PREDICTION OF FERTILIZERS FOR THE EFFICIENT

YIELD THROUGH MACHINE LEARNING

A Project report submitted in partial fulfillment of the requirements for the award of
the degree of

BACHELOR OF TECHNOLOGY
IN
ELECTRONICS AND COMMUNICATION ENGINEERING

Submitted by
P. PRUDHVIRAJU (317126512047) N. SUMITRANJALI(317126512041)
S. S MANOJ (317126512050) G. DAVID ELIEZER(317126512020)
Under the guidance
Of
Mr. N. RAM KUMAR
Assistant Professor

DEPARTMENT OF ELECTRONICS AND COMMUNICATION ENGINEERING

ANIL NEERUKONDA INSTITUTE OF TECHNOLOGY AND SCIENCES
(UGC AUTONOMOUS)
(Permanently Affiliated to AU, Approved by AICTE and Accredited by NBA & NAAC with ‘A’ Grade)
Sangivalasa, bheemili mandal, visakhapatnam dist.(A.P)2020-2021

i
 ACKNOWLEDGEMENT

We would like to express our deep gratitude to our project guide N.Ram Kumar, Assistant
professor ,Department of Electronics and Communication Engineering, ANITS, for his/her
guidance with unsurpassed knowledge and immense encouragement. We are grateful to
Dr. V. Rajyalakshmi, Head of the Department, Electronics and Communication
Engineering, for providing us with the required facilities for the completion of the project
work.
We are very much thankful to the Principal and Management, ANITS, Sangivalasa, for
their encouragement and cooperation to carry out this work.
We express our thanks to all teaching faculty of Department of ECE, whose suggestions
during reviews helped us in accomplishment of our project. We would like to thank all
non-teaching staff of the Department of ECE, ANITS for providing great assistance in
accomplishment of our project.
We would like to thank our parents, friends, and classmates for their encouragement
throughout our project period. At last but not the least, we thank everyone for supporting
us directly or indirectly in completing this project successfully.

PROJECT STUDENTS
Prudhviraju Polamarasetty (317126512047),
Sumitranjali Nagothi (317126512041),
Satyanarayana Manoj Samudrala (317126512050),
David Eliezer Gorremuchu(317126512020)

iii
 ABSTRACT
Richness of soil assumes a vital part in gaining great yield from the harvest. Present day
headways in innovation are end up being a help in catalyzing the harvest yield. A large
portion of the ranchers have faith in legends in the public eye and develop without earlier
information or appropriate examination of manures that are most appropriate for the given
soil and harvest type. Soil testing, determining the best reasonable compost will expand
the agrarian creation by improving the supplement content accessible in the dirt. Utilization
of wrong manures definitely impacts the yield and soundness of the harvest. ML is an up-
and-coming field of informatics that can be applied effectively to the horticultural area, so
we proposed an ML model which break down the given informational collection with the
distinctive ML models like Decision Tree, Random Forest, Gradient Boost, Ada Boost,
Gaussian NB and anticipate the most appropriate fertilizer by choosing the best suitable
model. ML methods helps in compost prediction consequently, assists the ranchers with
improving the harvest yield.

iv
 CONTENTS
ABSTRACT iv

LIST OF SYMBOLS viii

LIST OF FIGURES ix

LIST OF TABLES xi

LIST OF ABBREVATIONS xii

CHAPTER 1 INTRODUCTION 01
1.1 Project Outline 01
1.2 Project Objective 02
CHAPTER 2 METHODOLOGY 03

2.1 Introduction to Machine Learning 04

2.1.1 Supervised Learning 05
2.1.2 Unsupervised Learning 05
2.1.3 Reinforcement Learning 05

2.2 Technologies Used 06

2.2.1 Python 06
2.2.2 Libraries 07
2.2.3 Pandas 07
2.2.4 Matplotlib 09
2.2.5 NumPy 10
2.2.6 Scikit Learn 11

v
 2.2.7 Seaborn 12
2.2.8 Jupyter Notebook IDE 14
2.2.9 Jupyter Kernals 15
2.3 Algorithms Used 16
2.3.1 Decision Tree (DT) 16
2.3.2 Random Forest (RF) 18
2.3.3 Gaussian Naïve Bayes (NB) 19
2.3.3.1 Bayes’ Therom 20
2.3.3.2 Naïve Bayes classifier 20
2.3.3.3 Gaussian Naïve Bayes 21
2.3.4 Boosting 22
2.3.4.1 Ada Boost (AB) 23
2.3.4.2 Gradient Boost (GB) 24

CHAPTER 3 DATA VISUALIZATION 25

3.1 Cleaning of Data set 25
3.2 What is Data Visualization? 28
3.2.1 Why visualization? 28
CHAPTER 4 RESULTS 38
4.1 Evaluation Parameters 38
4.1.1 Confusion Matrix 38
4.1.2 Accuracy 39
4.1.3 Precision 39
4.1.4 Recall 39
4.1.5 F1-Score 39
4.2 Gradient Boost 43
4.3 Decision Tree 44
4.4 Random Forest 45
4.5 Gaussian NB 46

vi
 4.6 Ada Boost 47
4.7 Comparison Table 48
CHAPTER 5 CONCLUSION AND FUTURE WORK 49

REFERENCES 50
APPENDIX 52

vii
 LIST OF SYMBOLS

Symbol Description Page no

H(s) Entropy 16
P(x) Probability of the attribute 17
IG (S, A) Information Gain 17
TN True Negative 38
TP True Positive 38
FP False Positive 38
FN False Negative 38

viii
 LIST OF FIGURES

Figure no Title Page no

Fig. 2.1 Steps involved in prediction of fertilizers. 03
Fig. 2.2 Machine Learning Pipeline. 04
Fig. 2.3 A Trained Model of Supervised Learning. 06
Fig. 2.4 Seaborn Plots. 13
Fig. 2.5 Jupyter IDE 14
Fig. 2.6 Decision tree 17
Fig. 2.7 Random Forest Classifier 19
Fig. 2.8 Formulas for Bayes’Theorem 20
Fig. 2.9 Gaussian NB Classifier 22
Fig.2.10 Steps in Ada Boost Classifier 23
Fig. 2.11 Steps in Gradient Boost Classifier 24
Fig. 3.1 Steps involved in Data Cleaning 25
Fig. 3.2 Observations of Null Values and Unique Values of Parameters 26
Fig. 3.3 Types of Crops in that are considered 27
Fig. 3.4 Different Fertilizers Used 27
Fig. 3.5 Types of Data Visualization Techniques 28
Fig. 3.6 Histogram of Humidity 30
Fig. 3.7 Histogram of Moisture 31
Fig. 3.8 Histogram of Nitrogen 32
Fig. 3.9 Histogram of Phosphorous

ix
Fig. 3.10 Histogram of Potassium 33
Fig. 3.11 Correlation Matrix 35
Fig. 3.12 Temperature vs Fertilizer 36
Fig. 3.13 Moisture vs Fertilizer 37
Fig. 3.14 Nitrogen vs Fertilizer 37
Fig. 3.15 Potassium vs fertilizer 37
Fig. 4.1 Confusion Matrix of Gaussian NB 40
Fig. 4.2 Confusion Matrix of Decision Tree 40
Fig. 4.3 Confusion Matrix of Random Forest 41
Fig. 4.4 Confusion Matrix of Gradient Boost 41
Fig. 4.5 Confusion Matrix of Ada Boost 42
Fig. 4.6 Gradient Boost Classifier 43
Fig. 4.7 Decision Tree Classifier 44
Fig. 4.8 Random Forest Classifier 45
Fig. 4.9 Gaussian NB Classifier 46
Fig. 4.10 Ada Boost Classifier 47

