A Project on

CROP PRODUCTION PREDICTION

Submitted for partial fulfilment of award of

MASTER OF SCIENCE
DR. HOMI BHABHA STATE UNIVERSITY, MUMBAI

Degree in
DATA SCIENCE

By
Simran Kishan Kanojia

Under the Guidance of

Ms. Prachee Angane

Department of Mathematics
The Institute of Science
Mumbai – 400032

May, 2023
 Declaration

I hereby declare that the project work entitled “Crop Production Prediction”
carried out at the Department of Mathematics, The Institute of Science, Mumbai,
is a record of an original work done by me under the guidance of Dr. Selby Jose,
The Institute of Science, and this project work is submitted in the partial
fulfilment of the requirements for the award of the degree of Master of Science
in Data Science, Dr. Homi Bhabha State University. The results embodied in this
report have not been submitted to any other University or Institute for the award
of any degree or diploma.

SIMRAN KANOJIA
DS2216
 Acknowledgment

One thing that deserves my utmost gratitude is the opportunity to work

on this research project.

Above all, I would like to extend my heartfelt gratitude to my parents for

their unwavering support and encouragement throughout. Their unique insights
and guidance were priceless in our success of this project, and I feel incredibly
lucky to have had such a wonderful support system.

I am immensely grateful to Dr. Selby Jose, Head of the Department of

Mathematics, and the entire Department of Mathematics for providing us with
this unique opportunity to work on this project.

I would like to take this opportunity to express my heartfelt gratitude to

our project guides, Ms. Prachee Angane and Ms. Juilie Shinde for their invaluable
contributions to this project. Their expertise in the field has been invaluable in
helping me to navigate the challenges and complexities of this project, and their
guidance and feedback have helped me to refine my ideas and approach to the
work. They have been a source of inspiration and encouragement, and I am truly
grateful for their unwavering commitment to my success. I would like to extend
my deepest appreciation to everyone who came across and motivated and
supported. I am truly grateful for the opportunity to work with the talented and
dedicated group members. Their hard work, dedication, and commitment to
excellence have been a constant source of inspiration and motivation for me.

This project as a whole has been a journey of growth and learning for me,
and I am grateful for the experience. Finally, I would like to thank my mentor for
their guidance and support, which helped me navigate the challenges and
opportunities that arose during the project.

SIMRAN KISHAN KANOJIA

Contents

1. Introduction

2. Preliminaries

2.1 Basic Definitions and Results

2.2 Machine Learning Process
2.3Ensemble Learning
2.3.1 Bagging
2.4 Fitted Models
2.4.1 Random Forest Regression
2.4.2 Decision Tree Regression
2.4.3 Ridge Regression
2.4.4 Lasso Regression
2.4.5 Elastic Net Regression

3. Analysis of crop production in India

4. Prediction of Crop Production

5. Conclusion

6. Scope of the project

7. Bibliography
Chapter 1

Introduction
Agriculture, since its invention and inception, be the prime and pre-
eminent activity of every culture and civilization throughout the history of
mankind. It is not only an enormous aspect of the growing economy, but it’s
essential for us to survive. It’s also a crucial sector for Indian economy and also
human future. It also contributes an outsized portion of employment. Because the
time passes the requirement for production has been increased exponentially. So
as to produce in mass quantity people are using technology in an exceedingly
wrong way. New sorts of hybrid varieties are produced day by day. However,
these varieties don’t provide the essential contents as naturally produced crop.
These unnatural techniques spoil the soil. It all ends up in further environmental
harm. Most of these unnatural techniques are wont to avoid losses.
But when the producers of the crops know the accurate information on the
crop yield it minimizes the loss. Machine learning, a fast-growing approach that’s
spreading out and helping every sector in making viable decisions to create the
foremost of its applications. Most devices nowadays are facilitated by models
being analysed before deployment. The main concept is to increase the
throughput of the agriculture sector with the Machine Learning models. Another
factor that also affects the prediction is the amount of knowledge that’s being
given within the training period, as the number of parameters was higher
comparatively. The core emphasis would be on precision agriculture, where
quality is ensured over undesirable environmental factors.
So as to perform accurate prediction and stand on the inconsistent trends
in the season a crop is chosen for cultivation and the area under cultivation various
Machine Learning models like Random Forest Regression Model, Linear
Regression, Ridge, Lasso, Elastic Net, Decision Tree Regression are applied to
urge a pattern. By applying these machine learning models, I came into a
conclusion that Random Forest algorithm provides the foremost accurate value.
System predicts crop prediction from the gathering of past data. Using the past
information on Year, Area, State and UT the major two crops Rice and Wheat
was cultivated the model was trained to give meaningful prediction of production
in tonnes.
Chapter 2

Preliminaries

2.1 Basic Definitions and Results

Definition 2.1.1: Machine Learning (ML)

Machine learning is a branch of artificial intelligence (AI) and computer
science which focuses on the use of data and algorithms to imitate the way
that humans learn, gradually improving its accuracy.

