Crop Recommendation and Yield

Estimation Using Machine Learning

A. Ashwitha1,∗ and C. A. Latha2

1
Dept of ISE, MSRIT and Research Scholar-Dept of CSE, AMCEC, Bangalore,
India and affiliated to Visvesvaraya Technological University, Belagavi, Karnataka,
India
2
Department of CSE, AMCEC, Bangalore, India and affiliated to Visvesvaraya
Technological University, Belagavi, Karnataka, India
E-mail: ashwitha.a.1990@[Link]; drlathaca@[Link]
∗
Corresponding Author

Received 18 September 2021; Accepted 05 December 2021;

Publication 22 January 2022

Abstract
In most developing countries like India, Agriculture is seen as one of
the most widely followed habitations and contributes majorly to the coun-
try’s economy. Agriculture provides the primary source of food, income,
livelihood and employment to the majority of rural populations in India.
Many crops are destroyed every year due to a lack of technical knowledge
and unpredictable weather patterns such as temperature, rainfall, and other
atmospheric parameters, which play a massive role in deciding the crop
yield and profit. Therefore, choosing the right crop to grow and enhancing
crop yield is an essential aspect of improving real-life farming scenarios.
One of the motives is to collect and integrate the agricultural data from
specific regions that may be used to analyse the optimal crop and estimate
the crop yield. This script is novel by using simple crop, soil and weather
parameters like crop, the area under cultivation, nitrogen, phosphorus and
potassium content of the soil, season, average rainfall and temperature of
a district in Karnataka, India. The user can predict the most suitable crop

Journal of Mobile Multimedia, Vol. 18 3, 861–884.

doi: 10.13052/jmm1550-4646.18320
© 2022 River Publishers
862 A. Ashwitha and C. A. Latha

and its estimated yield for a chosen year. This model uses primary classifica-
tion, techniques like the random forest, k-NN, logistic regression, decision
tree, XGBoost, SVM and gradient boosting classifier for determining the
most suitable crop and regression algorithms like Linear Regression, k-
NN, DBSCAN, Random Forest and ANN algorithm to estimate the yield
of the most optimal crop identified earlier. The algorithm that has the least
mean error is chosen for prediction and estimation and thus gives better
results than the particular machine learning algorithm domain. There is a
web interface for ease of use for end-users. Therefore, this project assists
the farmers in choosing the suitable crop that can be grown in a particular
region during a specific season or specific period and estimate its yield and
predict if the recommended crop is profitable. Hence this project helps the
farmers in preserving their time by assisting them in the decision-making
process.

Keywords: Agriculture, machine learning, classification algorithms, regres-

sion algorithms, crop recommendation, yield estimation.

1 Introduction
Agriculture is the Indian economy’s lifeline, as it employs around 70–75 per
cent of the people directly or indirectly. Agricultural growth is closely propor-
tional to India’s economic growth. However, due to a variety of constraints,
such as changing climatic conditions, bad soil conditions, soil erosion, land
drifting, and floods, Indian farmers face various obstacles. If agriculture fails,
it has a significant impact on the productivity of other industries, which harms
the country’s gross domestic product ratio. India’s growth in agriculture has
remained stagnant, and the government requires a technological revolution
to fulfil the country’s expanding population demands. Thus it is essential for
us to adopt new technologies in agriculture and enhance crop production.
This will help the farmers to avoid or minimise loss and achieve profit.
Initially, the crop and yield predictions were performed based on farmer’s
experience in a particular region. And they don’t have adequate knowledge
on the changing weather patterns, soil components of phosphorus, potas-
sium, nitrogen and the soil ph. Therefore this inadequate knowledge may
lead to soil acidification, reduced yield, and damaging the top layer of the
soil by the application of insufficient and improper amounts of nutrients
to the land and without the crop rotation. The crop yield mainly depends
on environmental changes, weather conditions, water resources, and rainfall.
 Crop Recommendation and Yield Estimation Using Machine Learning 863

Hence the farmers may not be able to achieve the expected or higher yield.
Machine learning, data mining and deep learning are the essential analytical
technologies that support accurate decision making in crop yield prediction,
which includes some of the assisting conclusions on which crop to grow and
the decisions regarding the crops in the growing season on the agricultural
land.

