ISSN 2278-3091

Putul Siddharth et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(3), May - June 2021, 1844 – 1850
Volume 10, No.3, May - June 2021
International Journal of Advanced Trends in Computer Science and Engineering
Available Online at http://www.warse.org/IJATCSE/static/pdf/file/ijatcse501032021.pdf
https://doi.org/10.30534/ijatcse/2021/501032021

Agrobot in field of Machine Learning

Putul Siddharth1, B V A N S S Prabhakar Rao2
1
Department of Computer Science and Engineering, Vellore Institute of technology, Chennai, India,
sidputul1997@gmail.com
2
prabhakarrao@vit.ac.in

 From the application, the project will assist farmers in making

ABSTRACT better and faster decisions about which crop to grow and in
risk management. Other features, such as blog entries, crop
Farming is still one of the largest sectors of India from a information, agricultural loan information, and vacant land
financial standpoint as well as employing almost two-thirds of information would also be able to pique the attention of both
our working population. Yet the farming industry suffers from novices and experts.
not being attuned with modern technological advancements.
Hence farming industry has not reached its optimum potential 1.1 Objectives
and has a lot of areas for improvement. Our project tries to
tackle some of these problems. The project’s main focus is on Following are the objectives of the paper:
a crop recommendation system that was built using Machine  To collect agricultural datasets that has soil
Learning techniques. Along with this, we have worked to chemistry and location-related information for
make a website as a UI for our recommendation system. different crops. Doing data processing to help in
Finally, we have also added a lot of agriculture-related building the recommender systems. Carrying out
infotainment as there is a lot of misinformation related to dataset analysis on various ML models.
agriculture and also to pique the interest of the user in the field  Build crop predictor system using soil chemistry
of modern agriculture. Through the whole process of dataset with the help of Random Forest as our ML
implementing this project, our team learned about Machine algorithms.
Learning, Web Development, Content Creation, and most  Build crop recommender system using
importantly the scientific method of Agriculture for the 21st location-based dataset and crop-based dataset with
century. Our design ideology revolved around making our the help of KNN algorithm.
system a holistic place for anyone who wants to get an  To host a website to make the system user-friendly.
introduction to modern farming. The germination of the  Implementing account creation system in website.
project is from the need to modernize agriculture and thus the  Providing agriculture-related blog posts and other
project intends to tackle issues from the grass root. information for users.
Key words: Decision tree, KNN, Logistic regression, Naïve 2. LITERATURE SURVEY
Bayes, Random forest, Recommendation system, Support
vector machine. The paper[1] focused on the need for crop yield prediction in
the modern-day world and uses various lazy and eager learner
1. INTRODUCTION algorithm to tackle this issue. They aim at finding a suitable
data model that gives the highest accuracy against their
Agriculture is one of India's most important industries, but dataset.
more can be done to carry it into the twenty-first century by The paper[2] focused on crop recommendation systems using
making it more scientific and industrially precise. There are data mining techniques. A system is developed in which the
many approaches to achieving this aim, and our project user has to enter the soil nutritional attributes through an
addresses some of them. Our framework aims to assist in crop interface and the system will map it to the crop dataset with
recommendation, as well as provide agriculture-related the data values collected over the years and thus a
information. corresponding class of crop is predicted using the Naive
India, as a predominantly agricultural nation, requires Bayes Algorithm. There will be an admin panel that is
agricultural innovation. One way we're attempting to achieve responsible for handling and managing the entire application.
this is by offering a crop recommendation system to a farmer The system has been developed to provide benefits to the
or other consumer, allowing them to choose the best crop to farmer in terms of crop and risk management.
grow based on their area and type of soil. This is intended to The paper[3] emphasizes the need for a Fuzzy Model-based
eliminate inefficiency in the agriculture sector, which is crop recommender framework for farmers. Here the dataset is
already an unorganized sector. prepared through Questionnaire Assessment from the district
Jaipur of state Odisha. The questionnaire consists of twenty