Definition 2.1.2: Algorithm

An algorithm is a computer program that adapt and evolve based on the data
it is processed to produce predetermined outcomes.

Definition 2.1.3: Model

A model is a program that can find patterns or make decisions from a
previously unseen dataset.

Definition 2.1.4: Predictor Variable (x)

A predictor variable is a variable or set of features whose values will be used
to predict the value of the target variable.

Definition 2.1.5: Response Variable (Y)

It is the feature or the output variable that needs to be predicted by using the
predictor variable. It is that feature of the dataset about which you want to
gain a deeper understanding.
Definition 2.1.6: Training Data
Training data is the data you use to train an algorithm or machine learning
model to predict the outcome you design your model to predict.

Definition 2.1.7: Testing Data

An unseen data used to test your model is called testing data, and you can
use it to evaluate the performance and progress of your algorithms training
and adjust or optimize it for improved results.

2.2 Machine Learning Process

The machine learning process involves constructing a predictive model to
address a specific problem. The process typically consists of seven key
steps:

1) The Objective of the Problem

Clearly defining and understanding the problem statement that needs to
be solved.

2) Data Gathering
Identifying and selecting the appropriate sources from which to obtain
the necessary data. This step aims to collect the data required for training
the model.

3) Preparing Data
Manipulating and organizing the data in a suitable format for model
training. This step involves data cleaning tasks such as removing
duplicates, correcting errors, handling missing values, normalizing data,
and converting data types.

4) Data Exploration
Visualizing the data to identify relevant relationships between variables,
detect class imbalances, or perform other exploratory analyses.
 Additionally, splitting the data into training and evaluation sets is done
in this step.

5) Building a Model
Developing a predictive model that performs better than a baseline or
reference model. This step involves selecting an appropriate algorithm or
model architecture and training it using the prepared data.

6) Model Evaluation and Optimization

Assessing the performance of the trained model and optimizing its
effectiveness. This includes scaling up the model to detect potential
overfitting issues, regularizing the model to improve generalization, and
tuning model parameters for better performance.

7) Predictions
Applying the trained model to new, unseen data (test set) that was
previously withheld from the model. This data, for which the class labels
are known, is used to evaluate the model's performance and obtain a
better estimate of how the model will perform in real-world scenarios.

2.3 Ensemble Learning

Ensemble learning is a machine learning technique that combines multiple
models to improve their performance. It reduces variance and bias of the
model and improves its accuracy and robustness. It is an approach in which
two or more models are fitted to the same dataset, and the predictions of
each model are combined.
There are mainly three types of Ensemble learning methods -
1) Bagging -
It stands for Bootstrap Aggregating and involves training multiple
models on different subsets of the data and combining their predictions.
The goal is to reduce the variance of the model and prevent overfitting.

2) Boosting –
It involves training multiple models sequentially and giving more weight
to the misclassified data. The goal is to improve the accuracy of the
model and reduce the bias. Involves ML model like XGBoost.
3) Stacking -
It involves training multiple models and using their predictions as input
to a meta-model. The goal is to improve the accuracy and reduce the
variance of the model by combining the strengths of multiple models.
Involves methods to build new prediction models such as KNN (K
Nearest Neighbour) or SVM (Support Vector Machine).

2.3.1 Bagging
Bagging is a machine learning technique that involves training multiple
models on different subsets of the data and combining their predictions. The
goal of bagging is to reduce the variance of the model and prevent
overfitting. This method is commonly used to reduce variance and maintain
bias.
The name "bagging" comes from the fact that the technique involves
creating random subsets of the training data, also known as "bags". Each
bag contains a random sample of the training data, and a model is trained
on each bag.
The models can be of different types, but decision trees are often used
because they tend to overfit the training data. By training multiple trees on
different subsets of the data, bagging can help to reduce the variance of the
model and improve its accuracy. Once the models are trained, their
predictions are combined to make a final prediction. This can be done by
taking the average of the predictions for regression tasks, or by using a
voting scheme for classification tasks.
Machine learning algorithms that use the bagging method are as follows –

1) Random Forest -
It is an ensemble learning method that combines multiple decision trees
to make predictions. It uses bagging to reduce the variance of the model
and improve its accuracy.

