therefore by considering a simplified reality,
There are completely different foretelling methodologies
to permit a quick assessment of the impact of temperature
developed and evaluated by the researchers everywhere the
change in agriculture. Agriculture should adapt to those
globe within the field of agriculture or associated sciences. A
climate changes, and it will do therefore by developing
number of such studies are: Agricultural researchers in
models which will in theory optimize management practices,
alternative countries have shown that tries of crop yield
maximize the rotations of the new crop to manage the
maximization through pro-pesticide state policies have LED
changes of soil, novel breeding programs. By maximizing the
to hazardously high chemical usage. These studies have
worth of foretelling, the seasonal climate changes may be
reported a correlation between chemical usage and crop
ascertained and recorded in an exceedingly timely manner.
yield [1]. Agriculture is associate trade sector that's
Later on, by victimization computer code supported machine
benefiting powerfully from the event of detector technology,
learning, one will timely assess the temperature change
knowledge science, and machine learning (ML) techniques
impact and check attainable situations that incorporate
within the latest years. These developments return to satisfy
ascertained changes in climatic conditions and water
environmental and population pressures round-faced by our
distribution. data {processing} is that the process of
society, wherever reports indicate a requirement for robust
analyzing the experimental knowledge collected over a
international agriculture yield increase to produce food for a
amount and varied locations from completely different
growing population on a hotter planet. Most of the work
views, extract trends or patterns {of data of knowledge of
tired the sector of yield foretelling via cubic centimeter
info} and switch them into helpful information for users.
makes use of some kind of remote sensing knowledge over
Users will then additionally reason and/or summarize the
the farm. Agriculture seeks to extend and improve the crop
relationships ascertained from the collected knowledge, and
yield and therefore the quality of the crops to sustain human

© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 748
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 02 | Feb 2020 www.irjet.net p-ISSN: 2395-0072

typically predict what knowledge to expect. Machine- yield graphs, the best time of sowing, plant growth and
learning techniques are a part of data processing and harvesting of the plant can be found out. Also, the optimal and
knowledge exploration and focus exclusively on worst environmental condition can also be incurred. The
characteristic correlations or patterns among massive model focuses on all type of farms, and smaller farmers can
datasets or massive relative databases. The patterns, also be benefitted. This model can be further enhanced to find
associations, or relationships among all this knowledge will the yield of every crop, and for pesticide recommendation.
additional be reborn into information that's offered to the Also, it can be modified to suggest about the fertilizers and
user as historical patterns and future trends. This irrigation need for crops.
information provided by machine learning will facilitate
farmers with crop cultivation by predicting probabilities of 3. Related Work
crop losses or stop losses altogether.
1. A Scalable Machine Learning System for Pre-Season
2. Literature Review Agriculture Yield Forecast:

Nowadays many experts are applying automated farming. The system projected during this work is created by a neural
Since Decision Tree is an well-known algorithm it was used network wherever inputs area unit treated on an individual
for prediction which is a supervised learning algorithm and basis. Static soil information in handled by fully-connected
multiple linear regression which is generalized prediction layers whereas dynamic meteorological information is
model. An attempt has been made to research the influence handled by continual LSTM layers. This explicit design was
made by decision tree induction technique of climatic trained with historical information for many soil properties,
parameters on soybean productivity. For easy understanding precipitation, minimum and most temperature against
of end-user different kind of rules were created from the historical yield labels at county level. When training, the
Decision tree. The paper from Md. Tahmid Shakoor & co model was tested in an exceedingly separate information set
paper helped us for selecting various attributes like land and showed comparable results with existing yield
capability classification, soil depth, slope, drainage, texture, prognostication ways that create use of in-depth remote
erosion, and permeability [4]. Two supervised classification sensing data. the most important lesson learnt from our
machine learning algorithms has been implemented in this experiments is that it's attainable get ascendable yield
study. Our system takes the necessary weather and soil forecast as a result of the projected neural network model
properties data for a given coordinate automatically from an will notice and exploit redundant info each within the soil
appropriate source. Another advantage is that their system and within the weather information. To boot, the model
worked on large regions, and provides forecasts at a might be able to learn AN implicit illustration of the cycles of
resolution compatible with best input data resolution, which the crops evaluated during this paper, considering the
in the case is originally from the soil data. The ability of seasonal atmospherically information used as input.
forecasting crop before the beginning of the crop season. This
provides users with the capability to perform strategy 2. Machine learning approach for forecasting crop yield
changes, like choosing a more robust genetic variation before based on climatic parameters
planting or even changing the crop type, in order to
The present study provides the potential use of information
accommodate for extreme climatic variations further ahead
mining techniques in predicting the crop yield supported the
in the crop cycle [2].
environmental condition input parameters. The developed
The algorithm developed introduces a data-driven model to webpage is user friendly and therefore the accuracy of
predict and forecast crop yield using joint dependencies of predictions square measure higher than seventy-five per cent
soil and climate features. Although there are several all told the crops and districts designated within the study
techniques existing to obtain rainfall predictions, the indicating higher accuracy of prediction. The user-friendly
algorithm discussed in this paper succeeded in emphasizing web content developed for predicting crop yield may be
on Rainfall along with the crop yield prediction. This utilized by any user their alternative of crop by providing
designed model took into account the most relevant environmental condition knowledge of that place.
environment as well as soil parameters that affect the crop
3. Crop Prediction on the Region Belts of India: A Naïve Bayes
growth, in a way that each of those parameters received
MapReduce Precision Agricultural Model
equal weight in the final prediction. The outcomes of this
research can benefit the agriculturists/farmers by knowing The planned work introduces efficient degree economical
the investment capital on the crop to be sown, even before crop recommendation system. Use of naïve mathematician
the sowing season begins. The predictive pattern of the makes the model terribly economical in terms of
algorithm can benefit local self-government and financial computation. The system is scalable because it may be wont
institutions to allocate suitable funds or fiscal loans to to take a look at on totally different crops. From the yield
farmers. Naive Bayes is used for the large dataset can also be graphs the simplest time of sowing, plant growth and gather
beneficial. Use of naïve Bayes and decision tree makes the of plant may be known. Conjointly the best and worst
model very efficient in terms of computation. The system is condition may also be incurred. The model focuses on all
scalable as it can be used to test on different crops. From the style of farms, and smaller farmers may also be benefitted.

© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 749
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 02 | Feb 2020 www.irjet.net p-ISSN: 2395-0072

This model may be more increased to seek out the yield of this situation, the crop yield rate plays a significant role in the
each crop, and for chemical recommendation. Conjointly it economic growth of the country. So, there is a need to
may be changed to recommend concerning the fertilizers and increase crop yield rate. Some biological approaches (e.g.
irrigation want of crops. seed quality of the crop, crop hybridization, strong
pesticides) and some chemical approaches (e.g. use of
4. Evaluation of Predictive Data Mining Algorithms in Soil fertilizer, urea, potash) are carried out to solve this issue. In
Data Classification for Optimized Crop Recommendation. addition to these approaches, a crop sequencing technique is
required to improve the net yield rate of the crop over the
In this study, we've given the analysis potentialities for the season. One of existing system we identified is Crop Selection
classification of soil by mistreatment well-known Method (CSM) to achieve a net yield rate of crops over the
classification algorithms as J48, BF Tree, and OneR and Naïve season. We have taken example of CSM to demonstrate how it
Bayes; in data processing. The experiment was conducted on helps farmers in achieving more yield
information instances from Kasur district, Pakistan. We have
ascertained the comparative analysis of those algorithms Crop can be classified as:
have the various level of accuracy to determine the
effectiveness and potency of predictions. However, the a) Seasonal crops— crops can be planted during a season. e.g.
advantages of the higher understanding of soils classes will wheat, cotton.
improve the productivity in farming, reduce dependence on
fertilizers and build higher prognostic rules for the advice of b) Whole year crops— crops can be planted during the entire
the rise in yield. In the future, we have a tendency to contrive year. e.g. vegetable, paddy, Toor.
to form a Soil Management and
c) Short time plantation crops— crops that take a short time
5. Agricultural Production Output Prediction Using for growing. e.g. potato, vegetables, ratio. d) Long-time
Supervised Machine Learning Techniques plantation crops— These crops take a long time for growing.
e.g. sugarcane, Onion. A combination of these crops can be
Two supervised classification machine learning formula has selected in a sequence based on yield rate per day. Illustrates
been enforced during this study. the choice Tree Learning- sequences of crops with cumulative yield rate over the
ID3 (Iterative Dichotomiser 3) and KNNR discover the season. CSM method, shown in may improve the net yield
patterns within the knowledge set containing average rate of crops using the limited land resource and also
temperature and precipitation worth obtained throughout increases re-usability of the land.
the cropping amount of six major crops in 10 major cities of
Bangladesh for the past twelve years and provides the Basically, in crop selection method makes use of technique
prediction. ID3 uses the choice tree table that consists of the where it recommends different set of crops for same area
ranges of the precipitation, temperature and yield knowledge. over the years. There are various options are available to
The research provides an answer to the current downside select for farmers. They can choose one of the options and
that was much required for farmers in People's Republic of observe the results. The combination which will give high
Bangladesh. Though the research is restricted to some yield for same area is generated as output for that area. In
mounted dataset, the long run ahead promises addition of a this way CSM method tries to predict the suitable crops for
lot of knowledge which will be analyzed with more machine given area. Farming Systems in India are strategically utilized,
learning techniques to come up with crop predictions with according to the locations where they are most suitable. The
higher exactness. Moreover, the analysis will result in profits agricultural systems that significantly follows to the
and invention of advanced farming techniques which will agriculture of India are subsistence farming, organic farming,
improve our economy and can facilitate United States stand industrial farming. Regions all over India differ in types of
out as a technologically advanced country. farming they use; some are based on horticulture, ley
farming, agroforestry, and many more. The surveyed
4. EXISTING SYSTEM research papers have given a rough idea about using ML with
only one attribute. We have the aim of adding more attributes
An agro-based country depends on agriculture for its to our system and ameliorate the results, which can improve
economic growth. When a population of the country the yields and we can recognize several patterns for
increases dependency on agriculture also increases and predictions. This system will be useful to justify which crop
subsequent economic growth of the country is affected. In can be grown in a particular region

© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 750
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 02 | Feb 2020 www.irjet.net p-ISSN: 2395-0072

5. PROPOSED SYSTEM

In our system we are making use of a classification algorithms to improvised the crop yields.

5.1. Data Acquisition: are assessed and the factors strongly leading to accurate
prediction of the crops
Dataset must have following attributes
5.2 Preprocessing:
 Soil Parameters:
Soil Type The dataset that is used needs to be pre-processed because
Soil Ph value of the presence of redundant attributes, noisy data in it.
 Climatic Parameters: Initially, data cleaning operation is performed where the
Humidity redundant factors are determined and are not considered for
Temperature the prediction of crops. Over18 which are either having the
Wind same values for all the employees or are completely
Rainfall unrelated to the prediction task. As part of the exploratory
 Production data analysis, the categorical factors are split and are
Cost of cultivation assigned values as 0 and 1 based on whether the factor is
Previous year yield details for that region present or not. These assigned values assist in further
classification based on that particular factor.
In this project we are performing crops prediction for
district level. So main aim is to find the dataset which 5.3 Classifier Models:
contains production details of past 10-12 years also details
about climatic parameters and soil parameters like rainfall, 5.3.1 Decision Tree Classifier:
temperature, moisture, soil contents etc. details. These
The decision tree is method of selecting best root nodes until
factors will help in the prediction of the crops by using
we get elements of same class we keep on splitting the tree
various classifiers on the given dataset. Thus, various factors
on the basis of attributes. With versatile features helping

© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 751
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 02 | Feb 2020 www.irjet.net p-ISSN: 2395-0072

actualize both categorical and continuous dependent 6. RESULTS AND ANALYSIS

variables, it is a type of supervised learning algorithm mostly
used for classification problems. What this algorithm does is, We tested decision tree, naïve bayes classifier, and KNN
it splits the population into two or more homogeneous sets classifier with sample dataset containing attributes like crop
based on the most significant attributes making the groups name, cost of cultivation, cost of irrigation, cost of production
as distinct as possible. The decision tree algorithm will give which are independent variables and yield per hectare is
us best split on different features for selection of most dependent variable.
suitable crop among the population. The feature selection
methodology of Decision tree classifier makes it suitable for The result obtained are represented using confusion matrix
prediction of suitable crops. The Selection attributes of which shows relation between prediction of algorithms and
Decision tree classifier are as follow. actual values obtained when sample are tested after training
dataset.
5.3.1.1 Gini Index
1. Decision Tree Classifier:
Gini index says, if we select two items from a population at
random then they must be of same class and probability for Confusion matrix for decision tree classifier:
this is 1 if population is pure. Used to calculate impurity for
[ [221, 49],
the features of given classes.
[67, 163]]
5.3.1.2 Entropy
Accuracy for Decision-Tree: 76.8%
A decision tree is built top-down from a root node and
involves partitioning the data into subsets that contain Precision for Decision-Tree:0.767
instances with similar values (homogeneous). If the sample
is completely homogeneous the entropy is zero and if the Specificity for Decision-Tree:0.708
sample is equally divided then it has entropy of one.
2. KNN Classifier:
5.3.1.3 Information Gain
Confusion matrix for KNN:
The information gain is based on the decrease in entropy
after a dataset is split on an attribute. Constructing a [[269, 30],
decision tree is all about finding attribute that returns the
highest information gain (i.e., the most homogeneous [23, 178]]
branches. This attribute selection methods will play vital role
in prediction of crop. Accuracy for KNN: 89.4%

