© 2022 IJRTI | Volume 7, Issue 11 | ISSN: 2456-3315

Agricultural Data Analysis using Machine Learning

Algorithms
Appa Rao Bobbili1, Dr. Sreedevi M2
1
Student, 2Professor
Amrita Sai Institute of Science and Technology
Autonomous NAAC with A Grade, Andhra Pradesh, India,

Abstract: Agriculture is undoubtedly the largest livelihood provider in India and also contributes a significant figure to the
economy of our Country. The technological factors affecting the crop production includes practices used and also
managerial decisions. So, predicting the crop yield prior to its harvest would help farmers to take appropriate steps. We
attempt to resolve theissue by building a user-friendly prediction system. The results of the prediction are suggested to the
farmer such that suitable changes can be made in order to improve the produce. There are different techniques or
algorithms which help to predict crop yield. By analyzing all the parameters like location, soil nutrients, pH value, rainfall,
moisture a potential solution can be obtained to overcome the situation faced by farmers. This paper focuses on the analysis
of the agriculture data and finding optimal yield to provide an insight before the actual crop production using data mining
techniques and Machine Learning algorithms.

Keywords: Yield, Random Forest regress or, Decision Tree regress or, GDP, Digitalisation.

I. INTRODUCTION
Today, India is one of the leading producers across the world in the agriculture sector[1]. Agriculture is the broadest economic
sector and plays an outstanding role in the socio-economic part of India. Agriculture is an eccentric business crop production which
is influenced by many climate and economic factors. Andhra Pradesh, basically being an agro-Based economy contributes more
than 29% of the GDP as against 17% in the country's GDP. Periodical advice to the farmers either in terms of improved agricultural
strategies or advancements in factors affecting the production of crops may strengthen the state in the agriculture sector. Yield
prediction is one among the agricultural advancements. Due to these kinds of innovations agriculture is driving the interest of
modern man. In the past farmers used to predict their yield from previous experiences[2]. Digitalisation in farming gives awareness
about the cultivation of the crops at the right time and at the right place even to young farmers. These kinds of advancements need
the use of data analytics. This is one such system that can be used to address yield prediction. The main objectives are:
1) To analyse different parameters (soil nutrients, rainfall, area etc)
2) To use machine learning techniques to predict crop yield.
3) To provide an easy to use User Interface

II. HOW DATA MINING IS USED IN AGRICULTURE SECTOR

Data mining techniques are used in performing several activities in the agricultural sector such as pest identification, detection and
classification and prediction of crop diseases. It can also be used in yield prediction, input management (planning of irrigation and
pesticides), fertilizer suggestion and predicting soil. In a world full of data , data mining is the computational process for discovering
new patterns[3]. Data mining techniques provide a major advantage in agriculture for detection and prediction for optimizing the
pesticides. Techniques for agriculture related activities provide a lot of information. The yield of agriculture primarily depends on
diseases, pests, weather conditions, planning of various crops for the harvest productivity are the results.
Crop production for reliable and timely requirements for various decisions for agriculture marketing. Predictions are very useful
for agriculture data. For instance, by applying data mining techniques, the government can fully benefit from data about farmers'
buying patterns and also to achieve a superior understanding of their land to achieve more profit on the farmer's part.
Data mining techniques followed in two ways[4]:
1) Descriptive data mining.
2) Predictive data mining.
Descriptive data mining tasks characterize the final properties of the info within the database while predictive data mining is
employed to predict the direct values supported patterns determined from known results. Prediction involves using some variables
or fields within the database to predict unknown or future values of other variables of interest.
As far as data mining techniques are concerned, in most cases predictive data mining approaches are employed. Predictive data
mining techniques are employed to predict future crop, forecasting, pesticides and fertilizers to be used, revenue to be generated
and so on. These techniques are used for pre-harvest forecasting for the agriculture field and are able to provide a lot of data on
agricultural-related activities. Data of agriculture in data mining can be presented in the form of datasets.