Need for Prediction

Food is one of humanity’s most fundamental needs. The agricultural sector
and its output determine it. In addition, the agrarian yield generation is
strongly reliant on Indian food businesses. However, many farmers have
committed suicide due to unfavourable climatic conditions and low yield
output. Over 59,000 farmers are said to have attempted suicide in the last
30 years. Everyone nowadays is talking about agriculture and its production,
but no one has come up with a methodology to assist them by increasing their
crop productivity. A growing number of scholars are focusing their efforts on
the agricultural industry. However, only a rudimentary attempt has been made
in this regard. Motivation for the proposed system comes from the farming
point of view that there is much work for the farmers to be done manually.
This work focuses on algorithms that can forecast the most suited crop and
estimate crop yield, assisting farmers in selecting and cultivating the most
profitable crop, lowering the risk of loss and boosting the productivity and
value of their agricultural area. Hence the proposed model assists the farmers
in the decision-making process by recommending them the most optimal
crop that, when grown, can produce greater yield so that with the help of
technology, the agricultural sector in India will be progressed with innovative
ideas. The past agricultural data is used in analysing crop production with
the help of machine learning techniques, and it also provides a new way of
understanding and analysing the various crops in the region of Karnataka and
their crop productivity.

2 Literature Survey
Various machine learning, data analytics, data mining, and deep learning
techniques are previously used to analyse the agricultural data and draw
valuable conclusions.
[1] have developed a web page using data mining to determine the crop
yield on the basis of several climatic parameters as input. The C 4.5 algorithm
864 A. Ashwitha and C. A. Latha

has been used to find the most predominant climatic input parameter on the
yield of the crop in specific districts of Madhya Pradesh. The environmental
data with respect to the previous years like temperature, rainfall were col-
lected. It focuses on giving the idea of how the growth of the crop is affected
by various input parameters of climate. [2] Developed a Java application to
determine the crop yield rate by using two methods, namely, the K-nearest
neighbours method and the Naive Bayes method. It helps in achieving the
maximum rate of yield for the crops. It also assists in choosing an optimal
crop suitable for a specific region and the environmental conditions. From
the input dataset, the accuracy of the two methods is compared, and the
best method is chosen for predicting. [3] proposed a system that uses data
mining techniques to predict the harvest yield of the crop. This forecast helps
the farmer to know his requirements and can plan accordingly. This system
makes use of the Random forest algorithm for predicting the output of crop
yield. [4] have analysed the agricultural data using data analytics and machine
learning to assist the farmers by predicting the productivity of the crop in
order to maximise crop production. They have made use of three algorithms,
namely, 1. Naı̈ve Bayes, 2. K-means Clustering 3. A priori Algorithm for final
data accumulation and to predict the crop and its estimated yield. [5] have
developed a model that uses the id3 algorithm for predicting crop yield
and getting excellent quality and improved yield for the crop of Tomato.
The proposed model is executed using PHP. The various parameters used
in this study are Area, Temperature, humidity, and tomato crop production.
[6] describes the rainfall and crop yield prediction using machine learning
techniques. This paper describes the various machine learning techniques
for predicting the rain and also predicting crop yield. It mentions the effi-
ciency and the performance of various machine learning algorithms like
SVM, decision tree, linear regression and KNN. It is concluded that the
SVM algorithm has the highest efficiency or performs the best for rainfall
prediction.

3 Proposed Approach
This paper proposes a method for crop recommendation and yield estimation
which is very important in the field of agriculture. Outcome can be visualised
in the form of a user-friendly web interface.
The overview and the workflow of the project can be seen in Figure 1,
which involves the steps involved and the sequential order that has to be
 Crop Recommendation and Yield Estimation Using Machine Learning 865
y g p

Figure 1 Overview and analysis of proposed project.

performed in order to achieve the final results. The design of the proposed
project of crop recommendation and yield estimation system involves the
following steps:
A. Data Collection and integration: It is the most crucial and essential
step. The proposed model begins with the collection of the necessary
agricultural data that consists of various parameters. It requires integrat-
ing data from multiple sources such as the soil parameters that include
the nitrogen, phosphorus and potash values as are necessary for the
crops, the weather parameters such as the season, rainfall, temperature,
and Ph concerning a particular region and crops.
B. Data Pre-Processing: The next step after data collection and integration
is the data pre-processing, which also means cleaning the data and
preparing it for modelling. The initial part involves handling the missing
values, removing the noisy data, and handling the outliers. The next part
is encoding the categorical values into numerical values using the label
encoder to be easier in modelling the data. The next part is scaling or
normalising the data by using the Minmax scalar or Standardscalar.
C. Splitting the data: Once the data is pre-processed, the data is divided
into two, i.e. the train data, which is used for training the model and the
test dataset, which is used to test the model.
D. Applying Machine Learning Algorithms: In this step, we apply
various classification and regression machine learning algorithms and
evaluate each model against the performance metrics.
866 A. Ashwitha and C. A. Latha