x
 LIST OF TABLES

Table no Title Page no

Table 3.1 Observations made from Correlation Matrix 35

Table 4.1 Confusion Matrix Representation 38

Table 4.2 Results of Gradient Boost Classifier 43

Table 4.3 Results of Decision Tree Classifier 44

Table 4.4 Results of Random Forest Classifier 45

Table 4.5 Results of Gaussian NB Classifier 46

Table 4.6 Results of Ada Boost Classifier 47

Table 4.7 Comparison Table for all proposed algorithms 48

xi
 LIST OF ABBREVATIONS

AB Ada Boost
GB Gradient Boost
DT Decision Tree
RF Random Forest
ML Machine Learning
IoT Internet of Things
N Nitrogen
P Phosphorus
K Potassium

xii
 CHAPTER -1
INTRODUCTION

1.1 PROJECT OUTLINE

Agribusiness expects a critical part in monetary headway similarly as occupation for
country like India. India is the second greatest maker of horticulture products. One of the
major issues looked by the ranchers is picking the correct manure to the particular yield.
There are numerous expensive composts available in the market, yet they fail to give great
outcomes since ranchers do not think about their individual requirements which results in
awful yield. The choice of compost relies upon numerous components consequently it
cannot be universalized.
These days there is a great deal of advancement in modern innovation which assists with
picking a superior compost, one of the arising fields is Machine learning. Machine
Learning is a part of man-made reasoning (AI) zeroed in on building applications that gain
from information and improve their precision after some time without being modified to
do so. It has various calculations with the assistance of which we can make right predictions
a portion of the algorithms are Random Forest, Decision Tree, Ada Boost, Gradient Boost
and Gaussian NB. Machine Learning does complex calculations and an assortment of exact
information to work keenly. It utilizes past information to peruse the examples and
afterward play out the proposed task as indicated by the characterized rules and calculations
dependent on the examination it delivered.
The central point influencing crop yield are the crop type, soil type the supplements like
N, P, K present in the soil, temperature dampness and mugginess. Without considering
these in the event that we purchase a manure it wouldn’t give accurate results. This project
focuses on comparing five different ML models and concludes the accurate model to
predict the suitable fertilizer when provided with the macro nutrient (NPK) values obtained
from soil test along with the temperature, humidity, crop type, soil type by the user.

1
1.2 PROJECT OBJECTIVE
Fertilizers are chemical substances supplied to the crops to increase their productivity.
These are used by the farmers daily to increase the crop yield. The fertilizers contain the
essential nutrients required by the plants, including nitrogen, potassium, and phosphorus.
They enhance the water retention capacity of the soil and increase its fertility. Fertility of
soil plays a crucial role in getting good yield from the crop. A fertile soil will contain all
the major nutrients for basic plant nutrition like Nitrogen(N), Potassium(k),
Phosphorous(P). This project helps the farmer to predict the suitable fertilizer for the given
nutrient levels of the soil obtained from the soil test. Every combination of soil and plant
are unique, and they require different form of nutrients. So, the type of fertilizer required
for them also vary. Farmers may not know the exact requirement by the soil or plant until
they get the result. Hence, farmer in one region may end up with good yield due to the
right selection of fertilizer while farmer in a different region with same type of soil and
plant yield improper result. This ensures that the fertilizer being used in the farm is based
on any unclear predictions. The key is to get this balance right and to maintain a level of
nutrients in soils that will support our crops. So, there is a need to establish a platform to
suggest the right fertilizer for a given crop and soil type.

2
 CHAPTER -2
METHODOLOGY

An informational collection is gathered which contains the data of temperature, mugginess,

N, P, K values crop type, soil type and the most appropriate manure for the given
conditions. further this informational index is checked for any invalid qualities or any
copies whenever discovered they are taken out. Informational index is then changed over
into reference diagrams, relationship networks. Charts are plotted against every single
boundary so the framework can have a superior comprehension of the information. 75% of
the information is feed to the machine for training it and remaining information is utilized
to test it to see whether it’s making the correct prediction, distinctive ML algorithms like
Random Forest, Decision Tree, Ada Boost Gradient boost and Gaussian NB are utilized to
do this. when each model does the forecast we ascertain the precision with assistance of
confusion matrix, F-Score, precision and recall by contrasting these qualities we choose
the best appropriate model to do the prediction. Once the best suited Model is chosen, the
system is provided with the macro nutrient (N, P, K) values, temperature, humidity, crop
type, soil type by the user and gives the best preferred fertilizer for the given conditions.
Fig 2.1 shows the complete methodology of fertilizer prediction.

Collection of data set

Cleaning of data set

Visualization of data set

Applying different Machine Learning

techniques like Decision Tree, Random
Forest and Gaussian NB

Selecting the most accurate model

Optimizing the model for further

development.
Fig 2.1: Steps involved in prediction of fertilizers.

3
2.1 INTRODUCTION TO MACHINE LEARNING
Machine learning is a subfield of artificial intelligence (AI). The goal of machine learning
generally is to understand the structure of data and fit that data into models that can be
understood and utilized by people. Although machine learning is a field within computer
science, it differs from traditional computational approaches. In traditional computing,
algorithms are sets of explicitly programmed instructions used by computers to calculate
or problem solve. Machine learning algorithms instead allow for computers to train on data
inputs and use statistical analysis in order to output values that fall within a specific range.
Because of this, machine learning facilitates computers in building models from sample
data in order to automate decision-making processes based on data inputs.

Fig 2.2: Machine Learning Pipeline

In machine learning, tasks are generally classified into broad categories. These categories
are based on how learning is received or how feedback on the learning is given to the
system developed. Machine learning implementations are classified into three major
categories as follows: -

4
2.1.1 Supervised Learning

In supervised learning, the computer is provided with example inputs that are labelled with
their desired outputs. The purpose of this method is for the algorithm to be able to “learn”
by comparing its actual output with the “taught” outputs to find errors and modify the
model accordingly. Supervised learning therefore uses patterns to predict label values on
additional unlabelled data.

2.1.2 Unsupervised Learning

When an algorithm learns from plain examples without any associated response, leaving
to the algorithm to determine the data patterns on its own. This type of algorithm tends
to restructure the data into something else, such as new features that may represent a class
or a new series of un-correlated values. They are quite useful in providing humans with
insights into the meaning of data and new useful inputs to supervised machine learning
algorithms.

2.1.3 Reinforcement Learning

When you present the algorithm with examples that lack labels, as in unsupervised
learning. However, you can accompany an example with positive or negative feedback
according to the solution the algorithm proposes comes under the category of
Reinforcement learning, which is connected to applications for which the algorithm must
make decisions (so the product is prescriptive, not just descriptive, as in unsupervised
learning), and the decisions bear consequences. In the human world, it is just like learning
by trial and error. Errors help you learn because they have a penalty added (cost, loss of
time, regret, pain, and so on), teaching you that a certain course of action is less likely to
succeed than others. An interesting example of reinforcement learning occurs when
computers learn to play video games by themselves. In this project we have used different
supervised algorithms. Fig 2.3 shows the trained supervised model

5
 Fig 2.3: A trained model of Supervised Learning

2.2 TECHNOLGIES USED

2.2.1 Python:

Python is an interpreted high-level, general-purpose programming language. Python's

design philosophy emphasizes code readability with its notable use of significant
indentation. Its language constructs and object-oriented approach aim to help programmers
write clear, logical codes for small and large-scale projects.

Python is dynamically typed and garbage-collected. It supports multiple programming

paradigms, including structured (mainly procedural), object-oriented and functional
programming. Python is often described as a "batteries included" language due to its
comprehensive standard library.

Guido van Rossum began working on Python in the late 1980s as a successor to the ABC
programming language and first released it in 1991 as Python 0.9.0. Python 2.0 was
released in 2000 and introduced new features, such as list comprehensions and a garbage

6
collection system using reference counting. Python 3.0 was released in 2008 and was a
significant revision of the language that is not entirely backwards compatible, and much
Python 2 code does not run unmodified on Python 3. Python 2 was discontinued version
2.7.18 in 2020.