1844
Putul Siddharth et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(3), May - June 2021, 1844 – 1850

Questions on different attributes of soil and weather It has the following attributes:
parameters along with rice production. This questionnaire was  Nitrogen (N): Part of the "Enormous 3(NPK)"
asked from the maximum qualified people of the district in supplements needed by plants for their food. Any
their native language. The system asks for location, crop plant ingests Nitrogen the most of all nutrients.
preferences, and season as input. Now the system mapped this Nitrogen is fundamental in the development of
set of input to the dataset to check the yield of the proteins which are urgent for the development of any
corresponding crop. The system used three input two output plant. Soil Nitrogen esteem is accordingly
Mamdani Fuzzy inference models for prediction. The study fundamental in deciding the kind of harvests that
was limited only to rice production and has not taken many could develop there as various yields require a
weather attributes into account. shifting level of nitrogen.
The paper[4] is about a crop monitoring framework that uses  Phosphorus (P): Second of the “Big 3(NPK)”
machine learning and the Internet of Things. The soil nutrients required by plants for their sustenance.
parameters are collected with the help of sensors then these Phosphorus is used by plants to store energy and also
collected data are uploaded into NodeMcu which has a to do photosynthesis. It also aids in plant growth and
built-in Wi-Fi module. Now, these uploaded data are sent to development so it becomes crucial for any crop. As
the ThingSpeak cloud for generating responses to the android mentioned for N, soil P values also differ from crop
application. Finally, the responses generated from the cloud to crop.
are sent to the ML model for triggering appropriate crop  Potassium (K): Third of the “Big 3(NPK)” nutrients
recommendations. The system also has a price prediction required by plants for their sustenance. Potassium is
section that estimates the price of the yield using the linear responsible for plant immunity and increasing yield.
regression model. K helps strengthen root systems in dry/cold
conditions. As stated for N & P, every crop needs a
3. METHODOLOGY different amount of Potassium to thrive.
 pH: pH can be considered an important factor as it
3.1 Dataset Collection affects many other soil attributes and processes that
For purpose of this paper, we searched for datasets that would influence plant growth. The behavior of
allow us to make a crop recommender with multiple microorganisms, as well as the solubility and
attributes. We ideally wanted a single dataset that would have availability of nutrients, are all pH-dependent
soil chemistry data and location data respective to different processes so when the pH of the soil is about seven,
crops. After searching the web extensively for datasets most crops thrive.
matching our requirements and after analyzing different  Temperature: Temperature is perhaps the main
datasets, we have settled on three different datasets which will variables in cultivating as irregular characteristics in
be discussed below. it can prompt demise of harvests, failure of manures
among different issues. Right temperature is
3.1.1Soil chemistry based dataset additionally crucial for give better yield
The soil chemistry dataset, a part of which is displayed in quantitatively and subjectively.
Table 1. is the main dataset for the project. The dataset has  Humidity: A feature that is indirectly dependent on
2200 rows and 8 columns. air temperature thus making it very important for
crop production. More humidity tends to cause foliar
Table 1: Soil chemistry based dataset or root diseases in plants thus increasing pesticide
costs and decreasing yield. On the other hand, low
humidity leads to slow growth and more leaf drops.
 Rainfall: Another very important environmental
feature for crop production. Less of it can starve
crops or increase production costs as artificial
watering would have to be ramped up. More of it
would mean the inundation of crops or dilution of
pests and fertilizers making crops vulnerable to
diseases.
 Crop: The dataset maps specific crops like rice with
the above 7 attributes. Since most crops have a
tolerance range in each of the 7 attributes that is why
there are multiple entries per crop.

This dataset was used to build the prediction system. This

dataset has 22 varieties of crop like rice, watermelon, etc. with
100 entries for each type to get a better model for each crop
and thus making the dataset more uniform avoiding any bias
due to less data for any specific crop.

