SMART FARMING USING MACHINE

LEARNING
Shubham Pawar1, Sumit Dere1, Ashitosh Akangire 1 , Harshvardhan Kamble1,
Prof. Seema Shrawne2
1
B.Tech Student, Dept of Computer Engineering & IT , VJTI College, Mumbai, Maharashtra, India.
2
Assistant Professor, Dept of Computer Engineering & IT, VJTI College, Mumbai, Maharashtra, India.

Abstract — Agriculture plays an important role in food to a net importer of food. All these trends
Indian economy. But now-a-days, agriculture in indicate that the agricultural sector in India is facing
India is undergoing a structural change leading to a a crisis today.
crisis situation. The only remedy to the crisis is to It is argued that the consequence of the
do all that is possible to make agriculture a agricultural crisis in India is very vast and likely to
profitable enterprise and attract the farmers to hit all the other sectors and the national economy in
continue the crop production activities. several ways. The only remedy to the crisis is to do
As an effort towards this direction, this research all that is possible to make agriculture a profitable
paper would help the farmers in making appropriate enterprise and attract the farmers to continue the
decisions regarding the cultivations with the help of crop production activities.
machine learning. This paper focuses on predicting In the past farmers used to predict their yield from
the appropriate crop based on the climatic situations previous year yield experiences. Thus, for this kind
and the yield of the crop based on the historic data of data analytics in crop prediction, there are
by using supervised machine learning algorithms. In different techniques or algorithms, and with the help
addition, a web application has been developed. of those algorithms, we can predict crop yield.
Nowadays, modern people don't have awareness
about the cultivation of the crops at the right time
I. INTRODUCTION
and at the right place. By analyzing all the issues
and problems like weather, temperature, and several
Agriculture is the backbone of the Indian economy.
factors, there is no proper solution and technologies
But agriculture in India is undergoing a structural
to overcome the situation faced. Accurate
change leading to a crisis situation. The relative
information about history of crop yield is an
contribution of agriculture to the GDP has been
important thing for making decisions related to
declining
agricultural risk management.Therefore, this paper
over time steadily. It is alarming that India is
proposes an idea to predict the crop and yield of the
moving from being a self-reliant nation of
crop based on the climatic conditions and historic
data related to the crop .The farmer will check the
production of the crop as per the acre ,before named Crop Advisor has been developed as a
cultivating onto the field. user-friendly web page for predicting the influence
of climatic parameters on crop yields.C4.5
The main objectives are: algorithm is used to produce the most influencing
a. To use machine learning techniques to predict climatic parameter on the crop yields of selected
crop and yield of the crop. crops in selected districts of Madhya Pradesh.
b. To analyze the data properly and to process the https://ieeexplore.ieee.org/document/6921718
data to get better predictions.
c. To improve the performance of machine learning In [3]. Analysis of Crop Yield Prediction by making
models. Use of Data Mining Methods. IJRET: The paper
d. To build an easy to use web application. provided in the International Journal of Research in
Engineering and Technology. In this paper, the main
aim is to create a user-friendly interface for farmers,
II. PROBLEM STATEMENT
which gives the analysis of rice production based on
the available data. For maximizing the crop
Crop selection and crop yield prediction are
productivity various Data mining techniques were
important agricultural problems. The aim of this
used to predict the crop yield.
project is to predict suitable crop based on the given
https://pdfs.semanticscholar.org/3376/e91c3a77a54
climate parameters and location and also to predict
7ce51cfe4a2e68ea6f35ffe63.pdf
the yield of that crop based on the season and area
of the field using machine learning algorithms.
In [4]. Random Forests for Global and Regional
Crop Yield Predictions. institute on the
III. LITERATURE REVIEW Environment, University of Minnesota, St. Paul,
MN 55108, United States of America. The
In [1] Predicting the yield of the crop using a generated outputs show that RF is an effective and
machine learning algorithm. International Journal of different machine-learning method for crop yield
Engineering Science Research Technology. This predictions at regional and global scales for its high
paper focuses on predicting the yield of the crop accuracy.
based on the existing data by using the Random https://www.researchgate.net/publication/30379496
Forest algorithm. Real data of Tamil Nadu were 5_Random_Forests_for_Global_and_Regional_Cro
used for building the model. p_Yield_Predictions
http://www.ijesrt.com/issues%20pdf%20file/Archiv
e-2018/April-2018/1.pdf In[5] Crop Prediction using Machine Learning
This research work helps the beginner farmer in
In [2]. Machine learning approach for forecasting such a way to guide them for sowing the reasonable
crop yield based on parameters of climate. The crops by deploying machine learning. Naive Bayes,
paper was provided at the International Conference a supervised learning algorithm puts forth in the
on Computer Communication and Informatics way to achieve it. The proposed supervised machine
(ICCCI). In the current research, a software tool learning using naive Bayes Gaussian classifier with
boosting algorithm is developed to predict the crop includes Temperature, Humidity, Soil pH, Rainfall
at high accuracy. and class label is the Crop.

