CROP RECOMMENDATION SYSTEM

USING MACHINE LEARNING

A Synopsis submitted to

JSS Science and Technology University

Master of Science
in

Data Science
by
Anuprasad H A

Under the supervision of

Department of Information Science and Engineering

2023-24
 ABSTRACT

Agriculture is the important aspect in terms of existence and livelihood of humans, as it has
become a necessary and crucial factor in our daily life; evolution of farming techniques is
developed that leads to the improvement in supply of food to the country’s GDP and also
made easy for farmers to grow the accurate and suitable crop without any risk and loss in the
productivity. To maintain the sustainability in rate and quality of production, we need to
involve and introduce the best portable and noncomplex strategies such as machine learning
tools to carry on the needed operations and procedures to get desired and expected crop by
coordinating and initiating data exchange between the scientific and practical platforms,
which is the trending existing system. The concept and ideology of this project named crop
recommendation system defines and states that the respective crop has to be grown on basis
of various relative parameters like soil characteristics or features such as nitrogen,
phosphorus, potassium that are extracted from the soil through filtration process, and weather
conditions which are embedded in a dataset in form of structured data. This dataset is taken
over by machine learning algorithms that will perform some operations like classification and
will be finding the accuracy where in detail to be explained we will be splitting the given
dataset into training and testing data and compare the results of those algorithms based on
accuracy that each model gives and that will be our preferred algorithm. We used machine
learning techniques like decision tree, random forest, KNN along with few dimensionality
reduction techniques like PCA, LDA and cross-validation.

1
CHAPTER 1 : INTRODUCTION

Agriculture is the foundation of all economies. In a nation like India, where the population is
expanding and there is a constant need for food, agriculture must increase to satisfy the needs.
India's fundamental and preeminent cultural practise has long been thought of as agriculture.
Ancient people's needs were satisfied because they farmed their own land. As a result, many
organisms, including people, animals, and birds, cultivate and consume natural crops. The
greenery-based goods produced on the land that the animal consumed promote a long and
healthy existence.

OVERVIEW

Machine learning (ML) is used to find a solution when the link between the variables used as
input and output is ambiguous or hard to determine. The automated collection of structural
descriptions from examples of the described item is referred to in this sense as "learning."
Machine learning (ML) does not rely on the validity of the underlying structure of the data
model, unlike traditional statistical methods. This trait is particularly useful for simulating
complex non-linear behaviours, such as a function for agricultural yield prediction. The best
application of ML for predicting agricultural production.

Precision agriculture allows for the precise application of inputs such as seed, water, pesticides,
and fertilizers to crops at the optimal time to maximize productivity, quality, and yields.

Farmers can better understand their fields by deploying sensors for data collection and mapping
fields, allowing them to conserve resources and reduce negative environmental effects.

C
H
 Agriculture is the important factor of economy in India. In recent years due to industrialization;
excessive use of pesticides the strength of soil is getting affected. Many of the methods followedby
agriculture are not sufficient to increase the productivity. The common difficulty present and thus
among the Indian farmers is they don’t have any information regarding the right crop based on
their soil requirements so it affects the productivity.
Indian farmers face a lot of challenges in making decisions about which farming technique to opt
for and which crop should be selected for which climate. The common problem existing among
Indian farmers that they don’t choose proper crops so as to obtain max yield which arebased on
topographic features and financial aspects.

PROBLEM STATEMENT

 Instability:
Agriculture in India is largely depending on monsoon. As a result, production of foodgrains
fluctuates year after year. A year of abundant output of cereals is often followed by a year of
acute shortage.

 Cropping Pattern:

The crops that are grown in India are divided into two broad categories: food crops andnon-food
crops. While the former comprises food-grains, sugarcane and other beverages, the latter includes
different kinds of fibers and oilseeds.

 Land Ownership:

Although the ownership of agricultural land in India is fairly widely distributed, there is some
degree of concentration of land holding. Inequality in land distribution is also due to the fact
that there are frequent changes in land ownership in India.

 Land Tenure:

The land tenure system of India is also far from perfect. In the pre-independence period, most
tenants suffered from insecurity of tenancy. They could be evicted any time.
5

 When it comes to crop yield prediction, there are several factors to consider. Among
these arestudies of climate data, satellite imagery, soil conditions, and the possibility of
pest attacks.

 All of this adds up to provide a comprehensive picture of the best times for crop
production. There are also what-if scenarios and alternative action plans in place to deal
with any unforeseenissues.

 Businesses in the AgTech industry are now using neural network algorithms to forecast
crop yield.

 Backpropagation algorithms aid in determining the appropriate yield weight value for
calculating the error derivative. Crop yield estimation accuracy is important for
agronomic production.

 As a result, crop yield prediction is critical for the global food production ecosystem.
With better data, it is possible to make more informed decisions.

 Crop yield prediction data is also useful to government agencies so that they can plan for
national food security

M
E
6

In Indian economy and employment agriculture plays major role. The most common problem
faced by the Indian farmers is they do not opt crop based on the necessity of soil, as a result they
face serious setback in productivity. This problem can be addressed throughprecision agriculture.
This method takes three parameters into consideration, viz: soil characteristics, soil types and
crop yield data collection based on these parameters suggesting the farmer suitable crop to be
cultivated. Precision agriculture helps in reduction of non-suitable crop which indeed increases
productivity, apart from the following advantages like efficacy in input as well as output and
better decision making for farming.