Crop Recommendation: This module uses machine learning classifica-

tion algorithms for modelling the data and recommending the suitable
crop. The classification algorithms use dare:
• Decision tree algorithm
• K-Nearest neighbours algorithm
• Random forest algorithm
• Logistic regression algorithm
• SVM algorithm
• XGBoost algorithm
• Gradient Boosting classifier
• Naive Bayes algorithm
Yield Estimation: This module uses machine learning regression algo-
rithms to model the data and estimate the crop yield. The regression
techniques used are:
• Linear Regression algorithm
• K-Nearest Neighbours algorithm
• Algorithm
• DBSCAN algorithm
• Random Forest algorithm
• Stochastic Gradient Descent Algorithm
• K means clustering
• Agglomerative Clustering
E. Data Exploration and Visualization: This step involves exploring the
data through visualisations to understand the properties and charac-
teristics of the dataset and the results obtained in a better and easily
understandable way.
F. Comparison of the Machine learning models: The classification
Models of crop recommendation and the regression models of yield
estimation are compared and evaluated against various performance
metrics. The crop recommendation models are compared with accu-
racy, precision and recall. The yield estimation models are assessed
against mean absolute error. The algorithms that perform best in crop
recommendation and yield estimation respectively are chosen.
G. User Interface: The final output of the proposed project is a web
application developed using a flask. The machine learning models of
both crop recommendation and yield estimation are transformed into
pickle models for integrating it with the flask application. The Web
Application helps the users in accessing and using it efficiently. By
 Crop Recommendation and Yield Estimation Using Machine Learning 867

giving the required parameters for crop recommendation and yield esti-
mation, the user will know the best crop suitable for a particular land
and its estimated yield in an optimised manner.

4 Crop Prediction Algorithms

The list of algorithms that compared used for crop prediction is as follows:

Decision Tree
A decision tree functions as a decision support tool, with each branch repre-
senting a tree-like model decision. The decision tree results are insignificant
for small datasets because of the number of options available at each branch,
and the tree’s height is easy to overfit [7, 10, 12, 17, 18, 21, 23–25].

Logistic Regression
The logistic model is used in statistics to model the probability of a particular
classification of an entity or an event occurring, such as fail or pass, win
or lose, or positive or negative. This model can be transformed to replicate
events, such as determining whether the image has a cow, buffalo, horse or
pony. Each detected object in the image would be assigned a max probability
of one, all of which adds up to one [14, 22].

Support Vector Machine

Decision planes and boundaries are used in SVM. It divides the complete set
of data into two categories. It works better when the data is multidimensional,
and the number of dimensions exceeds the number of samples. When there is
quite some separation between the classes, it works reasonably well. Feature
scaling is required to achieve better outcomes. When the hyperplane has three
or more properties, it is not easy to visualise [12, 23, 26].

XGBoost
This is a decision tree algorithm that makes use of a gradient boosting
framework. It can handle missing values. It has built-in regularisation, which
prevents the model from overfitting. It is more likely to be overfitting than
bagging. It requires a solid dataset and correctly chosen hyperparameters,
which results in improved accuracy [22].
868 A. Ashwitha and C. A. Latha

Gradient Boosting
Gradient boosting is a technique that enhances the weak classifiers, like
decision trees, for regression, classification, and other problems. A Gradient
boosted decision tree can easily outperform a random forest in multiple
scenarios. It carries out model construction in a stage-wise manner, sim-
ilar to previous boosting approaches, then generalises them by allowing
optimisation of any differentiable loss function [23].

Naı̈ve Bayes
The naive assumption taken for building this classifier is that the presence of
one feature or attribute is independent of the existence of another set of traits
or characteristics. In comparison to numerical variables, it works particularly
well with categorical input variables. It can accurately estimate the class of a
test dataset and operates quickly. The loss of categorical variables that were
not observed in the training dataset might significantly impact the outcomes
because it is foolish to assign zero probability to them. The underlying
limitation is the assumption that the presence of one attribute is dependent
on the presence of other qualities [12].

