The International journal of analytical and experimental modal analysis ISSN NO: 0886-9367

IMPROVING CROP PRODUCTIVITY THROUGH CROP RECOMMENDATION

SYSTEM

PERKA SAMPATH*, DR B NARSIMHA**

PG SCHOLAR*, PROFESSOR & HEAD**

DEPARTMENT OF CSE, HOLY MARY INSTITUTE OF TECHNOLOGY & SCIENCE, HYDERABAD, TELANGANA-501 301

ABSTRACT sample from a field to know about features

of that soil like pH values, organic Carbon,
Crop recommendation system or prediction
iron, zinc etc here we using 10 parameters
system is the art of predicting crop yields to
for improving a productivity of the crops to
improve the production and production
get a efficient result analysis. The purpose of
before the harvest actually takes place, it
soil testing in high-yield farming is to
takes typically a couple of months in
determine that relative ability of a soil to
advance. Crop prediction depends on the
supply crop nutrients during a particular
computer programs that describe the plant-
growing season, to determine the needs, and
environment and the soil features
for diagnosing problems such as the
interactions in quantitative terms. The soil
excessive salinity or alkalinity. Soil testing
testing will start with the collections of a soil
is used to guide the nutrient management
sample from the field. The first basic
decisions related to manure and sludge
principles of the soil testing is that a field
applications of the soil with the objective of
can be sampled in such a way that by getting
maximizing economic benefits while
a chemical analysis of the soil sample and
minimizing the potential for negative
also majorly depend on temperature and
impacts on water quality and also features.
rainfall will accurately reflect the field’s true
The nutrient requirements of soil using
nutrient status on a particular area to help
agriculture for crop production are examined
out farmers to improve the production
to determine the need data mining system to
KEYWORDS: Recommender System, detect the crops suits for the soil by
Machine Learning and Data Science,
Knowledge Discovery in Databases, Naive analyzing features of that particular soil of
Bayes that region. This helps in giving information
INTRODUCTION The soil testing to the farmer about the suitable crop for
program starts with the collection of a soil particular regions.

Volume XIV, Issue III, March/2022 Page No : 1797

The International journal of analytical and experimental modal analysis ISSN NO: 0886-9367

EXISTING SYSTEM  Pritam Bose [6] developed a SNN

 Many crop prediction yield models model to have a spatiotemporal
have been developed. Clustering analysis w ith crop estimation. An
approaches such as k-means, automatic system to gather the
kmeans++ are used to perform information about soil nature,
grouping o f data as clusters to weather conditions was developed
predict crop yield is used [1]. [7] with clustering techniques to
Tripathy et al., [2] provided a system extract the knowledge and use it by
to have management of pesticides for farmers in crop cultivation.
crop cultivation using data mining DISADVANTAGES
process.  An existing system’s
 Essential parameter for agriculture recommendation is based on soil and
analysis is nature o f soil. Diverse not based on Cr o p
varieties o f soil are available in this Recommendation Based on
India. Crops are cultivated Production.
depending on the type of soil in the  Farmers will be given
land. The role of soil in improving recommendation by considering not
crop cultivation is discussed [3]. the season of crop production.
Data mining techniques are applied
to analyze the soil parameter PROPOSED SYSTEM
 JRip, J48 and Naive Bayes Prediction o f crops was done according to
techniques are applied [4] w hich farmer’s experience in the past years.
produces more reliable results in Although farmer’s knowledge sustains,
analyzing red and Black soil. Impact agricultural factors has been changed to
of parameters o f agriculture in crop astonishing level. There comes a need to
management is studied to improve indulge engineering effect in crop
productivity [5]. Neural networks, prediction. Data mining plays a novel role in
soft computing, big data and fuzzy agriculture research [11]. This field uses
logic methods are being used to historical data to predict; such techniques
examine the agricultural factors. are neural networks, K-nearest Neighbor. K-
means algorithm does not use historical data

Volume XIV, Issue III, March/2022 Page No : 1798

The International journal of analytical and experimental modal analysis ISSN NO: 0886-9367

but predicts based on-computing centers o f  An exact recommendation based on

the samples and forming clusters. Recommend based on production
Computational cost of algorithm acts as a SYSTEM IMPLEMENTATION:
major issue. Use of Artificial Neural
PROBLEM ANALYSIS:
Network is a boon to agriculture field which
computes accurately even with more input. As fertilizer is composed of chemicals and

An architecture developed in [11] uses harmful for the plants and society, this

input; selects needed features; classification procedure has to be changed to find out the

and association rule mining is applied and suitable crop for the soil. The nutrient

visualized. requirements of soils using in agriculture for

Bangladesh has its high production as rice. crop production are examined to determine

Statistical Methodologies has been used to the need of a data mining system to detect

predict its crop production. Shakil Ahamed the crops suite for the soil by analyzing the