Python is a multi-paradigm programming language. Object-oriented programming and

structured programming are fully supported, and many of its features support functional
programming and aspect-oriented programming (including metaprogramming and
metaobjects (magic methods)). Many other paradigms are supported via extensions,
including design by contract and logic programming.

2.2.2 Libraries:

Python's sizeable standard library, commonly cited as one of its greatest strengths provides
tools suited to many tasks. For Internet-facing applications, many standard formats and
protocols such as MIME and HTTP are supported. It includes modules for creating
graphical user interfaces, connecting to relational databases, generating pseudorandom
numbers, arithmetic with arbitrary-precision decimals, manipulating regular expressions,
and unit testing.

Specifications cover some parts of the standard library (for example, the Web Server
Gateway Interface (WSGI) implementation follows PEP, but most modules are not. They
are specified by their code, internal documentation, and test suites. However, because
mostof the standard library is cross-platform Python code, only a few modules need
altering or rewriting for variant implementations.

2.2.3 Pandas

Pandas is a software library written for the Python programming language for data
manipulation and analysis. In particular, it offers data structures and operations for

7
manipulating numerical tables and time series. It is free software released under the three
clause BSD license. The name is derived from the term "panel data", an econometrics term
for data sets that include observations over multiple periods for the same individuals. Its
name is a play on the phrase "Python data analysis" itself. Wes McKinney started building
what would become pandas at AQR Capital while he was a researcher from 2007 to 2010.

Features:

• Data Frame object for data manipulation with integrated indexing.

• Tools for reading and writing data between in-memory data structures and different file
formats.

• Data alignment and integrated handling of missing data.

• Reshaping and pivoting of data sets.

• Label-based slicing, fancy indexing, and sub setting of large data sets.

• Data structure column insertion and deletion.

• Group by engine allowing split-apply-combine operations on data sets.

• Data set merging and joining.

• Hierarchical axis indexing to work with high-dimensional data in a lower-dimensional

data structure.

• Time series-functionality: Date range generation and frequency conversion, moving

window statistics, moving window linear regressions, date shifting and lagging.

• Provides data filtration.

8
2.2.4 Matplotlib

Matplotlib is a plotting library for the Python programming language and its numerical
mathematics extension NumPy. It provides an object-oriented API for embedding plots
into applications using general-purpose GUI toolkits like Tkinter, wxPython, Qt, or GTK.
A procedural "pylab" interface is also based on a state machine (like OpenGL), designed
to resemble MATLAB, though its use is discouraged closely. SciPy makes use of
Matplotlib.

John D. Hunter originally wrote Matplotlib. Since then, it has an active development
community and is distributed under a BSD-style license. Michael Droettboom was
nominated as matplotlib's lead developer shortly before John Hunter's death in August 2012
and was further joined by Thomas Caswell. Matplotlib 2.0.x supports Python versions 2.7
through 3.6. Python 3 support started with Matplotlib 1.2. Matplotlib 1.4 is the last version
to support Python 2.6. Matplotlib has pledged not to support Python 2 past 2020 by signing
the Python 3 Statement.

Tools required:

• Base map: map plotting with various map projections, coastlines, and political boundaries

• Cartopy: a mapping library featuring object-oriented map projection definitions and

arbitrary point, line, polygon and image transformation capabilities. (Matplotlib v1.2 and
above)

• Excel tools: utilities for exchanging data with Microsoft Excel • GTK tools: interface to
the GTK library

• Qt interface

• Mplot3d: 3-D plots

9
• Nat grid: interface to the nat grid library for gridding irregularly spaced data.

• matplotlib2tikz: export to Pgfplots for smooth integration into LaTeX documents

• Seaborn: provides an API on top of Matplotlib that offers sane choices for plot style and
colour defaults, defines simple high-level functions for common statistical plot types, and
integrates with the functionality provided by Pandas

2.2.5 NumPy

NumPy is a library for the Python programming language, adding support for large,
multidimensional arrays and matrices, along with an extensive collection of high-level
mathematical functions to operate on these arrays. NumPy, Numeric, was created by Jim
Hugunin with contributions from several other developers. In 2005, Travis Oliphant
created NumPy by incorporating features of the competing Num array into Numeric, with
extensive modifications. NumPy is open-source software and has many contributors.

Features:

•NumPy targets the C Python reference implementation of Python, which is a non

optimizing bytecode interpreter. Mathematical algorithms are written for this version of
Python often run much slower than compiled equivalents.

• NumPy addresses the slowness problem partly by providing multi-dimensional arrays and
functions and operators that operate efficiently on arrays, requiring rewriting some code,
primarily inner loops, using NumPy.

• Using NumPy in Python gives functionality comparable to MATLAB since they are both
interpreted, and they both allow the user to write fast programs as long as most operations
work on arrays or matrices instead of scalars. In comparison, MATLAB boasts many

10
additional toolboxes, notably Simulink, whereas NumPy is intrinsically integrated with
Python, a more modern and complete programming language.

• Moreover, complementary Python packages are available; SciPy is a library that adds
MATLAB-like functionality, and Matplotlib is a plotting package that provides MATLAB
like plotting functionality. Internally, both MATLAB and NumPy rely on BLAS and
LAPACK for efficient linear algebra computations.

• Python bindings of the widely used computer vision library OpenCV utilize NumPy
arrays to store and operate on data. Since images with multiple channels are simply
represented as three-dimensional arrays, indexing, slicing or masking with other arrays are
very efficient ways to access specific pixels of an image.

• The NumPy array as a universal data structure in OpenCV for images, extracted feature
points, filter kernels and many more vastly simplifies the programming workflow and
debugging.

2.2.6 Scikit-learn

Scikit-learn (formerly scikits.learn and sklearn) is a free software machine learning library
for the Python programming language.

It features various classification, regression and clustering algorithms, including support

vector machines, random forests, gradient boosting, k-means and DBSCAN, and is
designed to interoperate with the Python numerical and scientific libraries NumPy and
SciPy.

Scikit-learn is written mainly in Python and uses NumPy extensively for high-performance
linear algebra and array operations. Furthermore, some core algorithms are written in
Cython to improve performance.

11
A Cython wrapper around LIBSVM implements support vector machines, logistic
regression and linear support vector machines by a similar wrapper around LIBLINEAR.
In such cases, extending these methods with Python may not be possible.

Scikit-learn integrates well with many other Python libraries, such as Matplotlib and Plotly
for plotting, NumPy for array vectorization, Pandas data frames, SciPy, and many more.

2.2.7 Seaborn

Seaborn is a Python data visualization library based on matplotlib. It provides a high-level

interface for drawing attractive and informative statistical graphics.

Seaborn helps to explore and understand the data. Its plotting functions operate on
dataframes and arrays containing whole datasets and internally perform the necessary
semantic mapping and statistical aggregation to produce informative plots.

Its dataset-oriented, declarative API lets you focus on what the different elements of the
plots mean rather than on the details of how to draw them.

There is no universally best way to visualize data. Different plots best answer different
questions. Seaborn makes it easy to switch between different visual representations by
using a consistent dataset-oriented API.

When statistical values are estimated, seaborn uses bootstrapping to compute confidence
intervals and draw error bars representing the estimate's uncertainty.

Statistical analyses require knowledge about the distribution of variables in your dataset.
The seaborn function displot() supports several approaches to visualizing distributions.
These include classic techniques like histograms and computationally-intensive
approaches like kernel density estimation.

