Volume 9, Issue 1, January – 2024 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

A Novel Strategy for Predicting

Agriculture Crop and its Yield
Shobhit Tembhre1 (Assistant Professor); Gopireddy Venkata Rukhmanand Reddy2; Chadive Devendra Reddy3
Sudalakuntaakhil4; Penekalapati Pavan Kalyan5
School of Computer Science and Engineering, Presidency University, Bengaluru,India

Abstract:- Agriculture holds a pivotal role in India, knowledgeable mathematician predicting the likelihood of a
making substantial contributions to the nation's GDP particular class's membership, such as estimating the
(23%) and offering significant employment opportunities probability of specific crop yields under givenconditions.
(59%), thereby securing the country's food needs.
However, the sector faces challenges due to climate II. LITEREACTURE REVIEW
change, impacting crop predictability and resulting in
lower yields. Recognizing the potential of technology to  To achieve accurate harvest predictions, they propose the
address these issues, particularly machine learning, this utilization of a supervised machine learning method
research project aims to guide inexperienced farmers in employing a boosted naïve Bayes Gaussian classifier.
adopting advanced crop prediction methods. Machine This model predicts the output seed based on specific
literacy, a form of machine learning, is proposed as a input parameters. Our initiative aims to greatly benefit
solution to enhance agricultural practices by leveraging less-experienced farmers lacking crop forecasting
data and experience to make informed decisions, expertise for sustainable agriculture. Furthermore, future
ultimately improving yields and sustainability in the face enhancements may involve providing tailored
of climate change. recommendations for fertilizers and guidelines suitable
for different farms and crops, considering the input
I. INTRODUCTION provided.[1]
 The utilization of a real-time dataset significantly
Agriculture has traditionally been recognized as the improves the program's ability to make precise
primary source of daily necessities for humans, playing a predictions, providing farmers with more valuable
pivotal role in India as a major industry and primary guidance for crop selection. This model, which
occupation. The promotion of ecological diversity and the incorporates artificial neural networks (ANN) and linear
production of nutritious crops has been advocated, urging regression with both forward and backward propagation,
farmers to adopt natural observation methods and abstain demonstrated an 82% accuracy in predicting the
from using chemicals on both their fields and animals. dependent variable with minimal loss. The proposed
Nevertheless, the swiftly changing modern climate presents backpropagation model, integrating gradient descent and
a formidable obstacle, jeopardizing essential resources and the RELU activation function, is aimed at minimizing
undermining food supply and security. the mean square error (MSE). [2]
 The article delves into multiple machine learning
Adding to these difficulties is the diminishing
techniques for forecasting agricultural production,
contribution of the agricultural sector to the nation's GDP. focusing on variables like temperature, precipitation,
The sector's GDP share stood at approximately 17.2% in season, and region. Through experiments utilizing
2005, dwindled to 11.1 % in 2012, further decreased to 5% datasets from the Indian government, it was determined
in 2018, and reached a meager 2% in the initial quarter of
that the Random Forest Regressor stands out as the most
2020. As 80% of farmers reside in rural areas, reductions in
accurate predictor for crop yield forecasts. Furthermore,
crop production revenue at the industry level significantly the study highlights that the Simple Recurrent Neural
impact their way of life. The ongoing fluctuations in the
Network, functioning as a sequential model,
agricultural sector and the resulting economic repercussions
demonstrates superior accuracy in predicting rainfall
underscore the pressing need for sustainable practices and
compared to Long Short-Term Memory (LSTM) models.
strategies to address the challenges posed by a changing
[3]
climate.
 A model for forecasting soil series and suggesting an
In the context of crop prediction, machine learning appropriate crop yield for that particular soil is put forth.
leverages historical data from previous years and combines Six upazila’s in the Khulna region's soil datasets were
it with current monthly data to showcase the precision of used for the research. The model has been put to the test
meteorological information. Operating on the idea that using several machine learning algorithms. While KNN
computers can analyze data and make decisions with minimal and bagged trees have decent accuracy, SVM has the
human input, machine learning serves as a method for greatest accuracy of all the classifiers when it comes to
automating models. A practical analogy is a logical soil categorization. [4]
classifier within machine learning, resembling a