1845
Putul Siddharth et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(3), May - June 2021, 1844 – 1850

3.2 Crop Prediction

3.1.2 Crop and Location based dataset As we all know that crop prediction is of great use to global
food production as many of them rely on accurate prediction
Table 2: Crop based dataset to make timely import and export decisions and can make
both ends meet. The crop prediction starts with importing the
dataset from Kaggle and necessary packages and libraries are
installed. Heat map is then used to show the correlation
between different attributes of a dataset. Then cleaning the
dataset is done in case there are null values present in the
dataset.
After that Normalization of the dataset is performed using
Standard Scalar to increase the efficiency of our ML model.
Finally, we have used different machine learning algorithms
like DecisionTreeClassifier, GaussianNB, Support Vector
Machine, Logistic Regression, and RandomForestClassifier
against our dataset to compare the accuracy and the below
table showed the same.

The crop-based dataset and location-based dataset are the

other dataset used for the project.The initial aim was to find
one dataset that also had location-related information along
with soil chemistry and environmental factors related to crops.
Table 3: Location based dataset

.Figure 1: Snapshot of heat map

pH, Nitrogen (N), Phosphorus (P), and Potassium (K) are the
seed parameters that are used to classify the crop. Crop
parameters are used to characterize a diverse range of crops
grown in India's fertile lands. The soil dataset includes data
samples collected from various locations in Chennai and other Figure 2: Algorithm vs. Accuracy Comparison
Tamil Nadu districts. pH, Nitrogen (N), Phosphorus (P), and
Potassium (K) is the soil parameters that are used to classify Now we considered Random Forest as our classification
the soil. The data samples were collected from the algorithm for further prediction and use the pickle module to
Tamilnadu-Kancheepuram district's approved horticulture save our model. So whenever the user enters the required
department. To collect feedback from users, the device input through an interface, this model is used to predict the
employs an interface. Depending on the module they choose, crop and display it to the user.
the input form can vary.

1846
Putul Siddharth et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(3), May - June 2021, 1844 – 1850

3.3 Crop Recommendation provide blog-style content that any reader could understand
Along with crop prediction, our project also encourages the and that would allow them to follow or at the very least be
recommendation of a set of different crops to the farmers aware of modern agricultural methods. In addition, the system
based on two methods either via their city location or via their provides the consumer with information on crops, agricultural
soil nutrients value. Similar work is done for prediction, loans, and available vacant land for cultivation.
which involves dataset import, cleaning, and normalizing the This was done by creating HTML pages for each page that can
dataset, but for crop recommendations, we used k-nearest be displayed to the user. The pages were stylized using CSS
neighbors. Bootstrap. The contact information entered by the user, blog
posts, and vacant land information is stored in PHPMyAdmin
Databases.

3.5Architectural Diagram

Figure 3: Snapshot of heat map

In the first part, when the user inputs his city location through
the interface, the given input will be mapped to the Figure 5: Flowchart of how user uses the system
location-based dataset to gather the soil parameters of the
corresponding city. Then this list of parameters will be
The user first needs to have an account to access the
mapped to that crop dataset and with the help of the KNN
application. For that, they need to sign up by filling in some of
classification algorithm, this list of the parameter will find
the personal details and set up strong login credentials. Then
themselves a suitable class of crop and thus the result is
they can log in by entering correct credentials to utilize
displayed to the user through an interface.
different functionalities accessible. At the homepage, the user
enjoys the different blog posts related to topics like
agro-education, agro-news, etc. Then the user has crop
prediction functionality available in which they can get a
single predicted crop after entering the required details in the
portal. Also, the system provides a crop recommendation
facility in which users can get ten different options of the crop
suitable for their soil and location.
Now out of the recommended crops if the user is interested in
a particular crop but is unaware of its process of cultivation
then he can get that knowledge from the information sector of
our application. There he can know about health benefits,
climate required, soil requirement, land preparation, planting
Figure 4: System architecture diagram and spacing, irrigation, manures and fertilizers, harvest, yield,
and demand of that crop in the market.
In the subsequent part, the user needs to include soil Even if a farmer needs some loan to fund his seasonal
supplements for crop suggestions. The provided input will be agricultural operation and other activities but is unaware of
saved in a variable, and the crop dataset will be used to map it. the process of getting an agricultural loan, he can get
With the help of the K-Nearest Neighbor classification information from the Agricultural Loan sector available in the
algorithm, the input locates itself into a crop class and the information menu. There he can know about different types of
same will be displayed to the user using an interface. agricultural loans, agricultural loan interest for different banks
Subsequently, the user will get a list of crops suggested by the in India, the process to apply for an agricultural loan,
system. documents required, and eligibility criteria to borrow an
agricultural loan.
3.4 Agricultural Infotainment The system also proved beneficial for peasants as peasants
In contrast to other sectors, there is a lack of well-presented don’t own agricultural land and has to search for that vacant
agricultural infotainment. As a result, the project aimed to land here and there. So, the system will provide a feature for