III. PROPOSED SYSTEM

The main objectives of proposed work is to analyse the agricultural parameters using data mining algorithms and predict the yield.In
our proposed work, agriculture data has been collected from various sources which include:
Dataset in agricultural sector[5] ,Crop wise agriculture data:[6], Soil data of different districts:[7]
In this proposed system , we mainly focussed on Andhra Pradesh State in India. As the state has two major rivers flowing , it has a

IJRTI2211033 International Journal for Research Trends and Innovation (www.ijrti.org) 207
 © 2022 IJRTI | Volume 7, Issue 11 | ISSN: 2456-3315

diversity in factors useful for agriculture at district level. Periodical data about the crop , soil and water a particular region is the
major focus of this study.The final dataset has been tabulated as in table-1:

S. No Feature Description

1 Year The year in which the crop will be cultivated. Generally, the
upcoming year
2 Season One among Kharif,Rabi and Whole Year.
3 Crop Name of the crop
4 District Name of the district
5 pH Level This describes the nature of the soil
6 Nitrogen Amount of nitrogen present
7 Potassium Amount of potassium present
8 Phosphorus Amount of phosphorus present
9 Rainfall Expected rainfall in millimeters
10 Area Area of field in hectares

Table-1: Description of Input data

The below diagram depicts the system architecture of our proposed system. Our whole system can be divided into 2 modules as a
whole i.e., one model predicts the optimal yield and the other model analyses the patterns in the dataset. The operation of these
models as a whole is specified clearly in the below diagram.

Fig.1 The blueprint of the proposed system

IV. METHODS
In the implementation of this yield prediction system Regression Analysis is used.Regression Analysis is considered as one of the
oldest,and widely used multivariate analysis techniques in the social sciences. Unlike others regression stands as an example of
dependence analysis in which the variables are treated asymmetrically. In regression analysis, the object is to obtain a prediction
of one variable, based on given the values of the others[8].Random Forest and Decision Tree algorithms are generally used in
classification problems but these can also be used in regression problems as well.

A. Decision Tree Regression

The Decision Tree algorithm comes under supervised machine learning techniques. A decision tree arrives at an outcome by asking
a series of questions to the input data , each question narrows down the possible outcomes until the model gets enough potential to
make a unique prediction[9]. The order of the questions as well as their contents are being determined by the model.All the questions
that are raised have their answer as either true or false.

B. Random Forest Regression

Random Forest algorithm comes under the family of ensemble algorithms. This is also a supervised learning algorithm. This can
be implemented in classification and regression as well. Random forest algorithm basically works on Decision Tree principle by
constructing a number of decision trees having different sets of hyper-parameters for tuning and training on different subsets of
data[10].

V. EXPERIMENTAL RESULT
A. Decision Tree Regression
Decision Tree algorithm on applying on the dataset resulted 100% on data and 82%(approx.) on test data.Fig-2 shows the accuracy
of Decision tree algorithm on data:

IJRTI2211033 International Journal for Research Trends and Innovation (www.ijrti.org) 208
 © 2022 IJRTI | Volume 7, Issue 11 | ISSN: 2456-3315

Fig-2: Result of Decision Tree algorithm

B. Random Forest Regression

Random Forest algorithm on applying on the dataset resulted 98% on data and 90%(approx.) on test data.Fig-3 shows the accuracy
of Decision tree algorithm on data:

Fig-3: Result of Random Forest algorithm

Fig-4:Crop with reduced production and increased price

Fig-5: Crops with lower increase in production but are increasing in price

IJRTI2211033 International Journal for Research Trends and Innovation (www.ijrti.org) 209
 © 2022 IJRTI | Volume 7, Issue 11 | ISSN: 2456-3315