2) Bagged Decision Trees -

It involves training multiple decision trees on different subsets of the data
and combining their predictions. It is a type of bagging method that can
be used for both regression and classification tasks.

3) Bagged SVM –
It involves training multiple support vector machines (SVMs) on
different subsets of the data and combining their predictions. It is a type
of bagging method that can be used for classification tasks.

4) Bagged Neural Networks –

It involves training multiple neural networks on different subsets of the
data and combining their predictions. It is a type of bagging method that
can be used for both regression and classification tasks.
All of these algorithms use the bagging method to reduce the variance of the
model and improve its accuracy.

2.4 Fitted Models -

2.4.1 Random Forest Regression
Random Forest Regression is a supervised learning algorithm that uses
ensemble learning method for regression. Ensemble learning method is
a technique that combines predictions from multiple machine learning
algorithms to make a more accurate prediction than a single model.
The diagram above shows the structure of a Random Forest. A Random
Forest operates by constructing several decision trees during training time
and outputting the mean of the classes as the prediction of all the trees. To
get a better understanding of the Random Forest algorithm, let’s walk
through the steps:
1) Pick at random k data points from the training set.
2) Build a decision tree associated to these k data points.
3) Choose the number N of trees you want to build and repeat steps 1 and
2.
4) For a new data point, make each one of your N-tree trees predict the value
of y for the data point in question and assign the new data point to the
average across all the predicted y values.
A Random Forest Regression model is powerful and accurate. It usually
performs great on many problems, including features with nonlinear
relationships. Disadvantages, however, include the following: there is no
interpretability, overfitting may easily occur, we must choose the number of
trees to include in the model.

Advantages -
1) Random forest regression is a powerful and flexible modelling technique
that can handle complex data structures and high dimensional data.
2) It is less prone to overfitting than other modelling techniques, such as
decision trees, due to the use of bagging and random feature selection.
3) It can handle missing data and categorical variables without the need for
data pre-processing.
4) It provides feature importance measures that can be used to identify the
most important predictors in the model.

Disadvantages -
1) Random forest regression can be computationally expensive and require
a lot of memory, especially for large datasets.
2) It is not as interpretable as simpler models, such as linear regression.
3) It may not perform well on small datasets or datasets with a small number
of predictors.
4) It may not work well if there are strong correlations between predictors.
2.4.2 Decision Tree Regression
A Decision Tree is a predictive model that uses a set of binary rules in order
to calculate the dependent variable. Each tree consists of branches, nodes,
and leaves.
Terminologies used in this regression are as follows:
1) The root node represents the entire population and is divided into two
or more homogeneous sets.
2) A decision node is when a sub-node splits into further sub-nodes.
3) A leaf is when a node does not split. These are also referred to as
"Terminal Nodes".

A Decision Tree imposes a series of questions to the data, each question

narrowing possible values, until the model is trained well to make
predictions. These questions are determined completely by the model,
including their content and order, and are asked in a True/False form.
Decision trees are classified supervised learning model. It learns based on a
known set of input data with known responses to the data. It is able to make
a prediction by running through the entire tree, asking true/false questions,
until it reaches a leaf node. The final prediction is given by the average of
the value of the dependent variable in that leaf node.
Advantages -
1) Decision Tree Regression is a non-parametric algorithm that can
model complex relationships between input and output variables without
making assumptions about the underlying distribution of the data.
 2) It is easy to interpret and visualize, making it a useful tool for
understanding the patterns and relationships in the data.
3) It can handle both categorical and continuous input variables and can
perform variable selection automatically.
4) It can handle missing values and outliers in the data.
5) It can be used for both classification and regression problems.

Limitations -
1) Decision Tree Regression is prone to overfitting, especially when the
tree is deep and the training data is small.
2) It may not perform well when the input variables have complex
interdependencies or when there are many irrelevant variables in the data.
3) It may not generalize well to new data if the training data is not
representative of the entire population.