5 Yield Estimation Algorithms

The list of algorithms that compared used for yield estimation is as follows:

Linear Regression
The linear regression attempts to shape the relationship between two variables
with the linear equation of the observed data. One variable is explanatory,
while the other is considered to be dependent. For example, a person might
want a linear regression model to match people’s weight to their height [8,
14, 16, 19, 24, 26].

K-Nearest Neighbours
In n-dimensional space, the clustering process is classified based on the k-
nearest data points. The multi-class problem is simple to implement. It’s
straightforward to use. It is continually changing. The data quality determines
the accuracy of the algorithm. And when the size grows, the calculations grow
cubically. It is a computationally expensive algorithm [12, 15, 20, 22, 23].
 Crop Recommendation and Yield Estimation Using Machine Learning 869

Artificial Neural Network

An ANN is built from a network of connected units or nodes known as
artificial neurons, which are loosely modelled after the neurons in the human
brain. Each link, like synapses in a human brain, can send a signal to other
neurons. An artificial neuron that receives a signal analyses it and can signal
neurons to which it is attached. The “signal” at a connection is a multiplier,
and each neuron’s output is generated by the non-linear function of the sum of
its inputs. The connection between neurons is referred to as edges. Neurons
and edges usually have a weight that changes as learning progresses [11, 20,
22, 23].

DBSCAN
DBSCAN (Density-Based Spatial Clustering of Applications with Noise)
is used for spatial density-based clustering of data points in n-dimensional
space. It is capable of handling clusters of diverse forms and sizes. It is
noise-resistant. The DBSCAN algorithm is incompatible with data with
variable spatial density. It also does not perform well with high-dimensional
data [13].

Random Forest
This is a classification algorithm that entails the construction of many deci-
sion trees. It reduces overfitting and enhances accuracy. It works effectively
with both continuous and categorical data. The results are difficult to interpret
because of the vast number of decision trees. When dealing with climatic
variability, selecting the appropriate ranges of values to make decisions is
critical [22, 23].

Stochastic Gradient Descent

SGD is an optimisation approach used in Machine Learning applications
to discover the best-fit attributes that predict the actual values. Because the
network processes one of the data instances simultaneously, it simply lays
through the memory. It is analytically faster because it only analyses one
training sample at a time. Because of the large epoch values, the steps
taken towards the minima are small and noisy. Furthermore, it takes longer
to converge the loss function for obtaining the minima due to the noisy
behaviour [22].
870 A. Ashwitha and C. A. Latha

Machine Learning Regression Algorithm

The user can choose out of two functionalities: crop prediction and yield
estimation based on their requirement and enter the necessary inputs to
obtain the desired results. Crop and yield recommendation is an approach
that predicts the most optimal crop and the estimated yield by making
use of various parameters such as temperature, rainfall, soil constituents
and other environmental conditions. The main objective is to collect the
agricultural dataset of a particular region, which can be used to analyse
crop yield forecasting. The agrarian data is subject to several machine
learning, data mining and deep learning techniques, and the comparison
between these algorithms is made to recommend the crops more accu-
rately. The agricultural data plays an important role, and using a large
dataset with a more significant number of farming parameters, could help
achieve better results. The farmer or the user of the application will have
to enter the parameters such as selecting the district, choosing the season,
Average temperature in Celsius value, NPK values, average rainfall and
area of the field in hectare for the crop prediction and an added input of
crop chosen for the yield prediction. After entering the information, the
user will have to press the ‘Predict crop’ button on the crop prediction
page and the ‘Predict yield’ button on the yield prediction page. The cor-
rect predictions will be made on the user’s choice, and the result of the
best crop which can be grown and predicted yield will be displayed in
kilotons.
The pseudocode for the combination of results from the above set of
chosen algorithms is as follows:

1: pred = None # store the predicted output

2: data = [] #store the input
3:
4: index()
5: # Renders home page
6:
7: predict(pred)
8: # Renders the crop prediction page.
9: if pred!=None:
10: # Render page with result
 Crop Recommendation and Yield Estimation Using Machine Learning 871

11: else
12: # Render page without results
13: # Ask the users to fill the form
14: input(data)
15: submit()
16: if valid (data) == True:
17: # Predict crop
18: pred = predict crop(data)
19: # Redirect with output.
20: predict(pred)
21: else
22: # Show error message
23: end if
24: end if
25:
26: yield predict(pred)
27: # Renders the yield prediction page.
28: if pred!=None:
29: # Render page with result
30: else
31: # Render page without results
32: # Ask the users to fill the form
33: input(data)
34: submit()
35: if valid (data) == True:
36: # Predict yield
37: pred = predict yield(data)
38: # Redirect with output.
39: yield predict(pred)
40: else
41: # Show error message
42: end if
43: end if
44:
45: about()
46: # Render about us Page
872 A. Ashwitha and C. A. Latha

Flowchart
The flow diagram of the Methodology is depicted in Figure 2 below.