12
Some seaborn functions combine multiple kinds of plots to give informative summaries of
a dataset quickly. One, jointplot(), focuses on a single relationship. It plots the joint
distribution between two variables along with each variable's marginal distribution

Fig 2.4: Seaborn plots

13
2.2.8 Jupyter Notebook IDE

Jupyter Notebook (formerly IPython Notebooks) is a web-based interactive computational

environment for creating Jupyter notebook documents. The "notebook" term can
colloquially refer to many different entities, mainly the Jupyter web application, Jupyter
Python web server, or Jupyter document format, depending on context. A Jupyter
Notebook document is a JSON document, following a versioned schema, containing an
ordered list of input/output cells containing code, text (using Markdown), mathematics,
plots and rich media, usually ending with the ". ipynb" extension. A Jupyter Notebook can
be converted to several open standard output formats (HTML, presentation slides, LaTeX,
PDF, ReStructuredText, Markdown, Python) through "Download As" in the web interface
via the nbconvert library or "jupyter nbconvert" command-line interface in a shell. To
simplify the visualization of Jupyter notebook documents on the web, the nbconvert library
is provided as a service through NbViewer which can take a URL to any publicly available
notebook document and convert it to HTML on the fly display it to the user.

Fig 2.5: Jupyter IDE

14
Jupyter Notebook provides a browser-based REPL built upon many popular open-source
libraries:

• IPython

• ØMQ (ZeroMQ)

• Tornado (web server)

• jQuery

• Bootstrap (front-end framework)

• MathJax

Jupyter Notebook can connect to many kernels to allow programming in different

languages. By default, Jupyter Notebook ships with the IPython kernel. The 2.3 release
(October 2014) was 49 Jupyter-compatible kernels for many programming languages,
including Python, R, Julia, and Haskell.

The Notebook interface was added to IPython in the 0.12 release (December 2011),
renamed to Jupyter notebook in 2015 (IPython 4.0 – Jupyter 1.0). Jupyter Notebook is
similar to the notebook interface of other programs such as Maple, Mathematica, and
SageMath, a computational interface style that originated with Mathematica in the 1980s.
According to The Atlantic, Jupyter interest overtook the popularity of the Mathematica
notebook interface in early 2018.

2.2.9 Jupyter kernels

A Jupyter kernel is responsible for handling various requests (code execution, code
completions, inspection) and reply. Kernels talk to the other components of Jupyter using
ZeroMQ and thus can be on the same or remote machines. Unlike many other Notebook-

15
like interfaces, in Jupyter, kernels are not aware that they are attached to a specific
document and can be connected to many clients at once. Usually, kernels allow only a
single language, but there are a couple of exceptions.

The Jupyter Notebook has become a popular user interface for cloud computing, and major
cloud providers have adopted the Jupyter Notebook or derivative tools as a front-end
interface for cloud users. Examples include Amazon's SageMaker Notebooks, Google's
Colaboratory and Microsoft's Azure Notebook.

2.3 ALGORITHMS USED

We have used four different Machine Learning algorithms in this study to compare and
contrast their performances, the algorithms are as follows:

2.3.1 Decision Tree (DT):

Decision Tree makes use of a model, were in a structure that resembles a tree used to make
decisions and their likely outcomes, as well as chance event outcomes, resource costs, and
utility. Each node in the tree is a representation of a conditional statement(‘if’) and on the
whole, the decision tree can be viewed as a representation of a nested conditional.

A mandatory node that is present for all tree is root node. Root node definitely have leaf
nodes which return the decision based on the attribute or parameter on each node. For each
level on the tree two parameters are determined: Information Gain and Entropy.

Entropy: is nothing but quantitative measure randomness or uncertainty. If entropy is 0,

the data has no randomness. Zero value of entropy depicts the decision to be certain and
on the contrary higher values of entropy describes higher levels of uncertainty.

16
Information Gain: It is the change that occurs with entropy after deciding a particular
attribute with respect to the independent variables.

Where, p(x)=is the probability of the attribute

These parameters are again calculated after a decision is returned by a leaf node and the
attribute with the highest IG is removed from the list. This process keeps on repeating
resulting in the depletion of attributes and finally classification of the datasets.

Fig 2.6: Decision Tree

Steps in DT classification are:

1.Importing libraries (for DT) and input dataset in python.

2.The class variable split and the validation split is done to obtain the test and train
dataset.

17
3.The decision tree optimization is done by specifying the criterion for the attribute
selection

4. The evaluation parameters are obtained to compare with the other classifiers

5. Then the model is compiled and fit to give the prediction of faults in the dataset

2.3.2 Random Forest (RF)

Random forest is an ensemble learning (process by which multiple several classifiers
are created and combined to solve a problem) method that is applicable for
classification as well as regression by combining an aggregate of decision trees at
training time and the output of this algorithm is based on the output (can be either mode
or mean/average) of the individual trees that constitute the forest.
The design of RF is such that group of diverse (to an extent) models operate together to
solve a problem. This mathematically can be represented with correlation (low
correlation in this case). This method is proven to be a better option for several problem
throughout time because each tree try to make up for the mistakes made by any other.
DT are very specific on the data that they are trained on, even small changes made to the
set can disturb the structure of the tree
Each tree in RF get data such that each tree is not very similar to any other tree such that
making the behaviour of trees diverse. This is achieved by two methods:
Bagging: random sampling from a dataset and this is done without any replacement
(this is repeated for each individual trees). It is important to note that this method does
not take any unique subset of the dataset and training each DTs on them. Replacement
plays an important role in making sure that DTs make up for each other’s mistakes.

Feature Randomness: Splitting in DTs occur such that there is most separation
between the left and right nodes after considering every feature present in the dataset.
But in the case of RF, as there are only limited features in each tree the splitting occurs

18
with very different separation in individual trees and it is very unlikely that two trees
will be exactly the same.

Fig 2.7: Random forest classifier

2.3.3 Gaussian Naive Bayes

Gaussian Naive Bayes is a variant of Naive Bayes that follows Gaussian normal
distribution and supports continuous data. We have explored the idea behind Gaussian
Naive Bayes along with an example. Before going into it, we shall go through a brief
overview of Naive Bayes.

Naive Bayes are a group of supervised machine learning classification algorithms based on
the Bayes theorem. It is a simple classification technique, but has high functionality. They
find use when the dimensionality of the inputs is high. Complex classification problems
can also be implemented by using Naive Bayes Classifier.

19
2.3.3.1 Bayes Theorem
Bayes Theorem can be used to calculate conditional probability. Being a powerful tool in
the study of probability, it is also applied in Machine Learning.

Fig 2.8: Formula for bayes theorem

Bayes Theorem has widespread usage in variety of domains.

2.3.3.2 Naive Bayes Classifier
Naive Bayes Classifiers are based on the Bayes Theorem. One assumption taken is the
strong independence assumptions between the features. These classifiers assume that the
value of a particular feature is independent of the value of any other feature. In a supervised
learning situation, Naive Bayes Classifiers are trained very efficiently. Naive Bayed
classifiers need a small training data to estimate the parameters needed for classification.
Naive Bayes Classifiers have simple design and implementation and they can applied to
many real life situations.

20
2.3.3.3 Gaussian Naive Bayes
When working with continuous data, an assumption often taken is that the continuous
values associated with each class are distributed according to a normal (or Gaussian)
distribution. The likelihood of the features is assumed to be-

Sometimes assume variance is independent of Y (i.e., σi), or independent of Xi (i.e., σk)

or both (i.e., σ)
Gaussian Naive Bayes supports continuous valued features and models each as conforming
to a Gaussian (normal) distribution.
An approach to create a simple model is to assume that the data is described by a Gaussian
distribution with no co-variance (independent dimensions) between dimensions. This
model can be fit by simply finding the mean and standard deviation of the points within
each label, which is all what is needed to define such a distribution.

21
 Fig 2.9: Gaussian NB classifier

The above illustration indicates how a Gaussian Naive Bayes (GNB) classifier works. At
every data point, the z-score distance between that point and each class-mean is calculated,
namely the distance from the class mean divided by the standard deviation of that class.

Thus, we see that the Gaussian Naive Bayes has a slightly different approach and can be
used efficiently.
2.3.4 Boosting
Ensemble meta-algorithm that is mainly used to reduce bias (group of presumptions
made so the target function is relatively easier to learn) and variance (relative change
in the estimated target function when the training data is changed).