2.4.3 Ridge Regression

Ridge regression is a technique that estimates the coefficients of multiple
regression models when the independent variables are highly correlated or
multicollinear. It is a form of regularized regression that adds regularization
to the least square problem. It can be used to predict continuous values and
to reduced overfitting.
For any type of regression machine learning model, the usual regression
equation forms the base which is written as: Y = X𝛽 + e
Where,
Y is the dependent variable,
X represents the independent variables,
𝛽 is the regression coefficients to be estimated, and
e represents the errors are residuals.
Standardization -
In ridge regression, the first step is to standardize the variables (both
dependent and independent) by subtracting their means and dividing by
their standard deviations. This causes a challenge in notation since we must
somehow indicate whether the variables in a particular formula are
standardized or not. As far as standardization is concerned, all ridge
regression calculations are based on standardized variables. When the final
regression coefficients are displayed, they are adjusted back into their
original scale. However, the ridge trace is on a standardized scale.

Assumptions -
The assumptions of ridge regression are the same as that of linear
regression: linearity, constant variance, and independence.
To avoid overfitting the learning should constrain the solution in order to fit
a global pattern. Adding such a penalty will force the coefficients to be
small, i.e., to shrink them toward zeros.
Ridge regression impose a 𝑙 penalty on the coefficients, i.e., it penalizes with
the Euclidean norm of the coefficients while minimizing SSE. The objective
function becomes:
𝑅𝑖𝑑𝑔𝑒(𝜃) = 𝑦 − 𝑥𝜃 + 𝜆 𝜃

Advantages -
1) Ridge regression can help to prevent overfitting in linear regression
models by adding a regularization term to the loss function.
2) It can handle multicollinearity in the data by shrinking the coefficients
of correlated predictors towards each other.
3) It is computationally efficient and easy to implement.
4) It can improve the stability and generalizability of the model.
Disadvantages -
1) Ridge regression assumes that all predictors are equally important,
which may not be the case in some situations.
2) It may not perform as well as other regularization techniques, such as
Lasso regression, when there are a large number of predictors.
3) It may not work well if the data is not well-suited to linear modelling.
4) It may require some tuning of the regularization parameter to get
optimal results.

2.4.4 Lasso Regression

The word “LASSO” stands for Least Absolute Shrinkage and Selection
Operator. It is a statistical formula for the regularisation of data models and
feature selection.
Lasso regression is a regularization technique. It is used over regression
methods for a more accurate prediction. This model uses shrinkage.
Shrinkage is where data values are shrunk towards a central point as the
mean. The lasso procedure encourages simple, sparse models (i.e. models
with fewer parameters). This particular type of regression is well-suited for
models showing high levels of multicollinearity or when you want to
automate certain parts of model selection, like variable selection/parameter
elimination.
Lasso regression penalizes the coefficients by the l1 norm. This constraint
will the capacity of the learning algorithm.

To add such a penalty forces the coefficients to be small, i.e., it shrinks them
toward zero. The objective function to minimize becomes:
𝑙𝑎𝑠𝑠𝑜(𝜃) = 𝑦 − 𝑥𝜃 + 𝜆 𝜃
If a regression model uses the L1 Regularization technique, then it is called
Lasso Regression.

Mathematical equation of Lasso Regression:

Residual Sum of Squares + λ × (Sum of the absolute value of the magnitude
of coefficients) where, λ denotes the amount of shrinkage.
 • λ = 0 implies all features are considered and it is equivalent to the
linear regression where only the residual sum of squares is considered
to build a predictive model
• λ = ∞ implies no feature is considered i.e., as λ closes to infinity it
eliminates more and more features
• The bias increases with increase in λ and Variance increases with
decrease in λ

Advantages -
1) Lasso regression can help to prevent overfitting in linear regression
models by adding a regularization term to the loss function.
2) It can perform feature selection by setting the coefficients of irrelevant
predictors to zero.
3) It can handle multicollinearity in the data by shrinking the coefficients
of correlated predictors towards each other. 4) It is computationally
efficient and easy to implement.
Disadvantages -
1) Lasso regression may not perform as well as other regularization
techniques, such as Ridge regression, when there are a large number of
predictors.
2) It may not work well if the data is not well-suited to linear modelling.
3) It may require some tuning of the regularization parameter to get
optimal results.
4) It may not be able to handle situations where there are more predictors
than observations.
2.4.5 Elastic Net Regression
Elastic net (also called ELNET) regression is a statistical hybrid method that
combines two of the most often used regularized linear regression
techniques, lasso, and ridge, to deal with multicollinearity issues when they
arise between predictor variables. Regularization aids in solving the
overfitting issues with the models.
It is also used for regularizing and choosing the essential predictor variables
that significantly impact the response variable. Ridge employs an L2
penalty, while lasso employs an L1. Since the elastic net utilizes both the L2
and the L1 models, the question of choosing between either one does not
arise.
An elastic net is a combination of two regressions, lasso, and ridge, and
hence the resultant equation to calculate it is:

The elastic net penalty comes in two varieties: l1 and l2. The lasso and ridge
regression models are two types of regularization models that apply l1 and
l2 penalties, respectively. The absolute value of the coefficient’s magnitude
is added as a penalty term in the lasso regression model. The ridge
regression adds the coefficient’s squared magnitude as a penalty to the loss
function.
Thus, it deals with both multicollinearity problems and the selection of
regression coefficients. ELNET uses coefficient regression shrinkage
towards zero or equal to zero to reduce the occurrence of predictor variables.
The tuning parameter (λ1) multiplied by the sum of coefficient variables (l1
norms) absolute values is utilized for this purpose. The tuning parameter
(λ2) is multiplied by the l2 norm’s squared coefficient variables to handle
the high correlation between the predictor variables. The ELNET regression
approach helps produce a fitting, interpretable model by minimizing
unnecessary variables that do not appear in the final model to improve
prediction accuracy. The ELNET manages multicollinearity by maintaining
or excluding highly correlated predictor variables from the fitted model.
While building ElasticNet regression model, both hyperparameters (L2) and
(L1) need to be set.
ElasticNet takes the following two parameters:
• alpha – Constant that multiplies the penalty terms. Default value is set
to 1.0.
(alpha = 𝛾 + 𝜎)
• l1_ratio – The ElasticNet mixing parameter, with 0 ≤ l1_ratio ≤ 1.

Where, l1_ratio = 0 implies that the penalty is an L2

penalty.
l1_ratio = 1 implies that it is an L1 penalty.
0 < l1_ratio < 1 implies that the penalty is a combination of L1 and L2.
Chapter 3

Analysis of crop production in India

India is a global agricultural powerhouse and is unquestionably the largest
livelihood provider in India, more so vast in rural areas. Being that important
sector, it requires good quality of analytical view and better farming practices to
avoid damage and loss on the part of our hardworking farmers. For our
examination approach crop production details state-wise which then goes a level
deep with district details was extracted from a government owned website,
https://www.aps.dac.gov.in/. The data from 1997-2020 with major crops Rice and
Wheat is taken under study. Some other features that are available to us are the
crop season and area under cultivation.
This data is only about India. And so, the scope of research is also going to be limited to just India.

3.1 State and District-wise Production analysis (1997 – 2020)

3.1.1 Graph representing State and Union Territories Vs Production in

tonnes
3.1.2 Wheat production –
According to the procurement report of 2022 by Times Of India (TOI),

i. Punjab: As per the current trends, Punjab is the largest producer

of wheat in India, accounting for around 40% of the total wheat
production in the country. On analysis, the state shows up as the
largest producer with major wheat-growing districts in the state
including Sangrur, Firozepur, Ludhiana, Patiala and Bhatinda.
The climate of Punjab is well suited and the soil in is also rich in
nutrients, which provides the necessary conditions for wheat. In
addition, the state has a well-developed irrigation system.
 ii. Uttar Pradesh: Uttar Pradesh is the second-largest producer of
wheat in India currently, accounting for around 18% of the total
wheat production in the country. The state is known for its high-
yielding wheat varieties and favourable agro-climatic conditions.
The past 22 years analysis shows that the state stands up as the
second largest for wheat production in India. The major wheat-
growing districts in Uttar Pradesh include Hardoi, Shahjahanpur,
Budaun, Aligarh and Saharanpur.

iii. Haryana: The third-largest producer of wheat in India, accounting

for around 15% of the total wheat production in the country. Land
of Haryana is flat, covered with loamy soil which is very suitable
for agriculture. According to our analysis over the data, major
wheat-growing districts in Haryana include Sirsa, Hisar, Jind,
Fatehabad and Ambala.
iv. Madhya Pradesh: The state is said to be the fourth-largest
producer of wheat in India right now, accounting for around 10%
of the total wheat production in the country. The state has a
favourable climate and fertile soil for wheat cultivation. So, the
analysis over the data concludes the same and the major wheat-
growing districts in the state are Hoshangabad, Ujjain, Vidisha,
Sehore and Raisen.