Figure 2 Flowchart.

Results
Data Collection and Integration
The data available in this work is obtained for the years from 1997 to 2014 in
30 districts of Karnataka in India.
A sample of integrated dataset can be shown in Figure 3.
State = Karnataka
District Year Season Area Temp pH Rainfall N P K Crop
Udupi 2008 Rabi 39 25.15 5.3 123.21 45.3 8.1 6.2 maize
Chickballapura 2012 Kharif 134 26.32 6.5 82.34 56.8 9.4 4.5 Lobia
Bagalkot 2007 Kharif 1345 27.51 5.6 59.12 34.9 0 3.1 sunflower
Chickmaglur 2002 Kharif 3456 28.72 6.7 18.67 76.1 0 2.2 Tur
Bellary 2013 Kharif 7868 23.91 5.2 17.34 23.2 1.6 9.1 Tur
Dharwad 2001 whole 4565 34.22 6.3 34.15 59.3 7.8 5.3 Onion
Gulbarg 2007 Rabi 6786 27.43 5.4 23.34 20.4 9.0 8.5 sunflower
Bijapur 2012 Rabi 12 28.69 6.2 25.11 34.5 4.4 3.3 urad
Raichur 2011 Kharif 123 28.34 5.3 67.55 35.6 8.2 2.2 Mustard
Kolar 2002 Rabi 235 29.12 6.5 58.23 37.7 2.1 9.7 Mustard
Haveri 2002 Kharif 4345 30.23 5.8 23.34 53.2 0 1. Sugarcane
Koppal 2007 Rabi 6778 26.44 6.9 37.19 41.3 0 4.3 moong
Bangalore 2007 Rabi 345 25.55 5.7 45.2 78.3 3.5 4.7 jowar
Hasan 2012 Kharif 3556 27.45 6.8 52.4 34.4 5.7 6.8 raagi
Mysore 1998 Rabi 768 28.31 5.8 98.8 23.9 8.5 7.6 rice
Belgaum 2000 Rabi 997 29.12 7.5 79.2 25.3 9.3 3.2 urad
Dharwad 2008 Kharif 675 24.23 6.9 83.2 37.1 7.2 2.1 linseed
Uttarkarnataka 2003 Kharif 5678 23.34 5.2 62.3 59.2 6.2 1.2 moong
Koppal 2005 Rabi 3536 26.12 7.1 13.5 19.4 5.1 4.3 rice
kodagu 2011 Rabi 34346 28.67 5.4 92.8 28.7 0 8.3 gram
Yadgiri 2010 Kharif 3677 29.90 7.9 21.2 31.2 0 7.9 jowar

Figure 3 Sample data.

 Crop Recommendation and Yield Estimation Using Machine Learning 873

Data Pre-Processing
The dataset collected had no null or missing, or duplicate values so, it did not
require cleaning. After verifying this fact, the categorical are encoded into
numerical values as shown in Figure 4.

State = Karnataka
District Year Area Production Temp pH Rainfall N P K Crop
2 1997 11268 0.16 34.7 6.71 62.9 4.38 12.56 12.56 0

3 1997 10477 0.317 31.7 6.72 61.73 7.18 10.92 10.99 0

4 1997 6057 0.68 33.87 7.47 60.32 6.65 10.85 10.85 0

5 1997 48463 0.072 34.39 7.23 61.48 7.10 12.76 12.76 0

6 1997 20286 0.1196 33.04 7.19 64.18 4.02 12.69 12.69 0

2

Figure 4 Data after preprocessing.

Splitting the Data

The collected dataset containing 6730 rows are split into test and train data.
The test data has about 30% data or 2019 rows, whereas the training dataset
includes the remaining 4711 rows.
The size vs data graph for the data split is depicted in Figure 5, which aids
us in making appropriate analysis.

Figure 5 Plots of Data splitting for better visualisation.