Fig-6: Decision Tree for Suicide Prediction

S.No. File name Name of DB in program Details or Comments

1 apy Crop Production Main DB that has crop production info from 2000 to
Statistics(crop prod) 2014.For different district of each State it includes,
what are various crops produced, their area of
production and total produc- tion,and what type of
crop is it.(Kharif,Rabi)
2 Crops price Crop prices(crop price) Prices of some crops year wise change till 2013.For
different commodities it has data for its price in
rupees per quintal.
3 area cult Crop cultivation area(crop Area of land a crop is produced on year by year for
cult) major crops from 2000 to 2009.
4 culti cost Crop cultivation cost(culti State wise cost of cultivation of crops per hectare
cost) and per quintal.Three variant of cost are
there(actual paid out cost plus imputed value of
family labour(A2+FL),comprehensive cost including
imputed rent and interest on owned land and
capital(C2) and cost per quintal)
5 Mean Mean Data of mean temperature from 2000-2012 for whole
Temperatures Temperatures(temperature) year and over interval of two months. This is used to
determine effect of temperature on various crops
6 rainfall Rainfall Statistics(rainfall) State wise rainfall statics from year 2000-2015
cleaned annually and monthly in millimeter per square
meter(area)
7 Avg annual Crops Growth rate(growth) Growth rate of various crops from 1997 to 2012 over
Growth Rate a interval of five years.Growth rate represent increase
Major Crops in size, mass or number of crops over a period of time.
It is used in anal- ysis of preference of one crop over
other.
8 suicides 10 Suicide Statics(suicides) Data of no. of total cases of suicides in vari- ous states
14 from year 2010-2014.While analysis it is taken into
account to predict responsible factors.
9 IndiaExport Exports(exports) Data regarding the amount of export of vari- ous
materials and its price from 2003 to 2015.
File name Name of DB in program Details or Comments
10 data set Data(data) Combined data of various states from 2000 to 2014.
This is combined representation of all data in one
table. All other tables mentioned above are combine
using the common features and merged.

VI. CONCLUSION
Both Decision tree regression and Random Forest regression techniques are implemented on the input data to assess the best
performance yielding method. These methods are compared using performance metrics. According to the analyses of metrics both
the algorithms work well , but Random Forest regression gives a better accuracy score on test data than Decision tree regression.
The proposed work can also be extended to analyse the climatic conditions and other factors for the crop and to increase the crop
production.

IJRTI2211033 International Journal for Research Trends and Innovation (www.ijrti.org) 210
 © 2022 IJRTI | Volume 7, Issue 11 | ISSN: 2456-3315

VII. REFERENCES
[1] https://www.investopedia.com/articles/investing/100615/4-countries-produce-most-food.asp
[2] http://www.fao.org/fileadmin/templates/ess/documents/meetings_and_workshops/GS_SAC_2013/Improving_methods_for_crops_
estimates/Crop_Yield_Forec asting_Methods_and_Early_Warning_Systems_Lit_review.pdf
[3] https://ijaers.com/uploads/issue_files/3%20IJAERS-MAY-2017-60-
Different%20Types%20of%20Data%20Mining%20Techniques.pdf
[4] https://docs.oracle.com/cd/B12037_01/datamine.101/b10698/4descrip.htm
[5] https://data.gov.in/
[6] http://www.apagrisnet.gov.in/2018/weekly/October/weekly_report_(Rabi)_06_21-11-18.pdf
[7] http://dataverse.icrisat.org/file.xhtml?fileId=1185&version=RELEASED version=.3
[8] https://www.sciencedirect.com/topics/medicine-and-dentistry/regression-analysis
[9] https://gdcoder.com/decision-tree-regressor-explained-in-depth/
[10] https://www.analyticsvidhya.com/blog/2020/12/lets-open-the-black-box-of-random-forests/
[11] https://journalofbigdata.springeropen.com/articles/10.1186/s40537-017-0077-4

IJRTI2211033 International Journal for Research Trends and Innovation (www.ijrti.org) 211