22
2.3.4.1 Ada Boost (AB)

It is ML boosting algorithm, which aggregates several “weak classifiers” (nothing but

classifiers with performance just slightly better than the random chance) to form a
“strong classifier. In the case of AdaBoost the weak learners are also known as decision
stumps. The decision stumps are characterized as weak learners because they are trees
with only a single split. This algorithm mainly focuses on the difficult classification
instances rather than the ones that are clearly classified. The process starts with each
observation having equal weight by training a DT.After the evaluation of the first DT,
the modified such that the observations that are difficult to classify are assigned with
higher weights and vice versa. A new tree is now created and works with this highly
weighted data. This process continues based on the input parameters. These weak
learners are put together to form a new strong learner as mentioned above.

Fig 2.10: Steps in AdaBoost classifier

23
2.3.4.2 Gradient Boost (GB)

Training of models is done in a subsequent and additive method. So, the intuition behind
gradient boosting algorithm is to repetitively leverage the patterns in residuals and
strengthen a model with weak predictions and make it better. The main difference
between AB and GB is the way both these algorithms identify the faults of the weak
learners. The three causes of discrepancy between original and predicted values are
noise, variance, and bias. As noise cannot be reduced only the other two are reduced by
the use of ensemble methods While the AB uses weighted data points, GB uses gradients
in the loss function (LF). LF is used as a measure of how well our model matches the
training data that we are given. A main advantage of GB is that it allows the user to choose
the cost function and optimizes that cost function.

Fig 2.11: Steps in Gradient Boost classifier

24
 CHAPTER-3

DATA VISUALIZATION

3.1 CLEANING OF DATA SET

This is an essential step to perform before creating a visualization. Clean, consistent data
will be much easier to visualize. Clean data is data that is free of errors or anomalies which
may make it hard to use or analyze the data. Starting from a clean dataset allows you to
focus on creating an effective visualization rather than trying to diagnose and and fix issues
while creating visualizations. Data cleaning tasks will be very dependent on the dataset that
you’re working with. In most cases, data cleaning involves:
1.Removing unnecessary variables
2.Deleting duplicate rows/observations
3.Addressing outliers or invalid data
4.Dealing with missing values
5.Standardizing or categorizing values
6.Correcting typographical errors

Fig 3.1: Steps Involved in Data Cleaning

25
The following observations have been made during the process of data cleaning before
visualizing the data

Fig 3.2: Observations of Null Values and Unique Values of parameters

Figure 3.2 shows the number of null values and also the number of unique values of

parameters that are available in the existing data set.

26
 Fig 3.3 : Types of Crops in that are conisdered.

Figure 3.3 shows the different types of crops that have been considered for the project
and the count of each crop of in the dataset

Fig 3.4: Different fertilizers used.

Figure 3.4 shows the different types of fertilizers that have been considered for the project
and the count of each fertilizer of in the data set.

27
3.2 WHAT IS DATA VISUALIZATION?

With the help of data visualization, we can see how the data looks like and what kind of
correlation is held by the attributes of data. It is the fastest way to see if the features
correspond to the output. With the help of following Python recipes, we can understand
ML data with statistics.

Fig 3.5: Types of Data Visualization Techniques

3.2.1 Why Visualization?

Do you think giving you the data of let’s say 1 million points in a table/Database file and
asking you to provide your inferences by just seeing the data on that table is feasible? Unless
you’re a super human its not possible. This is when we make use of Data visualization,
wherein all the data will be transformed into some form of plots and analyzed further from
that. As being a human, we are more used to grasp a lot of info from diagrammatic
representation than the counterparts. Okay ! So since it is said that we need to convert the
data from a boring table into interesting pictorial form like a scatter plot or Bar chart, You
may wonder, How can I do it? Do I need to write my own code for that ? NO! Actually we
can make use of very good packages from some popular programming languages which are
readily available and can make the work pretty much simple with just a single line of code.
That’s the power of modern programming . As a human, we can just visualize anything in

28
either in 2-d or 3-d. But trust me almost of the data that you obtain in real world won’t be
this way. As a Machine learning engineer, working with more than 1000-dimensional data
is very common. So, what can we do in such cases where data is more than 3D? There are
some Dimensionality Reduction (DR) techniques like PCA, TSNE, LDA etc which helps
you to convert data from a higher dimension to a 2D or 3D data in order to visualize
them. There may be some loss of information with each DR techniques, but only they can
help us visualize very high dimensional data on a 2d plot. TSNE is one of the state-of-the-
art DR techniques employed for visualization of high dimensional data.

From perspective of building models, by visualizing the data we can find the hidden
patterns, explore if there are any clusters within data and we can find if they are linearly
separable/too much overlapped etc. From this initial analysis we can easily rule out the
models that won’t be suitable for such a data and we will implement only the models that
are suitable, without wasting our valuable time and the computational resources. This part
of data visualization is a predominant one in initial Exploratory Data Analysis (EDA) on
the field of Data science/ML.

The following techniques have been used to visualize the data from the collected data set
that consists of N, P, K values, temperature, humidity, moisture, crop type, soil type and
fertilizer name corresponding to the respective data.

a. Histograms
b. Correlation Matrix
c. Bar Plots

a. Histogram

Histograms group the data in bins and is the fastest way to get idea about the distribution
of each attribute in dataset. The following are some of the characteristics of histograms −

29
  It provides us a count of the number of observations in each bin created for
visualization.

 From the shape of the bin, we can easily observe the distribution i.e. weather it is
Gaussian, skewed or exponential.

 Histograms also help us to see possible outliers.

The following observations have been made while plotting the histograms of different input
parameters like temperature, moisture, humidity and the macro nutrient values present in
the soil like nitrogen(N), phosphorous(P), potassium(k) (these nutrient values can be
obtained by testing of the soil) excluding the crop type, soil type, and fertilizer names as
they have not been labelled with values. So, labelling has been done for those parameters
to procced further. Different histograms have been plotted so as to get a clear understanding
for the machine to read the data like the number of values of an input parameter that are
present at a particular instant.

These are some of the histograms that have been plotted i.e., humidity, moisture,
temperature, nitrogen, potassium and phosphorus.

Fig3.6: Histogram of humidity

30
Figure 3.6 shows the values of humidity that are in between 15-55 with a highest frequency
of 55, 55-60 occurred with a highest freaquency of 68, 60-65 occurred with a highest
frequency of 60, 65-70 occurred with highest frequency of 29, 70-74 occurred with highest
frequency of 5.

Fig 3.7: Histogram of Moisture

Figure 3.7 shows the values of moisture that are in between 0-30 with a highest frequency
of 24, 30-40 occurred with a highest freaquency of 58, 40-50 occurred with a highest
frequency of 52, 50-60 occurred with highest frequency of 25, 60-65 occurred with highest
frequency of 35.

31
 Fig 3.8: Histogram of nitrogen

Figure 3.2.1(c) shows the values of nitrogen that are in between 0-5 with a highest
frequency of 15, 5-10 occurred with a highest freaquency of 65, 10-15 occurred with a
highest frequency of 115, 15-20 occurred with highest frequency of 20, 20-25 occurred
with highest frequency of 43, 25-30 occurred with a highest frequency of 5, 30-35
occurred with a highest frequency of 40, 35-43 occurred with a highest frequency of 42

Fig 3.9: Histogram of Phosphorous

32
Figure 3.9 shows the values of phosphorous that are in between 0-10 with a highest
frequency of 85, 10-20 occurred with a highest freaquency of 43, 20-30 occurred with a
highest frequency of 38, 30-40 occurred with highest frequency of 35, 40-45 occurred with
highest frequency of 35.