874 A. Ashwitha and C. A. Latha

Data Analysis
After analysing the graphs in Figures 6 to 8 which consists of Predicted yield
vs Temperature for MOONG, we can identify that the optimal average annual
temperature range for max yield for MOONG (GREEN GRAM) lies between
33 to 34.5 degrees Celsius. In contrast, the least favourable temp range is
31-to-32.5-degree Celsius.

Figure 6 Predicted yield vs temperature over the years for MOONG (GREEN GRAM) in
HASSAN.

Figure 7 Predicted yield vs temperature over the years for MOONG (GREEN GRAM) in
BELGAUM.
 Crop Recommendation and Yield Estimation Using Machine Learning 875

Figure 8 Predicted yield vs temperature over the years for MOONG (GREEN GRAM) in
BIJAPUR.

The performance metrics are Accuracy for Crop prediction, whereas Root
Means Square Error(RMSE) for Yield prediction. After comparing all the
models as shown in the Figure 9, we concluded the Random Forest algorithm
gave the highest accuracy among algorithms used for crop prediction and
least RMSE among algorithms used for yield prediction. Random Forest
achieved an accuracy of 99.94%. And RMSE of 0.2716.

Figure 9 The Mean absolute error comparison of the Machine learning models.
876 A. Ashwitha and C. A. Latha

The Random Forest Regressor has the lowest mean absolute error of all
the machine learning algorithms, as shown in Figure 9. As a result, it’s best
for predicting crop yield.

Table 1 The mean absolute error comparison of the machine learning models for yield
estimation
Machine Learning Models Mean Absolute Error
Random Forest Regressor 0.270
Artificial Neural Net 0.3052
Linear Regression 0.478
KNN Regressor 0.303
DBSCAN 1.054
SGD Regressor 0.478

Figure 10 Accuracy Comparison of the classification models of machine learning for Crop
Recommendation.

Crop recommendation uses several classification algorithms, and each

algorithm is evaluated and compared to find the best performing algorithm for
the final prediction. From Figure 10, it is clear that Random Forest algorithm
has the highest accuracy and it performs better than other algorithms.
The accuracy of all the models can be seen in the accuracy table in
Table 2.
 Crop Recommendation and Yield Estimation Using Machine Learning 877

Table 2 Accuracy of ML models for crop recommendation

Machine Learning Models Accuracy
Random Forest 99.93
KNN 95.15
Logistic Regression 91.08
Decision Tree 93.24
Naive Bayes 99.75
XG Boost 99.81
Gradient Boosting 99.61
SVM 90.13

Figure 11 The actual versus predicted yield of each crop.

In Figure 11, there is a close relationship between the actual and predicted
data.
Crop recommendation and yield estimation features are included on a
web page that is easy to access and use for farmers. Pickle is used to
incorporating the algorithms chosen in the Flask application to develop the
Web application based on selecting algorithms that perform better in crop
suggestion and yield estimation.
The developed online application is divided into four web pages: home
page, crop recommendation, yield estimation, and about us page.
Crop Recommendation – The user proposes the most appropriate crop and
the projected return, as area selection, season entry, area values, temperature,
precipitation, ph, nitrogen and potash, is also advised. The production output-
input parameters are required. After the user enters suitable inputs, the
predicted crop and estimated yield is displayed as shown in Figure 12.
878 A. Ashwitha and C. A. Latha

Figure 12 Crop recommendation section of the web application.

Yield Estimation – Based on the input criteria District, season, rainfall,

temperature, ph, crop, nitrogen, phosphorus, and potash, this component of
the website determines the estimated yield of a specific crop. After the user
enters suitable inputs, the estimated yield is displayed as shown in Figure 13.

Figure 13 Yield Estimation section of the web application.