[12] applied clustering and classification features of the soil. This helps in giving

techniques on 15 districts of Bangladesh to information to the farmer about suitable

recommend for yield and planting of crops. crops that regulate the nutrition levels of the

Factors implementing crop yield were soil. Farmer can know which crop is suitable

considered. They are for that area and also he will get 100% yield

a. Environmental factors-rainfall, on that year. System used by administrator,

humidity, Minimum and maximum staffs and farmers. System is a browser

temperature based application which predicts crops

b. Biotic factors-soil pH and salinity based on the soil test results. System makes

c. Area factors-irrigated or use of data mining technique for crop

cultivated prediction. System makes use of previous

ADVANTAGES dataset for the crop prediction System makes

 Recommendation system acts as a use of navie bayesian algorithm for crop

good engine to provide suitable prediction System generates accurate results

items to users considering other based on the size of the dataset.

factors. This is extended to have its

support to agricultural sector

Volume XIV, Issue III, March/2022 Page No : 1799

The International journal of analytical and experimental modal analysis ISSN NO: 0886-9367

becomes the root of the decision tree and for

each outcome Oi we build a subsidiary
decision tree by invoking the same
procedure recursively on the set Si.
GRADIENT BOOSTING

Gradient boosting is a machine

learning technique used
in regression and classification tasks, among
others. It gives a prediction model in the
form of an ensemble of weak prediction
models, which are typically decision
DECISION TREE CLASSIFIERS trees.[1][2] When a decision tree is the weak
learner, the resulting algorithm is called
Decision tree classifiers are used
gradient-boosted trees; it usually
successfully in many diverse areas. Their
outperforms random forest.A gradient-
most important feature is the capability of
boosted trees model is built in a stage-wise
capturing descriptive decision making
fashion as in other boosting methods, but it
knowledge from the supplied data. Decision
generalizes the other methods by allowing
tree can be generated from training sets. The
optimization of an
procedure for such generation based on the
arbitrary differentiable loss function.
set of objects (S), each belonging to one of
the classes C1, C2, …, Ck is as follows: K-NEAREST NEIGHBORS (KNN)
Step 1. If all the objects in S belong to the
 Simple, but a very powerful
same class, for example Ci, the decision tree classification algorithm
for S consists of a leaf labeled with this
 Classifies based on a similarity
class measure
 Non-parametric
Step 2. Otherwise, let T be some test with
 Lazy learning
possible outcomes O1, O2,…, On. Each  Does not “learn” until the test
example is given
object in S has one outcome for T so the test
partitions S into subsets S1, S2,… Sn where  Whenever we have a new data to
classify, we find its K-nearest
each object in Si has outcome Oi for T. T neighbors from the training data

Volume XIV, Issue III, March/2022 Page No : 1800

The International journal of analytical and experimental modal analysis ISSN NO: 0886-9367

Many statisticians feel that logistic

Example
regression is more versatile and better suited
 Training dataset consists of k-closest for modeling most situations than is
examples in feature space discriminant analysis. This is because
 Feature space means, space with logistic regression does not assume that the
categorization variables (non-metric independent variables are normally
variables) distributed, as discriminant analysis does.
 Learning based on instances, and This program computes binary logistic
thus also works lazily because regression and multinomial logistic
instance close to the input vector for regression on both numeric and categorical
test or prediction may take time to independent variables. It reports on the
occur in the training dataset regression equation as well as the goodness
LOGISTIC REGRESSION of fit, odds ratios, confidence limits,
CLASSIFIERS
likelihood, and deviance. It performs a
Logistic regression analysis studies the comprehensive residual analysis including
association between a categorical dependent diagnostic residual reports and plots. It can
variable and a set of independent perform an independent variable subset
(explanatory) variables. The name logistic selection search, looking for the best
regression is used when the dependent regression model with the fewest
variable has only two values, such as 0 and 1 independent variables. It provides
or Yes and No. The name multinomial confidence intervals on predicted values and
logistic regression is usually reserved for the provides ROC curves to help determine the
case when the dependent variable has three best cutoff point for classification. It allows
or more unique values, such as Married, you to validate your results by automatically
Single, Divorced, or Widowed. Although the classifying rows that are not used during the
type of data used for the dependent variable analysis.
is different from that of multiple regression, NAÏVE BAYES
the practical use of the procedure is similar. The naive bayes approach is a supervised
Logistic regression competes with learning method which is based on a
discriminant analysis as a method for simplistic hypothesis: it assumes that the
analyzing categorical-response variables. presence (or absence) of a particular feature