v. Rajasthan: The state is the fifth-largest producer of wheat in India

currently, accounting for around 8% of the total wheat production
in the country. The Major wheat-growing districts in Rajasthan
include Ganganagar, Hanumangarh, Alwar, Bharatpur and Kota.
The state is better is terms of wheat production because its desert
area has a potential.
3.1.3 Rice production –
According to the procurement analysis report by Food Corporation of
India (FCI),

i. West Bengal: The state is the largest producer of rice in India,

accounting for around 13.95% of the country’s rice. The state has
a favorable climate and fertile soil for rice cultivation. Major rice-
growing districts in West Bengal include Burdwan, Hooghly, and
Nadia. Rice cultivation requires an ample of water and a type of
soil which holds the water, West Bengal have such type of soil.
 ii. Uttar Pradesh: The second-largest producer of rice in India,
accounting for around 13% of the total rice production in the
country. The state has a favorable climate and fertile soil for rice
cultivation. About 25% of the cultivated area of the state is
devoted to rice production. Also, Allahabad, Barabanki,
Gorakhpur etc., are the major rice-producing districts of the state.
In Uttar Pradesh, you will find Jaya, Panth-4, Mahsuri, Pusa
Basmati Rice and Kasturi Basmati Rice. Farmers also cultivate
the rice in an area of 59 hectares in UP.

iii. Punjab: The state is the third-largest producer of rice in India,

accounting for around 12% of the total rice production in the
country. The state has a favorable climate and fertile soil for rice
cultivation. Major rice-growing districts in Punjab include
Sangrur, Ludhiana, Patiala, Firozpur and Moga. This state is a
predominantly wheat-growing state, farmers have also used crop
rotation and perennial irrigation sources to grow rice.
iv. Andhra Pradesh: It is the fourth-largest producer of rice in India,
accounting for around 11% of the total rice production in the
country. The state has a favorable climate and fertile soil for rice
cultivation. Major rice-growing districts in Andhra Pradesh
include East Godavari, West Godavari and Krishna. Also, the
state is renowned as the ‘Rice Bowl of India’ as Krishna-
Godavari delta region is historically as know by the same name.

v. Telangana: Telangana is the fifth-largest producer of rice in India,

accounting for around 9% of the total rice production in the
country. The state has a favorable climate and fertile soil for rice
cultivation. Major rice-growing districts in Telangana include
Suryapet, Jagitial, and Peddapalli.
3.2 Agricultural area under cultivation Vs Production in tonnes

Agricultural area and production are closely related in India, where

agriculture is a major source of livelihood for millions of people.
The amount of agricultural area available for cultivation is a key factor
that determines the overall production levels of crops such as rice and
wheat. In general, states with larger areas under cultivation for these
crops tend to have higher production levels, while states with smaller
areas under cultivation may have to rely on higher yields per hectare to
achieve similar production levels.
According to the Ministry of Agriculture and Farmers' Welfare, the
state with the highest area under cultivation in India is Uttar Pradesh,
with around 25 million hectares of area used for agriculture. As per the
analysis on the past data, the state is the leading one using the land for
agriculture purpose.
3.3 Crop Vs Production in tonnes

Over the past few decades, India has seen significant growth in both
rice and wheat production. According to the United States Department
of Agriculture, the total rice production in India has increased from
around 80 million tonnes in 1997 to around 120 million tonnes in 2020-
21. Similarly, the total wheat production in India has increased from
around 60 million tonnes in 1997 to around 109 million tonnes in 2020-
21. So, both rice and wheat production have increased significantly
over the past few decades.
Also, the Rice is generally grown in greater quantities than wheat in
India. According to the Ministry of Agriculture and Farmers' Welfare,
in 2020-21, the total area under rice cultivation in India was around 43
million hectares, while the total area under wheat cultivation was
around 30 million hectares. So, rice is grown more than wheat in India.

CODE -

GRAPHICAL REPRESENTATION,
3.3.1 Rice

3.3.1.1 State, District-wise production of rice

As per the data from the Ministry of Agriculture and Farmers' Welfare,
the top rice-producing states in India from 1997 to 2020 were West
Bengal, Uttar Pradesh, and Andhra Pradesh. West Bengal consistently
maintained its position as the largest producer of rice in India during
this period, with a production of around 14-15 million metric tons per
year. Uttar Pradesh and Andhra Pradesh were also among the top
producers during this period, with production levels ranging from 10-
14 million metric tons per year.