IJISRT24JAN627 www.ijisrt.com 404

Volume 9, Issue 1, January – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
 The suggested approach facilitates the precise definition  Proposed System Design:
of features influencing crop yield. Its efficacy lies in The diagram provided below illustrates the
combining a self- experimental analysis using Deep Q- fundamental configuration of the system we are currently
Networks (DQN) with feature processing through designing (Figure 1). The initial step involves the loading of
Recurrent Neural Networks (RNN). Employing a Deep an agriculture- related dataset. To improve accuracy, data
Recurrent Q-Network (DRQN) presents a preprocessing is undertaken, encompassing tasks such as
comprehensive solution by autonomously uncovering the filling in missing values and converting strings to numeric
intricate relationships between crop yield and various formats. The construction of the model incorporates
environmental factors such as climate, soil, and machine learning techniques like Naive Bayes, Random
groundwater. This contrasts with conventional Forest Regressor, and Linear
supervised learning-based methods for crop yield
prediction. The advantage of this method is a reduced Regression. Following this, the model undergoes
reliance on expert input and historical knowledge during training using the pre-processed data. The accuracy of the
the model creation process for estimating crop model is subsequently evaluated through testing, employing
productivity. Consequently, the proposed technique various machine learning algorithms.
appears to employ a more comprehensive model for
predicting crop yield.[5] After determining the algorithm with the highest
 Leveraging a real-time dataset improves the program's accuracy, the cultivation of crops will be guided by the yield
accuracy in making predictions, providing farmers with predictions generated by the model. It is important to note
more valuable insights for crop selection. The model that our comprehension of the project is currently
achieved an 82% accuracy in predicting the dependent incomplete at this initial architectural phase, but these gaps
variable with minimal loss, utilizing artificial neural are expected to diminish as theproject progresses.
networks (ANN) and linear regression incorporating
forward and backward propagation. Additionally, the
model includes a feature that assesses the crop success
rate based on the farmer's input, ultimately suggesting
the crop with the highestprobability of success. [6]
 The investigation leverages readily available data from
government repositories, encompassing details such as
state, district, crop, season, year, and production
information. Various regression algorithms, including
Random Forest, Gradient Boost, and Decision Tree, are
tested, and ensemble methods are employed to enhance
accuracy. Distinguishing itself from other studies, this
research relies on easily accessible information such as
state, district, and season, rendering it practical for
farmers. The model demonstrates proficient accuracy in
predicting crop yields, with noticeable improvement
observed when utilizingensemble methods.[7]
 The study has successfully developed and deployed an
intelligent crop recommendation system, accessible to
farmers across India. This technology aids farmers in
choosing the optimal crop for cultivation, taking into
account diverse geographical and environmental
parameters. Potential future enhancements may involve
integrating a feature into the described model that
anticipates crop rotations. Consequently, the system
would consider the crop harvested in the previous cycle
when recommending the current crop, ensuring
maximum production.[8]

III. METHODOLOGY Fig 1 Proposed System Design Model

As commonly known, our application comprises two  Data Flow Diagram:

integral components: Machine Learning and Web The diagram below, denoted as the Data Flow Diagram
Development. I will commence by delving into the (Figure 2), illustrates the flow of data within the project.
foundational aspect of our application, which is the machine Farmers, acting as clients, initially input details like state
learning methodology. Before delving into the details, allow name, district, season, area, and production into the user
me to provide a concise overview. interface. These input values are then transmitted to the
machine learning model. Utilizing the knowledge gained
during training, the model takes on the responsibility of

IJISRT24JAN627 www.ijisrt.com 405

Volume 9, Issue 1, January – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
identifying the most suitable crop for achieving the highest storage of client details. accuracy among various models.
yield. Furthermore, in cases where a lower yield is Opting for Random Forest as our final model was decided
anticipated, the model suggests alternative crops for upon due to its superior accuracy.
maximizing productivity. The user interface displays either
the recommended crop or the projected yield based on the The finalized model is now saved as a pickle file,
model's output. prepared for future use when new data on crop or yield
predictions becomes available. Following this, using HTML
For a more comprehensive grasp of the details and CSS, we create a user interface or webpage. Users, upon
provided, please consult the image presented below, visiting the webpage, can register for the first time by
indicated as Figure 2. providing their details. This information is stored in the
database, allowing for subsequent logins and contributions.
In our quest to connect the machine learning model The model utilizes these input values for predictions, and
with the webpage, we chose the Flask framework, a Python- the results are showcased on the user interface. Integration
based solution. The initial step involves converting the model of the database and our ultimate model with the user
into a pickle file for easy invocation and utilization. interface is facilitated through a Python Flask framework.
Subsequently, Flask is linked to the database, enabling the

IV. RESULT

 Evaluation:

Fig 2 Data Flow Diagram

Let's initiate our examination by focusing on the precise predictions for both crop type and yield. Diverse
backend. We utilized a range of machine learning techniques were applied to enhance the model's accuracy,
techniques, including XGB Regressor, Random Forest, and Figure (3) illustrates a comparison of In the testing
Decision Tree Classifier, and Linear Regression, to craft our phase, the model effectively predicts the class label,
model. After preprocessing the data, we created train and test providing insights into the recommended crop and the
datasets. The model underwent training using the train dataset, expected yield. Ongoing evaluations of the model's
employing the mentioned algorithms to establish multiple performance involve the utilization of test data, with a
models. Python served as the programming language in this subsequent comparison of the results. The testing process
process. Evaluating the model's accuracy, we selected the has yielded notable success, achieving an accuracy rate of
final model based on the highest accuracy, ensuring more 98% when applying the model to the test dataset.