1847
Putul Siddharth et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(3), May - June 2021, 1844 – 1850

the user who wants to lend his land for cultivation on a profit extensive as was expected however they were unbiased
basis. The user needs to fill in some details like owner name, towards any specific crop because each of the 22 crops had
description about land, location, and type of peasants which 100 data instances each. After analyzing different ML
he needs, contact details, etc. Now the system will broadcast algorithms, Random forest was chosen as it gave 98.93
this information to all the users including the peasants. So the percent accuracy.
one who needs this can contact the owner and help For the recommendation system, 10 different suggestions
themselves. were generated for each entry using the KNN classification
Finally, after using the application if the user has some query algorithm. “predict_proba” function of MultiOutputClassifier
or complaint, he can get in touch with the admin by just filling was used for this.
in a contact form.

Figure 6: Flowchart of how admin uses the system

There is an admin panel in which the admin has the privilege

to add, edit and delete the blog shown on the user homepage.
Also, he can review and delete the vacant land post after
getting contacted by the corresponding landowner.
For adding or deleting posts our strategy is simple if the serial
number is zero then add all form data as a new post in the Figure 8: Snapshot of Crop Recommendation widget portal based
database but if the serial number is any other number then edit on location
that number of blogs. To add data to the database we will use
the “Posts” class in which we pass the object of Sqlalchemy.
We will pass form data through arguments and then add the
data to the database through “db.session.add ()”.

4. RESULTS AND DISCUSSION

Figure 7: Snapshot of Crop Prediction Figure 9: Snapshot of recommended crops based on location

Using Random forest as our classification algorithm, a The most relevant crops are listed above based on the location
prediction system was built. The datasets used were not as entered by user in the portal.

1848
Putul Siddharth et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(3), May - June 2021, 1844 – 1850

Crops grown before on soil will make the dataset more robust.
For now, the ml model works with acceptable accuracy.Due
to Covid-19, the lockdown has been imposed all over the
country and people are not in a state to move outside their
house. So this recommendation system will be able to help
farmers to choose the crops to grow with more precision
instead of going to any agricultural center to gain information
so that they can avoid losses and maximize profits. By the
creation of the E-commerce site, I hope casual readers will
become interested in agriculture and will thus help in making
farming scientific and efficient.