Volume XIV, Issue III, March/2022 Page No : 1801

The International journal of analytical and experimental modal analysis ISSN NO: 0886-9367

of a class is unrelated to the presence (or present some theoretical aspects of the naive
absence) of any other feature Yet, despite bayes classifier. Then, we implement the
this, it appears robust and efficient. Its approach on a dataset with Tanagra. We
performance is comparable to other compare the obtained results (the parameters
supervised learning techniques. Various of the model) to those obtained with other
reasons have been advanced in the literature. linear approaches such as the logistic
In this tutorial, wehighlight an explanation regression, the linear discriminant analysis
based on the representation bias. The naive and the linear SVM. We note that the results
bayes classifier is a linear classifier, as well are highly consistent. This largely explains
as linear discriminant analysis, logistic the good performance of the method in
regression or linear SVM (support vector comparison to others. In the second part, we
machine). The difference lies on the method use various tools on the same dataset (Weka
of estimating the parameters of the classifier 3.6.0, R 2.9.2, Knime 2.1.1, Orange 2.0b
(the learning bias). While the Naive Bayes and RapidMiner 4.6.0). We try above all to
classifier is widely used in the research understand the obtained results.
world, it is not widespread among RANDOM FOREST
practitioners which want to obtain usable Random forests or random decision forests
results. On the one hand, the researchers are an ensemble learning method for
found especially it is very easy to program classification, regression and other tasks that
and implement it, its parameters are easy to operates by constructing a multitude of
estimate, learning is very fast even on very decision trees at training time. For
large databases, its accuracy is reasonably classification tasks, the output of the random
good in comparison to the other approaches. forest is the class selected by most trees. For
On the other hand, the final users do not regression tasks, the mean or average
obtain a model easy to interpret and deploy, prediction of the individual trees is returned.
they does not understand the interest of such Random decision forests correct for decision
a technique. Thus, we introduce in a new trees' habit of overfitting to their training set.
presentation of the results of the learning Random forests generally outperform
process. The classifier is easier to decision trees, but their accuracy is lower
understand, and its deployment is also made than gradient boosted trees. However, data
easier. In the first part of this tutorial, we characteristics can affect their performance.

Volume XIV, Issue III, March/2022 Page No : 1802

The International journal of analytical and experimental modal analysis ISSN NO: 0886-9367

The first algorithm for random decision to grow in that field. Proposed system
forests was created in 1995 by Tin Kam predicts the crops using various data mining
Ho[1] using the random subspace method, techniques aspecially using a Naïve
which, in Ho's formulation, is a way to Bayesain algorithm to get accurate results.
implement the "stochastic discrimination" This system also useful to agricultural
approach to classification proposed by departments to predict the right crop in right
Eugene Kleinberg. time which gives the efficient results. If we
have such kind of an automation, then it will
An extension of the algorithm was
be useful to farmers and agricultural field.
developed by Leo Breiman and Adele
The goals that have been achieved by the
Cutler, who registered "Random Forests" as
developed system are, Simplified and reduce
a trademark in 2006 (as of 2019, owned by
the manual work of the agricultural
Minitab, Inc.).The extension combines
department, Large volumes of data can be
Breiman's "bagging" idea and random
stored and It provides Smooth work flow
selection of features, introduced first by
Ho[1] and later independently by Amit and REFERENCES
Geman[13] in order to construct a collection
[1] Lokesh.K,Shakti.J, Sneha Wilson,
of decision trees with controlled variance.
Tharini.M.S, “Automated crop prediction
Random forests are frequently used as based on efficient soil nutrient estimation
"blackbox" models in businesses, as they using sensor network”, July 2016,National
generate reasonable predictions across a Conference on Product Design (NCPD
wide range of data while requiring little 2016)
configuration.
[2] Rakesh Kumar, M.P. Singh, Prabhat
CONCLUSION Kumar and J.P. Singh (2015), “Crop
Selection Method to Maximize Crop Yield
Nowadays farmers facing lots of problems
Rate using Machine Learning Technique”,
in the agricultural field due the crop
International Conference on Smart
production and they don’t know the proper
Technologies and Management for
information regarding how to improve crop
Computing, Communication, Controls,
production for what they invest and also to
Energy and Materials (ICSTM).
cultivate. This proposed system helps the
farmers to know about what is a right crop

Volume XIV, Issue III, March/2022 Page No : 1803

The International journal of analytical and experimental modal analysis ISSN NO: 0886-9367

[3] S.Pudumalar, E.Ramanujam, ”Crop

Recommendation System for Precision
Agriculture”,2016, IEEE Eighth
International Conference on Advanced
Computing (ICoAC)

[4] Aakunuri Manjula, Dr.G .Narsimha

(2015), “XCYPF: A Flexible and Extensible
Framework for Agricultural Crop Yield
Prediction”, Conference on Intelligent
Systems and Control (ISCO)

Volume XIV, Issue III, March/2022 Page No : 1804