This method gives solutions like proposing a recommendation system through an ensemble
model with majority voting techniques using Random Forest and K - Nearest Neighbour as
learner to recommend suitable crop based on soil parameters with high specific accuracy and
efficiency.

C
H
7

Machine learning is a branch of computer science that employs statistical techniques to enable
computer systems to "learn" (i.e., progressively improve performance on a specific task) from
data without being explicitly programmed. Arthur Samuel coined the term "machine learning" in
1959 programming explicit algorithms with high performance is difficult or impossible; examples
include email filtering, network intruder detection, and computer vision.

Machine learning, which evolved from the study of pattern recognition and computational
learning theory in artificial intelligence, investigates the study and construction of algorithms that
can learn from and make predictions on data – such algorithms overcome strictly static
programme instructions by making data-driven predictions or decisions, by building a model from
sample inputs. Machine learning is used in a variety of computing tasks where designing and
Machine learning is closely related to (and frequently overlaps with) computational statistics,
which is also concerned with making predictions using computers. It has strong ties to
mathematical optimization, which provides the field with methods, theory, and application
domains. Machine learning is frequently confused with data mining, thelatter of which focuses on
exploratory data analysis and is referred to as unsupervised learning.

Machine learning is a method used in the field of data analytics to create complex models and
algorithms that lend themselves to prediction; in commercial use, this is known as predictive
analytics.

T
h
8

Agricultural crop recommendation systems are available in the market which consider various
parameters like weather at the time the crop is to be planted, soil type, topography of the
region, temperature and rainfall in the region, market prices of the crop, crop duration, etc.
Research has been carried out in this field and the following papers have been referred for the
purpose
of research and study.

Prof. Rakesh Shirsath and other co-author in paper [1] proposed a system which helps the users
to make decisions regarding the crop to be planted. The system used is a subscription-based
system which would have personalized information of every farmer registered. The system
includes a module which maintains the information of the previous crops planted collected
from various sources and shows a matching crop that can be planted. The whole process is
done with the help of artificial neutral networks. At the end a feedback system is provided so
that the developer can make changes required if the farmer finds some difficulty while using
the system.

Big Data Analysis Technology Application in Agricultural Intelligence Decision System paper
authors Ji-chun Zhao and Jianxin Guo in paper [2] considers knowledge database as big data
and inferences from the data is drawn. It considers various modules like users, knowledge
engineer, domain expert, man-machine interface, inference engine and knowledge base. The
knowledge acquisition system obtains knowledge for the decision system and establishes an
effective knowledge base to solve the problem. The paper uses various Hadoop modules for
the purpose of feature extraction. It uses the unstructured data and processes it using NoSQL,
Hive, Mahout and uses HDFS to store the data. The data was just presented for wheat crop and
other crops were not considered.

RSF as mentioned in paper [3] is a recommendation system for famers which considers a
location detection module, data analysis and storage module, crop growing database,
physiographic database. The similar location detection module identifies the locations which
are similar to the user’s locations and checks the similar crops that are planted in those
locations.
R
E
 9
Accordingly, using similarity matrix, the recommendations for the user is generated. Location
detection module uses the Google API services to get the current location of the user to
identify the similar locations. But the system does not get user feedback to improve the
process.

ensemble technique called Majority Voting Technique which combines the power of multiple
models to achieve greater prediction accuracy. The methods used are Random Trees, KNN,
CHAID and Naïve Bayes for ensemble so that even if one method predicts incorrectly, the other
models are likely to make correct predictions and since the majority voting technique is used,
the final prediction is correct one. If-then rules are the main components which are used in the
prediction process. The accuracy obtained is 88% using the ensemble model.

Paper [5] is a review paper for studying various algorithms and their accuracy in the
agricultural field proposed by Yogesh Gandge and Sandhya. It was observed that Multiple
Linear Regression gave an accuracy of 90-95% for rice yield. Decision tree using ID3
algorithm was considered for soybean crop and the recommendations were generated. The
third algorithm was SVM which was used on all the crops and the accuracy was good with
computationally less requirements. Neural network was used on corn data to achieve
95% of accuracy. Other algorithms were also used which are KNN, C4.5, K-means, J48,
LAD Tree and Naïve Bayes. The conclusion was that still improvement is needed for the
algorithms to achieve better accuracy.

InUse of Data Mining in Crop Yield Prediction [6], paper [6], the dataset used was collected
from Kaggle.com The author has analysed the data using WEKA tool for algorithms which are
LWL, J48, LAD Tree and IBK. The accuracy was measured using specificity, sensitivity,
accuracy, RMSE and mean absolute error. For each classifier, confusion matrix was used to get
the correctly identified instances. The observation was that better accuracy can be obtained if
pruning is used.

Paper [7] presented by Rakesh Kumar, M.P. Singh, Prabhat Kumar and J.P. Singh proposed
use of seven machine learning techniques i.e. ANN, SVM, KNN, Decision Tree, Random
Forest, GBDT and Regularized Gradient Forest for crop selection. The system is designed to
retrieve all the crops sowed and time of growing at a particular time of the year. Yield rate of

R
E
10
each crop is obtained and the crops giving higher yields are selected. The system also proposes
a sequence of crops to be planted to get the higher yields.

R
E