In[6] Smart Farming Prediction Using Machine Following figures are the snapshots of the datasets
learning that have been used for this project.
The paper is about using machine learning with
various environmental factors like soil, pressure,
weather, crop type to predict the maximized
profitable crop to grow. The paper mainly focuses
on the algorithms used to predict crop yield, crop
cost predictions.
https://www.ijitee.org/papers/Smart_Farming_Predi
ction_Using_Machine_Learning

In [7]Crop Prediction on the Region Belts of India:

A Naïve Bayes MapReduce Precision Agricultural
Model The planned work introduces an efficient
degree economical crop recommendation system.
From the yield graphs, the simplest time of sowing, Fig. 1. Dataset For Crop Prediction Problem
plant growth, and gathering of plants may be
known. Conjointly the best and worst condition may
also be incurred. The model focuses on all styles of
farms, and smaller farmers may also be benefitted.
This model may be increased to seek out the yield
of each crop, and for a chemical recommendation.
https://ieeexplore.ieee.org/document/8554948

IV. ANALYSIS AND DESIGN

To implement the system, we decided to focus on

Maharashtra State only in India. Historical data
about the crop and the climate at the district level Fig. 2. Dataset For Crop Yield Prediction Problem
was needed to implement the system. This data has
been gathered from the government website
Exploratory Data Analysis: It refers to the critical
www.data.gov.in which includes State, District,
process of performing initial investigations on data
Season, Crop, Area and Production. The data about
so as to discover patterns,to spot anomalies,to test
the climate conditions suitable for the particular
hypotheses and to check assumptions with the help
crops has been collected from the Kaggle which
of summary statistics and graphical representations.
Data Cleaning: It is the process of preparing data
for analysis by removing or modifying data that is
incorrect, incomplete, irrelevant, duplicated, or
improperly formatted.

Encoding: It is a required pre-processing step when

working with categorical data for machine learning
algorithms.

Feature Scaling: It is a technique to standardize the

independent features present in the data in a fixed
range. It is performed during the data
pre-processing to handle highly varying magnitudes
or values or units.

Data Partitioning: The Entire dataset is partitioned

into 2 parts: for example, say, 75% of the dataset is
used for training the model and 25% of the data is
set aside to test the model.

The Following Fig. 3 shows the proposed approach

to this problem. There are basically two modules: Fig. 3. Proposed Approach
the first one is the Crop prediction module which
predicts the most appropriate crop based on
temperature, humidity, soil pH and rainfall values V. IMPLEMENTATION
and second one is the Crop yield prediction module
which predicts the production of the crop based on For the first module which is basically a multiclass
area, season and location. The climate related data classification problem, we built the following
like temperature, humidity etc. can be acquired models and evaluated their performance.
through OpenWeatherMap API and location, soil 1. KNN
pH and area can be input through the user. 2. Support vector Machine
3. Random Forest
4. Naive Bayes

The metric we used for the evaluation is Cohen’s

Kappa Score.. It is a very good measure that can
handle very well both multi-class and imbalanced
class problems.
It basically tells you how much better your classifier
is performing over the performance of a classifier
that simply guesses at random according to the
frequency of each class.

Classification Cohen’s Kappa

Models Score

KNN 0.8825

SVM 0.8108

Random Forest 0.9356

Naive Bayes 0.9513

Fig. 4. Five Numbers Summary
Table 1: Comparison between various classifiers
Hence we divided the dataset into two sets based on
Naive Bayes Classification Model gives the highest the conditions that Area <= 24000 and Area >
Cohen’s Kappa score among all the classifiers. 24000. And then built the models for both the
Hence the Naive Bayes classifier has been selected datasets. The metrics we used for the evaluation are
in the project. R-Squared Value and the Mean Squared Error Value
(MSE).
For the second module which is basically a
regression problem, we built the following models [1] Performance of the models for the dataset with
and evaluated their performance. observations where Area <= 24000
1. Multi-linear Regression
2. Random Forest Regression
Regression R-Squared MSE
3. Support Vector Regression Model Value
4. KNN Regression
Random Forest 0.8285 0.22
Regressor
The dataset for this module contains outliers more
than 30% of the dataset. The following figure shows Support Vector 0.0581 1.138
the significant difference between min, median, Regressor
mean and max values due to which standardizing
Multiple Linear 0.1590 1.0168
the data was a big problem. Regressor

KNN Regressor 0.7039 0.3580

Table 2: Comparison between various regressors

[2] Performance of the models for the dataset with
observations where Area > 24000

Regression R-Squared MSE

Model Value

Random Forest 0.9179 0.1011

Regressor

Support Vector 0.0119 1.2166

Regressor

Multiple Linear 0.2351 0.9418

Regressor

KNN Regressor 0.8340 0.2043

Table 3: Comparison between various regressors

Random Forest Regression Model gives the highest

R-Squared value and least MSE among all the
regressors. Hence the Random Forest Regressor has Fig. 5. User Inputs The Data
been selected in the project.