IJISRT24JAN627 www.ijisrt.com 406

Volume 9, Issue 1, January – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165

Fig 3 Accuracy Comparison

V. CONCLUSION REFERENCES

The agricultural sector is facing significant challenges [1]. M. Kalimuthu, P. Vaishnavi and M. Kishore,
related to global sustainability, which calls for creative "Crop Prediction using Machine Learning," 2020
approaches to improve productivity while protecting Third International Conference on Smart Systems
environmental resources. The project showcases the efficacy and Inventive Technology (ICSSIT), Tirunelveli,
of machine learning in providing tailored crop India, 2020, pp.926-932, doi:10.1109/ICSSIT
recommendations and yield prediction based on different 48917.2020.9214190.
parameters. Its success signifies a step towards optimizing [2]. R. Medar, V. S. Rajpurohit and S. Shweta, "Crop
agricultural practices, promoting sustainability, and offering Yield Prediction using Machine Learning
valuabledecision-making tools for farmers in cultivatingcrops Techniques," 2019 IEEE 5th International
best suited for their specific conditions. Continued Conference for Convergence in Technology (I2CT),
advancements in this field hold significant potential for Bombay, India, 2019, pp. 1-5, doi:
revolutionizing farming practices and addressing agricultural 10.1109/I2CT45611.2019.9033611.
challenges globally. The primary driving forces behind crop [3]. A. Nigam, S. Garg, A. Agrawal and P. Agrawal,
prediction and agriculture initiatives are the need to "Crop Yield Prediction Using Machine
maintain food security in the face of an expanding global Learning Algorithms," 2019 Fifth International
population, maximize resource utilization, and boost Conference on Image Information Processing (ICIIP),
productivity. We are happy with our 98%accuracy rate. Shimla, India, 2019, pp. 125-130,doi:
10.1109/ICIIP47207.2019.8985951.
FUTURE SCOPE [4]. S. A. Z. Rahman, K. Chandra Mitra and S. M.
Mohidul Islam, "Soil Classification Using Machine
Further developments in artificial intelligence and Learning Methods and Crop Suggestion Based on
machine learning algorithms may result in increasingly Soil Series," 2018 21st International Conference of
complex crop prediction models. Better algorithms can shed Computer and Information Technology (ICCIT),
more light on the variables affecting crop yield, growth, and Dhaka, Bangladesh, 2018, pp. 1-4, doi:
disease resistance. Incorporating blockchain technology into 10.1109/ICCITECHN.2018.863194
the agricultural supply chain can improve traceability and [5]. D. Elavarasan and P. M. D. Vincent, "Crop Yield
transparency. This makes sure the information used to Prediction Using Deep Reinforcement Learning
forecast crops is reliable and contributes to the development Model for Sustainable Agrarian Applications," in
of a more dependable system. Working with emerging IEEE Access, vol. 8, pp. 86886-86901,2020, doi:
AgTech companies cancreate new opportunities for 10.1109/ACCESS.2020.2992480.
implementing cutting-edge tools and technologies into [6]. S. S. Kale and P. S. Patil, "A Machine Learning
agricultural and crop forecasting initiatives. Startups Approach to Predict Crop Yield and Success Rate,"
frequently offer original concepts and cutting-edge methods 2019 IEEE Pune Section International Conference
for solving problems. (PuneCon), Pune, India, 2019, pp. 1-5, doi:
10.1109/PuneCon46936.2019.9105741.

IJISRT24JAN627 www.ijisrt.com 407

Volume 9, Issue 1, January – 2024 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[7]. P. S. Nishant, P. Sai Venkat, B. L. Avinash and B.
Jabber, "Crop Yield Prediction based on Indian
Agriculture using Machine Learning," 2020
International Conference for Emerging Technology
(INCET), Belgaum, India, 2020, pp. 1-4, doi:
10.1109/INCET49848.2020.9154036.
[8]. Z. Doshi, S. Nadkarni, R. Agrawal and N. Shah,
"AgroConsultant: Intelligent Crop Recommendation
System Using Machine Learning Algorithms," 2018
Fourth International Conference on Computing
Communication Control and Automation
(ICCUBEA), Pune, India, 2018, pp. 1-6, doi:
10.1109/ICCUBEA.2018.8697349.

IJISRT24JAN627 www.ijisrt.com 408