Fig 3.10: Histogram of potassium

Figure 3.10 shows the values of potassium that are in between 0.0-2.5 with a highest
frequency of 255, 2.5-5.0 occurred with a highest freaquency of 0, 5.0-7.5 occurred with
a highest frequency of 10, 7.5-10.0 occurred with highest frequency of 17, 10-12.5
occurred with highest frequency of 15.12.5-15.0 occurred with highest frequency of 7,
15.0-17.5 occurred with highest frequency of 10, 17.5-20.0 occurred with highest
frequency of 11.

b. Correlation Matrix
Correlation coefficients are indicators of the strength of the linear relationship between
two different variables, x and y. A linear correlation coefficient that is greater than zero
indicates a positive relationship. A value that is less than zero signifies a negative
relationship. The correlation coefficient (ρ) is a measure that determines the degree to
which the movement of two different variables is associated. The most common

33
correlation coefficient, generated by the Pearson product-moment correlation, is used to
measure the linear relationship between two variables. However, in a non-linear
relationship, this correlation coefficient may not always be a suitable measure of
dependence.

The possible range of values for the correlation coefficient is -1.0 to 1.0. In other words,
the values cannot exceed 1.0 or be less than -1.0. A correlation of -1.0 indicates a
perfect negative correlation, and a correlation of 1.0 indicates a perfect positive
correlation. If the correlation coefficient is greater than zero, it is a positive relationship.
Conversely, if the value is less than zero, it is a negative relationship. A value of zero
indicates that there is no relationship between the two variables.

Note: When interpreting correlation, it's important to remember that just because two
variables are correlated, it does not mean that one causes the other

Following Correlation Matrix has been obtained for the given input parameters showing
how they are related to each other.

34
 Fig 3.11: Correlation Matrix
The results that are inferred from the above Correlation Matrix are tabulated below

Table 3.1: Observations made from Correlation Matrix

Temperature Humidity Moisture Nitrogen Potassium Phosphorous

Temperature 1.000000 0.094126 -0.029274 -0.025380 -0.029641 0.094987

Humidity 0.094126 1.000000 0.092427 0.009129 -0.031292 0.147147

Moisture -0.029274 0.092427 1.000000 -0.90268 0.057861 0.001713

Nitrogen -0.025380 0.009129 -0.090268 1.000000 -0.48831 -0.629624

Potassium -0.029641 -0.031292 0.057861 -0.488831 1.000000 0.081873

Phosphorous 0.094987 0.147147 0.001713 -0.629624 0.081874 1.0000000

35
 Correlation coefficients are used to measure the strength of the linear relationship
between two variables.
 A correlation coefficient greater than zero indicates a positive relationship while a
value less than zero signifies a negative relationship
 A value of zero indicates no relationship between the two variables being
compared.
 A negative correlation, or inverse correlation, is a key concept in the creation of
diversified portfolios that can better withstand portfolio volatility.
 Calculating the correlation coefficient is time-consuming, so data are often plugged
into a calculator, computer, or statistics program to find the coefficient.

c. Bar Plots
A bar plot shows categorical data as rectangular bars with the height of bars
proportional to the value they represent. It is often used to compare between values of
different categories in the data.

The following Bar Plots have been obtained.

Fig 3.12: Temperature vs Fertilizer

36
 Fig 3.13: Moisture vs Fertilizer

Fig 3.14: Nitrogen vs Fertilizer

Fig 3.15: Potassium vs Fertilizer

37
 CHAPTER-4
RESULTS AND DISCUSSION

4.1 EVALUATION PARAMETERS

4.1.1 Confusion Matrix: In every algorithm, a set of test data will be kept
aside for comparison by the Cross Validation method and the comparison
of the performance of the classifier on this test dataset with the true labelled
values is visualized as confusion matrix.

A Confusion matrix looks like:

Table 4.1: Confusion Matrix Representation

Predicted
Confusion Matrix Yes No

Yes TP FN
Actual No FP TN

Where,

TP = Nets that actually trojan nets and are predicted as

trojan nets. FN = Nets that are actually trojan nets but
predicted as normal nets.
FP = Nets that are actually normal nets but are predicted as
trojan nets. TN = Nets that are actually normal nets and are
predicted as normal nets.

38
4.1.2 Accuracy: Accuracy is defined as the fraction of the predictions that
the classifier got right

4.1.3 Precision: Precision is defined as how often the classifier predicts the output
correctly, irrespective of it being yes or no

4.1.4 Recall: Recall is defined as the number of actual positives our model labels as
positive

4.1.5 F1Score: F1 score is a function of Precision and Recall. It provides a balance

between precision and recall as it is the weighted average of the two

Confusion Matrices of all the proposed Machine Learning algorithms have been obtained i.e., for
Decision Tree, Random Forest, Gaussian NB, Ada Boost, Gradient Boost through which we can infer
that evaluation parameters like accuracy, recall, precision and f-score.

39
 Fig 4.1: Confusion Matrix for Gaussian NB

Fig 4.2: Confusion Matrix for Decision Tree

40
Fig 4.3: Confusion Matrix for Random Forest

Fig 4.4: Confusion Matrix for Gradient Boost

41
Fig 4.5: Confusion Matrix for Ada Boost

42
4.2 GRADIENT BOOST CLASSIFIER

Fig 4.6: Gradient Boost Classifier

Table 4.2: Results of Gradient Boost Classifier

Precision Recall F1-score
0 0.75 0.69 0.72
1 0.40 0.22 0.29
2 0.90 0.92 0.91
3 0.95 0.73 0.79
4 1.00 0.97 0.98
5 0.98 0.93 0.95
6 1.00 0.82 0.90
7 0.75 0.98 0.85
Accuracy 0.88

Macro average 0.83 0.78 0.80

Weighted average 0.88 0.88 0.87

43
4.3 DECISION TREE

Fig4.7: Decision Tree Classifier

Table 4.3: Results of Decision Tree Classifier

Precision Recall F1-score

0 0.80 0.62 0.70
1 057 0.44 0.50
2 0.90 0.92 0.91
3 0.88 0.93 0.90
4 1.00 1.00 1.00
5 0.98 0.93 0.95
6 0.95 1.00 0.97
7 0.95 1.98 0.97
Accuracy 0.93

Macro average 0.88 0.85 0.86

Weighted average 0.93 0.93 0.93

44
4.4 RANDOM FOREST:

Fig 4.8: Random Forest Classifier

Table 4.4: Results of Random Forest Classifier

Precision Recall F1-score

0 0.82 0.69 0.75

1 0.00 0.00 0.00
2 0.88 0.92 0.90
3 0.79 1.00 0.88
4 0.92 1.00 0.96
5 1.00 0.91 0.95
6 1.00 1.00 1.00
7 0.94 1.00 0.97
Accuracy 0.93

Macro average 0.79 0.82 0.80

Weighted average 0.90 0.93 0.91

45
4.5 GAUSSIAN NB

Fig 4.9: Gaussian NB Classifier

Table 4.5: Results of Gaussian NB Classifier

Precision Recall F1-score

0 0.62 0.77 0.69
1 0.40 0.22 0.29
2 1.00 0.92 0.96
3 0.94 1.00 0.97
4 1.00 1.00 1.00
5 0.98 0.93 0.95
6 0.95 1.00 0.97
7 0.95 0.98 0.97
Accuracy 0.94

Macro average 0.85 0.85 0.85

Weighted average 0.93 0.94 0.93

46
4.6 ADA BOOST

Fig 4.10: Ada Boost Classifier

Table 4.6: Results of Ada Boost Classifier

Precision Recall F1-score

0 0.82 0.69 0.75
1 0.00 0.00 0.00
2 0.88 0.92 0.90
3 0.79 1.00 0.88
4 0.92 1.00 0.96
5 1.00 0.91 0.95
6 1.00 1.00 1.00
7 0.94 1.00 0.97
Accuracy 0.93

Macro average 0.79 0.82 0.80

Weighted average 0.90 0.93 0.91

47
4.7 COMPARISION TABLE
Table 4.7: Comparison Table for all proposed algorithms

Decision Tree Gaussian NB Random Forest Ada Boost Gradient Boost

Accuracy 93.33 93.68 92.5 55 87.7

Precision 93 93 90 90 88

Recall 93 94 93 93 88

F-Score 93 93 91 91 87

48
 CHAPTER 5

CONCLUSION AND FUTURE WORK

Utilizing ML, we could anticipate the most appropriate compost for the given harvest
soil n p k qualities assisting the ranchers with expanding the yield. This framework
does the forecast utilizing five distinct ML models. Decision Tree and Gaussian NB
have shown similarities in F-Score that is 93% but have shown 93.33% and 93.68%
respectively in terms of Accuracy. Based on Accuracy and F-score results conclusion
has been made that Gaussian NB makes the accurate prediction. Also further predicted
the suitable fertilizer through Gaussian NB model when provided with the parameters
by the end user. In addition to this work the dataset can be extended by getting more
values from the analysis of the soil and extend the work for more crops. A user
interface can be developed so that the farmers can directly get their results from the
given inputs. Using advanced algorithms in Machine Learning also increase the level
of accuracy.