 Crop Recommendation and Yield Estimation Using Machine Learning 879

Conclusion and Future Work

Farmers’ practises are among the most critical application areas in most
developing countries, such as India. In agriculture, the use of analytical
qualities of various technologies can affect the selection of multiple parame-
ters, allowing farmers to build better and increased returns. Farmers are now
unaware of the use of technology and the study of agricultural requirements,
which may lead them to select fake plants for cultivation, affecting yields and
thus income directly. A farmer-friendly system with a graphical user interface
that recommends the most suitable and profitable crop for a particular site and
is estimated in return is critical for preventing or minimising such loss. This
proposed approach assists farmers in deciding on the best crop for farming,
allowing the agriculture sector in India to progress through creative ideas
and the use of technology. Past agricultural data is used for crop production
research using machine learning algorithms, and it also presents a new
approach to examining and analysing the varied cultures in the Karnataka
region and their crop productivity. It is clear that the results of this method that
different crops yield differently depending on precipitation, soil, temperature,
season, and other factors. This technique allows farmers to recommend the
best crop for production in the upcoming agricultural season and bridges the
gap between agriculture and technology.
Crop analysis can be performed on a variety of other crops and in a variety
of locations. Future IoT sensors and applications can be used to collect real-
time agricultural farming data, and this data can also be integrated to prevent
the use of harvest prices and crop fertiliser requirements with existing market
data. Clustering techniques can also be combined to improve outcomes.
Every land data relevant to this can be obtained by providing GPS sites for
the specific country that can be accessed via the government’s Rain Prediction
System.

References
[1] Veenadhari, S., Bharat Misra, and C. D. Singh. “Machine learning
approach for forecasting crop yield based on climatic parameters.” Inter-
national Conference on Computer Communication and Informatics.
IEEE, 2014.
[2] Medar, Ramesh, Vijay S. Rajpurohit, and Shweta Shweta. “Crop yield
prediction using machine learning techniques.” IEEE 5th International
Conference for Convergence in Technology (I2CT). IEEE, 2019.
880 A. Ashwitha and C. A. Latha

[3] Kumar, Y. Jeevan Nagendra, et al. “Supervised Machine learning

Approach for Crop Yield Prediction in Agriculture Sector.” 5th Interna-
tional Conference on Communication and Electronics Systems (ICCES).
IEEE, 2020.
[4] Bhosale, Shreya V., et al. “Crop yield prediction using data analytics
and hybrid approach.” Fourth International Conference on Computing
Communication Control and Automation (ICCUBEA). IEEE, 2018.
[5] CH. Vishnu Vardhanchowdary, Dr. K. Venkataramana, “Tomato Crop
Yield Prediction using ID3.” International Jounal of Engineering
Reasearch and Technology (IJERT) (Volume 4 Issue 10 pp. 663–62),
2018.
[6] Girish L, Gangadhar S, Bharath T R, Balaji K S, Abhishek K T.
“Crop Yield and Rainfall Prediction in Tumakuru District using Machine
Learning.” National Conference on Technology for Rural Development
(NCTFRD-18), 2018.
[7] Nigam, A., Garg, S., Agrawal, A., & Agrawal, P. “Crop yield prediction
using machine learning algorithms.” Fifth International Conference on
Image Information Processing (ICIIP) (pp. 125–130). IEEE, 2019.
[8] Akshatha, Shailesh Shetty S, Anet P James, Athira M Saseendran,
Chaitra M Poojary, “Crop Analysis and Profit Prediction using Data
Mining Techniques” (Id:39), International Jounal of Engineering
Reasearch and Technology (IJERT) RTESIT – 2019 (VOLUME 7 –
ISSUE 08).
[9] Kamatchi, S. B., and Parvathi, R. (2019). Improvement of Crop Pro-
duction Using Recommender System by Weather Forecasts. Procedia
Computer Science, 165, 724–732.
[10] Nishiba Kabeer, Dr Loganathan. D and Cowsalya. T. “Prediction of Crop
Yield Using Data Mining.” International Journal of Computer Science
and Network (IJCSN) (2019).
[11] Talasila, Vamsidhar, Chitturi Prasad, Guttikonda Trinesh Sagar Reddy,
and Allada Aparna. “Analysis and Prediction of Crop Production in
Andhra Region Using Deep Convolutional Regression Network.” Inter-
national Journal of Intelligent Engineering and Systems (INASS), 2020.
[12] Palanivel, K., and Surianarayanan, C. “An approach for the prediction
of crop yield using machine learning and big data techniques.” Inter-
national Journal of Computer Engineering and Technology, 10(3), 110-
118.
[13] Pandey A, Purohit S, Jadhav S, Shah K. “Optimum Crop Prediction
using Data Mining and Machine Learning techniques.” International
 Crop Recommendation and Yield Estimation Using Machine Learning 881