REFERENCES

1. D Ramesh, B Vishnu Vardhan. “Data Mining

Techniques and Applications to Agricultural
Yield Data”. International Journal of Advanced
Research in Computer and Communication
Engineering Vol. 2, Issue 9, September 2013, pp.
3477-3480.
2. Bharath Kumar R, Balakrishna K, BencyCelso A,
Siddesha M, Sushmitha R, “Crop
Figure 10: Snapshot of Crop Recommendation widget portal based Recommendation System for Precision
on soil nutrients value. Agriculture,” International Journal of Computer
Sciences and Engineering, Vol.7, Issue.5,
pp.1277-1282, 2019.
3. MadhusreeKuanr, BikramKesariRath,
SachiNandanMohanty(2018), ‘ Crop
Recommender System for the Farmers using
Mamdani Fuzzy Inference Model’, International
Journal of Engineering & Technology, 7, (4.15),
(2018), 277-280
4. R.Pallavi Reddy, B. Vinitha, K. Rishita,
K.Pranavi(2020), ‘ Crop Monitoring and
Recommendation System using Machine
Learning and IOT’, ‘ International Journal of
Innovative Technology and Exploring Engineering
(IJITEE)’, 2278-3075, Volume-9 Issue-9, July 2020.
5. Bhanumathi, M.Vineeth and N.Rohit , “Crop Yield
Prediction and Efficient use of Fertilizers” , IEEE
International Conference on Communication and
Signal Processing, April 4-6, (2019), pp. 0769-0771.
6. M. Ayaz, M. Ammad-Uddin, Z. Sharif, A. Mansour
and E. -H. M. Aggoune, "Internet-of-Things
(IoT)-Based Smart Agriculture: Toward Making
the Fields Talk," in IEEE Access, vol. 7, pp.
129551-129583, 2019, doi:
10.1109/ACCESS.2019.2932609.
7. J. J. I. Haban, J. C. V. Puno, A. A. Bandala, R.
Figure 11: Snapshot of recommended crops based on soil nutrients KerwinBillones, E. P. Dadios and E. Sybingco, "Soil
value entered by user. Fertilizer Recommendation System using Fuzzy
Logic," 2020 IEEE REGION 10 CONFERENCE
5. CONCLUSION (TENCON), 2020, pp. 1171-1175, doi:
10.1109/TENCON50793.2020.9293780.
The recommendation system was created upon a sub-optimal
8. A.Chandavale, A.Dixit, A.Khedkar and
dataset. We are trying to improve the dataset by collecting
R.B.Kolekar, "Automated Systems for Smart
more data. We are also trying to include parameters so that the
Agriculture," 2019 IEEE Pune Section
dataset mimics the real world much better. Including
International Conference (PuneCon), 2019, pp. 1-6,
parameters like crop yield (both by weight and by money),
doi: 10.1109/PuneCon46936.2019.9105686.

1849
Putul Siddharth et al., International Journal of Advanced Trends in Computer Science and Engineering, 10(3), May - June 2021, 1844 – 1850

9. Suchithra M.S and Maya L. Pai, “Improving the

prediction accuracy of soil nutrient classification
by optimizing extreme learning machine
parameters”, China Agricultural University.
Production and hosting by Elsevier (2019), pp. 1-6.
10. Sadia Afrin, Abu Talha Khan, MahrinMahia,
RahbarAhsan, MahbuburRahmanMishal, Wasit
Ahmed and Rashedur M. Rahman, “Analysis of Soil
Properties and Climatic Data to Predict Crop
Yields and Cluster Different Agricultural
Regions of Bangladesh”, Journal of IEEE (2018),
pp. 80-82.
11. J. Li and L. Zhou, "Research on Recommendation
System of Agricultural Products E-Commerce
Platform Based on Hadoop," 2018 IEEE 9th
International Conference on Software Engineering
and Service Science (ICSESS), 2018, pp. 1070-1073,
doi: 10.1109/ICSESS.2018.8663921.
12. P. A, S. Chakraborty, A. Kumar and O. R.
Pooniwala, "Intelligent Crop Recommendation
System using Machine Learning," 2021 5th
International Conference on Computing
Methodologies and Communication (ICCMC), 2021,
pp. 843-848, doi:
10.1109/ICCMC51019.2021.9418375.
13. S. Namani and B. Gonen, "Smart Agriculture
Based on IoT and Cloud Computing," 2020 3rd
International Conference on Information and
Computer Technologies (ICICT), 2020, pp. 553-556,
doi: 0.1109/ICICT50521.2020.00094.
14. GizealewAlazieDagnaw, SisayEbabyeTsigie,
“Applying Logistic Regression Data mining
techniques for Ethiopian Government
Agricultural Open Data Sets”, International
Journal of Advanced Trends in Computer Science
and Engineering, Volume 10, No.2, March - April
2021
15. Joel Mathew Toms, Zebadiah S. Wahlang N., Dr. M.
RoshniThanka, Dr. E. Bijolin Edwin, “Innovative
Agricultural Information System with
UserFriendly Digital Assistance for Farmers”,
International Journal of Advanced Trends in
Computer Science and Engineering, Volume 10,
No.2, March - April 2021

1850