Top 5 districts with highest yield of crop in West Bengal,

CODE -

OUTPUT -
GRAPHICAL REPRESENTATION,

3.3.1.2 Year-wise production of rice

CODE -

GRAPHICAL REPRESENTATION,
3.3.2 Wheat
3.3.2.1 State, District-wise production of wheat

The top wheat-producing states in India from 1997 to 2020 were Uttar
Pradesh, Punjab, and Haryana. Punjab is said to be the leading state.

Top 5 districts with highest yield of crop in Punjab,

CODE -

OUTPUT -

GRAPHICAL REPRESENTATION,
3.3.2.2 Year-wise production of wheat

CODE -

GRAPHICAL REPRESENTATION,

3.4 Year-wise Production in tonnes

After a fluctuating growth the production dropped down to around

2019. According to the reports, COVID-19 disrupted certain activities
in agriculture and the supply chain as well. Few reports indicates that
harvesting activities were suspended due to lack of migrant workers,
particularly in Northwest India where wheat is cultivated.
3.5 Season Vs Production

Rice and wheat are two of the most important food crops grown in
India. Rice is mainly grown during the kharif season, while wheat is
mainly grown during the rabi season. The production of these crops
varies depending on the season, with different factors affecting the
production during each season. Despite rest of the challenges like
climate and pests, India has managed to maintain a steady increase in
the production of rice and wheat, which has helped to meet the growing
demand for food in the country.

3.5.1 Season based production of rice

Being a kharif crop, rice is grown in another season as well. This might
be possible due to the favourable climatic and soil conditions in certain
regions.
3.5.2 Season based production of wheat

Wheat is a rabi crop but in some regions the production is certainly

possible for the entire year as well.
Chapter 4

Prediction of Crop Production

I trained the data using Random Forest Regression model and predicted the crop
production. I choose Random Forest regression algorithm as this was the best fit
for the dataset under study, better predictions were made by this model as
compared to the rest models. In this case n_estimators refer to the number of trees
in the forest and I took it as 100 to avoid overfitting as Random Forest algorithm
does not allow overfitting if there are more trees. Since the data under study was
large enough and with many features, this algorithm has the ability to work upon
it and deal with higher dimensionality as well.

This model is said to have 97% accurate in forecasting crop productions.

CODE -
OUTPUT -

GRAPHICAL REPRESENTATION,
Chapter 5
Conclusion
Crop yield prediction is still remaining as a challenging issue for farmers. The
aim of this research is to propose and implement a rule-based system to predict
the crop yield production from the collection of past data. This has been achieved
by applying association rule mining on agriculture data from 1997 to 2020.
Farmers require modern technologies to help them raise their crops. Agriculturists
can be notified about accurate crop predictions on a timely basis. The agriculture
factors were analysed using a variety of machine learning approaches. The
features are chosen are determined by the dataset availability and research goal.
According to studies, models with more characteristics may not always deliver
the highest yield prediction performance. Models with more and fewer features
should be evaluated to discover the best performing model. The findings reveal
that while no definitive conclusion can be taken about which model is the best,
they do show that some machine learning models are utilized more frequently
than others. With the help of Random Forest Regression model, I got an accuracy
of about 97%.
Scope of the project
I have taken into consideration only two major factors i.e., the area and season of
the districts. There are other factors such as the fertility and type of soil present
in the area which would affect the crop yield. Also, one of the major factor
climatic conditions.
On proper analysis, I can help find the soil type and fertilizers and add
on the weather conditions which can be helpful in maximizing and predicting the
crop yield. P.C.A and many Deep Learning techniques can be used on images of
the field and crop to detect if they are infected by any disease or the presence of
weed, which also affects the quality of the crop, and can be separated from the
healthy crops at the earliest possible.
Bibliography
[1] https://www.academia.edu/

[2] https://apeda.gov.in/apedawebsite/SubHead_Products/Wheat.html

[3] https://www.aps.dac.gov.in/

[4] https://agricoop.nic.in/sites/pocketbook

[5] https://edis.ifas.ufl.edu/publications

[6] https://techvidvan.com

[7] www.ijraset.com/research-paper/agriculture-crop-yield-prediction