49
 REFERENCES

[1] Bondre, Devdatta A., and Santosh Mahagaonkar. "Prediction of crop yield and fertilizer
recommendation using machine learning algorithms." International Journal of Engineering
Applied Sciences and Technology 4, no. 5 (2019): 371-376.

[2] Archana, K., and K. G. Saranya. "Crop Yield Prediction, Forecasting and Fertilizer
Recommendation using Voting Based Ensemble Classifier." SSRG Int. J. Comput. Sci.
Eng 7 (2020).
[3] Kim, Yun Hwan, Seong Joon Yoo, Yeong Hyeon Gu, Jin Hee Lim, Dongil Han, and
Sung Wook Baik. "Crop pests prediction method using regression and machine learning
technology: Survey." IERI Procedia 6 (2014): 52-56.
[4] Kalimuthu, M., P. Vaishnavi, and M. Kishore. "Crop Prediction using Machine
Learning." In 2020 Third International Conference on Smart Systems and Inventive
Technology (ICSSIT), pp. 926-932. IEEE, 2020.
[5] Kumar, Y. Jeevan Nagendra, V. Spandana, V. S. Vaishnavi, K. Neha, and V. G. R. R.
Devi. "Supervised Machine learning Approach for Crop Yield Prediction in Agriculture
Sector." In 2020 5thInternational Conference on Communication and Electronics Systems
(ICCES), pp. 736-741. IEEE, 2020.
[6] Medar, Ramesh, Vijay S. Rajpurohit, and Shweta Shweta. "Crop yield prediction using
machine learning techniques." In 2019 IEEE 5th International Conference for
Convergence in Technology (I2CT), pp. 1-5. IEEE, 2019
[7] Nigam, Aruvansh, Saksham Garg, Archit Agrawal, and Parul Agrawal. "Crop yield
prediction using machine learning algorithms."In 2019 Fifth International Conference on
Image Information Processing (ICIIP), pp. 125-130. IEEE, 2019.
[8] Jahan, Nusrat, and Rezvi Shahariar. "Predicting fertilizer treatment of maize using
decision tree algorithm." Indonesian Journal of Electrical Engineering and Computer
Science 20, no. 3 (2020): 1427-1434.

50
[9] Kanuru, L., Tyagi, A.K., Aswathy, S.U., Fernandez, T.F.,Sreenath, N. and Mishra, S.,
2021, January.Prediction of Pesticides and Fertilizers using Machine Learning and Internet
of Things.In 2021 International Conference on Computer Communication and Informatics
(ICCCI) (pp. 1-6). IEEE. Materials Science and Engineering, vol. 1022, no. 1, p. 012104.
IOP Publishing, 2021.

[10] Motia, Sanjay, and S. R. N. Reddy. "Method for dataset preparation for soil data
analysis in decision support applications." In IOP Conference Series:

[11] Pandiarajaa, P. "A survey on machine learning and text processing for pesticides and
fertilizer prediction." Turkish Journal of Computer and Mathematics Education
(TURCOMAT) 12, no. 2 (2021): 2295-2302.

[12] Maeda, Yuichiro, Taichi Goyodani, Shunsaku Nishiuchi,and Eisuke Kita. "Yield
prediction of paddy rice with machine learning."In Proceedings of the International
Conference on Parallel and Distributed Processing Techniques and Applications (PDPTA),
pp.361-365. The Steering and Committee of The World Congress in Computer
Science,Computer Engineering and Applied Computing (WorldComp),2018

51
 APPENDIX

#importing the libraries

import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import seaborn as sns
from sklearn.model_selection import GridSearchCV
from sklearn.metrics import accuracy_score

df=pd.read_csv("C:/Users/S S Manoj/Downloads/Fertilizer Prediction1.csv")

df.head()
type(df)
df.nunique()
df.shape
df.isnull().sum()
df.dtypes
df.columns,df.dtypes
df['Fertilizer Name'].value_counts().to_dict()
data=df.drop(labels="Fertilizer Name",axis=1)
y=df['Fertilizer Name']
df.hist(figsize = (20,20))
plt.show()
corrmat=df.corr()
fig = plt.figure(figsize = (12,9))
sns.heatmap(corrmat, vmax =.8, square=True)
plt.show()
corrmat

52
from sklearn import preprocessing
label_encoder = preprocessing.LabelEncoder()
df['Fertilizer Name']= label_encoder.fit_transform(df['Fertilizer Name'])
df['Fertilizer Name'].unique()
df['Fertilizer Name'].value_counts().to_dict()
fig_dims = (20, 4)
fig, ax = plt.subplots(figsize=fig_dims)
sns.barplot(x='Moisture',data=df,y='Fertilizer Name')
sns.barplot(x='Temparature',data=df,y='Fertilizer Name')
fig_dims = (20, 4)
fig, ax = plt.subplots(figsize=fig_dims)
sns.barplot(x='Phosphorous',data=df,y='Fertilizer Name')

fig_dims = (20, 4)
fig, ax = plt.subplots(figsize=fig_dims)
sns.barplot(x='Nitrogen',data=df,y='Fertilizer Name')
fig_dims = (20, 4)
fig, ax = plt.subplots(figsize=fig_dims)
sns.barplot(x='Phosphorous',data=df,y='Fertilizer Name')
fig_dims = (20, 4)
fig, ax = plt.subplots(figsize=fig_dims)
sns.barplot(x='Potassium',data=df,y='Fertilizer Name')
data['Crop Type'].value_counts().to_dict()
from sklearn import preprocessing
label_encoder = preprocessing.LabelEncoder()
data['Crop Type']= label_encoder.fit_transform(data['Crop Type'])
data['Crop Type'].unique()
data['Crop Type'].value_counts().to_dict()
data['Soil Type'].value_counts().to_dict()

53
from sklearn import preprocessing
label_encoder = preprocessing.LabelEncoder()
data['Soil Type']= label_encoder.fit_transform(data['Soil Type'])
data['Soil Type'].unique()
data['Soil Type'].value_counts().to_dict()
from sklearn.model_selection import train_test_split
ip=data
op=df['Fertilizer Name']
X_train,X_test,y_train,y_test=train_test_split(ip,op,test_size=0.7,random_state=1)
X_train
from sklearn.naive_bayes import GaussianNB
from sklearn.metrics import classification_report, confusion_matrix
nb = GaussianNB()
nb.fit(X_train, y_train)
n_pred=nb.predict(X_test)
print('Accuracy of GaussianNB classifier on training set: {:.3f}'.format(nb.score(X_train,
y_train)))
print('Accuracy of GaussianNB classifier on test set: {:.3f}'.format(nb.score(X_test,
y_test)))
print('Classification Report:')
print(classification_report(y_test,n_pred))
print('\n')
print('Confusion Matrix:')
data=confusion_matrix(y_test, n_pred)
sns.heatmap(data,annot=True,cmap="summer")
plt.show()
from sklearn.tree import DecisionTreeClassifier
dtc = DecisionTreeClassifier().fit(X_train, y_train)
dtc.fit(X_train, y_train)