Journal for Research in Applied Science & Engineering Technology

(IJRASET), 2014.
[14] Rajeswari, S. R., et al. “Smart Farming Prediction Using Machine
Learning.” International Journal of Innovative Technology and Explor-
ing Engineering (IJITEE) 7 (2018).
[15] Kusum Lata, and Bhushan Chaudhari. “Crop yield prediction using
data mining techniques and machine learning models for decision sup-
port systems.” International Journal of Emerging Technologies and
Innovative Research (IJETIR) (2019).
[16] Devika, B., and B. Ananthi. “Analysis of crop yield prediction using
data mining technique to predict the annual yield of major crops.”
International Research Journal of Engineering and Technology 5.12
(2018): 1460–1465.
[17] Mohan, P., and Patil, K. K. “Deep Learning-Based Weighted SOM to
Forecast Weather and Crop Prediction for Agriculture Application.”
International Journal of Intelligent Engineering and Systems, 11(4),
167–176 (2018).
[18] Surya, P., and I. Laurence Aroquiaraj. “Crop Yield Prediction in Agri-
culture using Data Mining Predictive Analytic Techniques.” IJRAR-
International Journal of Research and Analytical Reviews (IJRAR) 5.4
(2018): 783–787.
[19] 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.
[20] Sangeetha, C., and Sathyamoorthi, V. “Decision Support System for
Agricultural Crop Prediction Using Machine Learning Techniques.” In
Proceedings of the International Conference on Intelligent Computing
Systems (ICICS 2017–Dec 15th- 16th 2017) organised by Sona College
of Technology, Salem, Tamil Nadu, India.
[21] Kumar, R., Singh, M. P., Kumar, P., and Singh, J. P. “Crop Selec-
tion Method to maximise crop yield rate using machine learning
technique.” In International Conference on Smart Technologies and
Management for Computing, Communication, Controls, Energy and
Materials (ICSTM) (pp. 138–145). IEEE, (2015, May).
[22] Manoj G S, Prajwal G S, Ashoka U R, Prashant Krishna, Anitha P, 2020,
“Prediction and Analysis of Crop Yield Using Machine Learning Tech-
niques”, International Journal of Engineering Research and Technology
(IJERT) NCAIT – 2020 (Volume 8 – Issue 15)
882 A. Ashwitha and C. A. Latha

[23] Rani, D. E., Satyanarayana, N., Vardhan, B. V., and Goud, O. S. C.

“Crop Yield Analysis using Combinatorial Multivariate linear Regres-
sion.” International Journal of Advanced Research in Engineering and
Technology (IJARET) (2020).
[24] Kevin Tom Thomas, Varsha S, Merin Mary Saji, Lisha Varghese, and Er.
Jinu Thomas. “Crop Prediction Using Machine Learning.” International
Journal of Future Generation Communication and Networking (2020).
[25] Ms. Fathima, Ms Sowmya K, Ms Sunita Barker, Dr Sanjeev Kulkarni.
“Analysis of crop yield prediction using data mining technique.” Inter-
national Research Journal of Engineering and Technology (IRJET) –
2020.
[26] Pavan Patil, Virendra Panpatil, Prof. Shrikant Kokate. “Crop Prediction
Using Machine Learning.” International Research Journal of Engineer-
ing and Technology (IRJET) (2020).

Biographies

A. Ashwitha pursed Bachelor of Engineering from Visvesvaraya Technolog-

ical University, Belgaum in 2011 and Master of Engineering from Bangalore
University in the year 2013. She is currently pursuing a Ph.D. from Visves-
varaya Technological University under the guidance of Dr. C. A. Latha
(Prof and Head of CSE, AMCEC, Bangalore) and working as an Assistant
Professor in Department of Information Science, M S Ramaiah Institute of
Technology. She had published 3 research papers in reputed international
journals. Her main research work focuses on IoT Data Analytics, Data
Mining and Computational Intelligence based education. She has 8 years of
teaching experience and 4 years of Research Experience.
 Crop Recommendation and Yield Estimation Using Machine Learning 883

C. A. Latha pursed a Bachelor of Engineering from Mysore University in

1991 and a Master of Technology from NITK, Surathkal in the year 2003.
She has pursued Ph.D. from Anna University, Chennai in the year 2012 and
currently working as Professor and Head, Department of Computer Science,
AMC College of Engineering, Bangalore since 2012. She is a member of
Computer Society India since 1992 and a life member of the Indian Society
for technical education since 1993. She has published more than 5 research
papers in reputed international journals and conferences and authored a book,
“programming in C and Introduction to Data structures”. She has reviewed
several research papers for Elsevier and many international conferences and
secured Best reviewer award in the year 2015 for her outstanding contribution
in reviewing by Elsevier. Her main research work focuses on Cryptography
Algorithms, Network Security.