5.3.1.4 C4.5 Algorithm Precision for KNN:0.921

The C4.5 algorithmic program uses info gain as ripping Specificity for KNN:0.8825
criteria. It will handle numerical and categorical information
similarly as missing values. To handle continuous values, it Accuracy Comparison:
generates threshold and so divides attributes with prices
quite the edge price and values up to the edge value. It offers
the subsequent edges. They’re explicable, in contrast to
different classifiers, that need to be seen as a recorder that
has a class to a given input instance. Call trees will be
envisioned as tree graphs wherever nodes and branches
represent the classification rules learnt, and leaves denote
the ultimate categorizations.

5.3.3 KNN

KNN may be a variety of instance-based learning, wherever

the performance is barely approximated regionally and
every one computation is delayed till it's the classification.
Both for classification and regression, a helpful technique
will be to assign weights to the contributions of the
neighbors, in order that the nearer neighbors contribute
additional to the typical than the additional distant ones.

© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 752
 International Research Journal of Engineering and Technology (IRJET) e-ISSN: 2395-0056
Volume: 07 Issue: 02 | Feb 2020 www.irjet.net p-ISSN: 2395-0072

Precision Comparison [3]. Neha Rale, Raxitkumar Solanki, Doina Bein, James
Andro-Vasko, Wolfgang Bein.”Prediction of Crop
Cultivation”.978-1-7281-0554-3/19/$31.00©2019 IEEE

[4]. Md. Tahmid Shakoor, Karishma Rahman, Sumaiya Nasrin

Rayta, Amitabha Chakrabarty.2017.”Agricultural Production
Output Prediction Using Supervised Machine Learning
Techniques”.978-1-5386-3831-6/17/$31.00 ©2017 IEEE

[5]. G Srivatsa Sharma, Shah Nawaz Mandal, Shruti Kulkarni,

Monica R Mundada, Meeradevi.2018.”Predictive Analysis to
Improve Crop Yield Using a Neural Network Model”.978-1-
5386-5314-2/18/$31.00 ©2018 IEEE

[6]. Rashmi Priya, Dharavath Ramesh.2018.”Crop Prediction

on the Region Belts of India: A Naïve Bayes MapReduce
Precision Agricultural Model”. 978-1-5386-5314-
Specificity Comparison: 2/18/$31.00 ©2018 IEEE

[7]Talha Siddique,Dipro Barus,Zanntual Fredous,Amitabh

Chakravarti. 2017. “Automated Farming Prediction”.0978-1-
5090-6182-2/17/$31 @2017 IEEE

[8]Takeshi Yoshida Noriyuki Murakami and Hiroyuki

Tauiji.2017. Hybrid Machine Learning Approach to
Automatic Plant PhenotypingFor Smart Agriculture”. 978-1-
5090-5888-4/16/$31.00 @IEEE 2016

BIOGRAPHIES

Prof. Shrikant Kokate

Qualification: ME Comp
Area of Interest: Data Science, Web
Technologies
7. CONCLUSION
Name: Pavan Patil
The project work introduces an efficient crop
Qualification:BE
recommendation system using classifier models. The system
Computer(Pursuing)
is scalable as it can be used to test on different crops. From
the yield graphs the best time of sowing, plant growth and
harvesting of plant can also be found out along with 2nd
prediction for crops. Decision tree shows poor performance Name: Virendra Panpatil
Author
when dataset is having more variations but naïve bayes Qualification:BE
provides better result than decision tree for such datasets. Photo
Computer(Pursuing)
The combination classification algorithm like naïve bayes
and decision tree classifier are better performing than use of
single classifier model. 3rd
Author
REFERENCES
Photo
[1]. S.Veenadhari, Dr Bharat Misra, Dr CD
Singh.2019.”Machine learning approach for forecasting crop
yield based on climatic parameters.”.978-1-4799-2352-
6/14/$31.00 ©2014 IEEE

[2]. Igor Oliveira, Renato L. F. Cunha, Bruno Silva, Marco A. S.

Netto.2018.”A Scalable Machine Learning System for Pre-
Season Agriculture Yield Forecast.”.978-1-5386-9156-
4/18/$31.00 ©2018 IEEE DOI
10.1109/eScience.2018.00131

© 2020, IRJET | Impact Factor value: 7.34 | ISO 9001:2008 Certified Journal | Page 753