54
d_pred = dtc.predict(X_test)
print('Accuracy of Decision Tree classifier on training set: {:.2f}'
.format(dtc.score(X_train, y_train)))
print('Accuracy of Decision Tree classifier on test set: {:.2f}'
.format(dtc.score(X_test, y_test)))
from sklearn.metrics import classification_report
k_range = range(1, 10)
scores=[]
for k in k_range:
clf2 = DecisionTreeClassifier(max_depth = k).fit(X_train, y_train)
scores.append(clf2.score(X_train, y_train))
print(max(scores))
print('Accuracy of Decision Tree classifier on training set: {:.2f}'
.format(clf2.score(X_train, y_train)))
print('Accuracy of Decision Tree classifier on test set: {:.2f}'
.format(clf2.score(X_test, y_test)))
print("Classification Report")
print(classification_report(y_test,d_pred))
print('\n')
data=confusion_matrix(y_test, d_pred)
sns.heatmap(data,annot=True,cmap="summer")
plt.show()
from sklearn.ensemble import RandomForestClassifier
classifier = RandomForestClassifier(n_estimators=50,max_depth=10, random_state=1)
classifier.fit(X_train, y_train)
from sklearn.metrics import classification_report, accuracy_score
print('Accuracy of Random Forest classifier on training set: {:.2f}'
.format(classifier.score(X_train, y_train)))
print('Accuracy of Random Forest classifier on test set: {:.2f}'

55
 .format(classifier.score(X_test, y_test)))
r_pred=classifier.predict(X_test)
print("Classification Report")
print(classification_report(y_test,r_pred))
print('\n')
data=confusion_matrix(y_test, r_pred)
sns.heatmap(data,annot=True,cmap="summer")
plt.show()
from sklearn.ensemble import GradientBoostingClassifier

#Create Gradient Boosting Classifier

gb = GradientBoostingClassifier()

#Train the model using the training sets

gb.fit(X_train, y_train)

#Predict the response for test dataset

g_pred = gb.predict(X_test)
x_pred = gb.predict(X_train)
from sklearn import metrics
# Model Accuracy, how often is the classifier correct?
print("Accuracy of gradient boost:",metrics.accuracy_score(y_test, g_pred))
print("Classification Report")
print(classification_report(y_test,g_pred))
print('\n')
data=confusion_matrix(y_test, g_pred)
sns.heatmap(data,annot=True,cmap="summer")
plt.show()
from sklearn.ensemble import AdaBoostClassifier

56
from sklearn import metrics
clff =
AdaBoostClassifier(n_estimators=100,learning_rate=1,random_state=1).fit(X_train,
y_train)
print('Accuracy of adaboost classifier on training set: {:.2f}'
.format(clff.score(X_train, y_train)))
print('Accuracy of adaboost classifier on test set: {:.2f}'
.format(clff.score(X_test, y_test)))
a_pred=classifier.predict(X_test)
print("Classification Report ")
print(classification_report(y_test,a_pred))
print('\n')
data=confusion_matrix(y_test, a_pred)
sns.heatmap(data,annot=True,cmap="summer")
plt.show()
print("Accuracy scores of each model")
print("Decision tree : ",accuracy_score(y_test,d_pred))
print("Random forest : ",accuracy_score(y_test,r_pred))
print("GaussianNB : ",nb.score(X_test, y_test))
print("Adaboost : ",clff.score(X_test, y_test))
print("Gradient Boost : ",accuracy_score(y_test,g_pred))
def one():
temperature=int(input('Enter temperature'))
humidity=int(input('Enter humidity'))
moisture=int(input('Enter moisture'))
soiltype=int(input('Enter soil Type 0-black 1-clayey 2-loamy 3-red 4-sandy'))
croptype=int(input('Enter Crop Type 0 :Tobacco 1 :Cotton 2 :Ground Nuts 3 :Maize 4
:millets 5 :oil seeds 6 :pulses 7 :paddy 8 :sugarcane 9 :Barley 10:wheat'))
nitrogen=int(input('Enter mitrogen value'))

57
 potassium=int(input('Enter potassium value'))
phosphorous=int(input('Enter phosphorous value'))

temp=nb.predict([[temperature,humidity,moisture,soiltype,croptype,nitrogen,potassium,p
hosphorous]])
return temp

def two():
temperature=int(input('Enter temperature'))
humidity=int(input('Enter humidity'))
moisture=int(input('Enter moisture'))
soiltype=int(input('Enter soil Type 0-black 1-clayey 2-loamy 3-red 4-sandy'))
croptype=int(input('Enter Crop Type 0 :Tobacco 1 :Cotton 2 :Ground Nuts 3 :Maize 4
:millets 5 :oil seeds 6 :pulses 7 :paddy 8 :sugarcane 9 :Barley 10:wheat'))
nitrogen=int(input('Enter mitrogen value'))
potassium=int(input('Enter potassium value'))
phosphorous=int(input('Enter phosphorous value'))

temp=dtc.predict([[temperature,humidity,moisture,soiltype,croptype,nitrogen,potassium,p
hosphorous]])
return temp

def three():
temperature=int(input('Enter temperature'))
humidity=int(input('Enter humidity'))
moisture=int(input('Enter moisture'))
soiltype=int(input('Enter soil Type 0-black 1-clayey 2-loamy 3-red 4-sandy'))

58
 croptype=int(input('Enter Crop Type 0 :Tobacco 1 :Cotton 2 :Ground Nuts 3 :Maize 4
:millets 5 :oil seeds 6 :pulses 7 :paddy 8 :sugarcane 9 :Barley 10:wheat'))
nitrogen=int(input('Enter mitrogen value'))
potassium=int(input('Enter potassium value'))
phosphorous=int(input('Enter phosphorous value'))

temp=classifier.predict([[temperature,humidity,moisture,soiltype,croptype,nitrogen,potas
sium,phosphorous]])
return temp

def four():
temperature=int(input('Enter temperature'))
humidity=int(input('Enter humidity'))
moisture=int(input('Enter moisture'))
soiltype=int(input('Enter soil Type 0-black 1-clayey 2-loamy 3-red 4-sandy'))
croptype=int(input('Enter Crop Type 0 :Tobacco 1 :Cotton 2 :Ground Nuts 3 :Maize 4
:millets 5 :oil seeds 6 :pulses 7 :paddy 8 :sugarcane 9 :Barley 10:wheat'))
nitrogen=int(input('Enter mitrogen value'))
potassium=int(input('Enter potassium value'))
phosphorous=int(input('Enter phosphorous value'))

temp=clff.predict([[temperature,humidity,moisture,soiltype,croptype,nitrogen,potassium,
phosphorous]])
return temp

def five():
temperature=int(input('Enter temperature'))
humidity=int(input('Enter humidity'))

59
 moisture=int(input('Enter moisture'))
soiltype=int(input('Enter soil Type 0-black 1-clayey 2-loamy 3-red 4-sandy'))
croptype=int(input('Enter Crop Type 0 :Tobacco 1 :Cotton 2 :Ground Nuts 3 :Maize 4
:millets 5 :oil seeds 6 :pulses 7 :paddy 8 :sugarcane 9 :Barley 10:wheat'))
nitrogen=int(input('Enter mitrogen value'))
potassium=int(input('Enter potassium value'))
phosphorous=int(input('Enter phosphorous value'))

temp=gb.predict([[temperature,humidity,moisture,soiltype,croptype,nitrogen,potassium,p
hosphorous]])
return temp

switcher = {
1: one,
2: two,
3: three,
4: four,
5: five
}
print("1:Guassian Naive bayes 2:Decision Tree 3:randomForest")
print("4:Adaboost 5:GradientBoost")
n=int(input('Enter model'))
func =switcher.get(n)
print(func())

60
61