To develop an user-friendly web application we

used Flask. Flask is an API of Python and it is
based on the WSGI toolkit and Jinja2 template
engine.

User selects the location, puts the soil ph value, gets

from ph meter and then puts the area which is in
acres.
The result shows the appropriate crop based on the
climatic conditions as well as the production in
tonnes. Web page also displays the data that user
inputs and the weather data. Fig. 6. Predictions
VI. CONCLUSION AND FUTURE SCOPE

Crop and yield of the crop prediction using

intelligent machine learning techniques may
improve the crop planning decisions. For the Crop
Prediction Module, the Cohen’s Kappa score we got
for the Naive Bayes Classification Model is about
95%. For the Crop Yield Prediction Module, the
R-Squared value we got for the Random Forest
Regression Model is more than 81%.
Accurate forecasts of the climate parameters and
better historic data of the crop would result in
accurate crop and its yield forecast in the future.
Also, the developed webpage is user friendly and
can be made more informative by providing
additional useful information like intercropping,
fertilizers etc. to the user. We can create more
interactive User Interface by adding chatbots and
speech recognition systems.
 REFERENCES [11] Ramesh Medar, Vijay S. Rajpurohit, Shweta Shweta KLS
GIT, Belagavi. Crop Yield Prediction using Machine Learning
Techniques
[1] P.Priya, U.Muthaiah. Balamurugan.Predicting the yield of
the crop using a machine learning algorithm. International
[12] ‘District Wise rainfall data’ Available:
Journal of Engineering Science Research Technology.
http://hydro.imd.gov.in/hydrometweb/(S(ctep2djbwexxp0qr1k
nt3a45))/DistrictRaifall.aspx
[2]. S.Veenadhari, Dr. Bharat Misra, Dr. CD Singh.Machine
learning approach for forecasting crop yield based on climatic
[13] Suvidha Jambekar, Shikha Nema, Zia Saquib. dept.of
parameters. International Conference on Computer
Electronics and Communication Usha Mittal Institute of
Communication and Informatics (ICCCI)
Technology Mumbai, Maharashtra, India. Prediction of Crop
Production in India Using Data Mining Techniques
[3]D Ramesh, B Vishnu Vardhan. Analysis Of Crop Yield
Prediction Using Data Mining Techniques. IJRET:
[14]Predict Crop Production in India Using Machine Learning
International Journal of Research in Engineering and
Technique: A Survey
Technology
Bhavna Sharma, Pratap Singh Yadav, Sunita Yadav
[4]. Jig Han Jeong, Jonathan P. Resop, Nathaniel D. Mueller,
[15]Crop Selection Method to Maximize Crop Yield Rate
David H. Fleisher, Kyungdahm Yun, Ethan E. Butler,
using Machine Learning Technique. Rakesh Kumar, M.P.
Soo-Hyung Kim. Random Forests for Global and Regional
Singh, Prabhat Kumar, J.P. Singh
Crop Yield Predictions. Institute on the Environment,
University of Minnesota, St. Paul, MN 55108, United States of
[16]Machine learning approach for forecasting crop yield
America.
based on climatic parameters S. Veenadhari, Dr. Bharat
Mishra, Dr. C. D. Singh
[5]M. Kalimatthu, M. Kishore P.Vaishnavi
Crop Prediction using Machine Learning naive Bayes
[17]Crop yield prediction: two-tiered machine learning model
Gaussian classifier with boosting algorithm.
approach Ayush Kapoor, Mrityunjay V. Latte, Sushila Sidnal
[6].. Rashmi Priya, Dharavath Ramesh.2018.” Crop Prediction
[18]Crop Yield Analysis Using Machine Learning Algorithms.
on the Region Belts of India: A Naïve Bayes MapReduce
Fatin Farhan Haque, Ahmed Abdelgawad, Venkata Prasanth
Precision Agricultural Model”.
Yanambaka, Kumar Yelamarthi College of Engineering and
Technology Central Michigan University, ET100 Mount
[7]. Igor Oliveira, Renato L. F. Cunha, Bruno Silva, Marco A.
Pleasant, MI 48858, USA
S. Netto.2018.” A Scalable Machine Learning System for
PreSeason Agriculture Yield Forecast.”

[8]Renato L. F. Cunha, Bruno Silva, Marco A. S. Nett. A

Scalable Machine Learning System for Pre-Season
Agriculture Yield Forecast. 2018 IEEE 14th International
Conference on e-Science (e-Science)

[9] "data.gov.in", [online] Available: https://data.gov.in/.

[10]Aruvansh Nigam, Saksham Garg, Archit Agrawal, Parul

Agrawal Jaypee Institute of Information Technology, India.
Crop Yield Prediction Using Machine Learning Algorithms