International Journal of Engineering and Technical Research (IJETR)

ISSN: 2321-0869 (O) 2454-4698 (P) Volume-7, Issue-6, June 2017

An Effective Algorithm for Correlation Attribute

Subset Selection by Using Genetic Algorithm Based
On Naive Bays Classifier
Mr. Shiv Kumar Sharma, Mr. Ashwani Kumar, Mr. Ram Kumar Sharma

Abstract In recent years, application of feature selection II. PROBLEM IDENTIFICATION

methods in various datasets has greatly increased. Feature
Feature subset selection can be used as the technique to
selection is an important topic in data mining, especially for
high dimensional datasets. Feature selection (also known as identifying and removing as many redundant and irrelevant
subset selection) is a process commonly used in machine features as possible. This is because:
learning, wherein subsets of the features available from the data (i) redundant features do not help in getting a better predictor
are selected for application of a learning algorithm. The main for that they provide mostly information which is already
idea of feature selection is to choose a subset of input variables present in other feature(s), and
by eliminating features with little or no predictive information. (ii) Irrelevant features do not contribute to the predictive
The challenging task in feature selection is how to obtain an accuracy. A number of approaches to feature subset selection
optimal subset of relevant and non redundant features which
have been proposed in the literature, a few of them only are
will give an optimal solution without increasing the complexity
of the modeling task. Feature selection that selects a subset of
referred here.
most salient features and removes irrelevant, redundant and (iii) Relevant: These are features which have an influence on
noisy features is a process commonly employed in machine the output and their role cannot be assumed by the rest.
learning to solve the high dimensionality problem. It focuses But in my research paper we find the many assumption of
learning algorithms on most useful aspects of data, thereby Co-related feature selection problem in data mining.
making learning task faster and more accurate. A data Feature Selection is the essential step in data
warehouse is designed to consolidate and maintain all features mining. Individual Evaluation and Subset Evaluation are two
that are relevant for the analysis processes. major techniques in feature selection. Individual Evaluation
means assigning weight to an individual feature. Subset
Index Terms co-relation based GA, Feature Selection,
Evaluation is construction of feature subset. The general
Feature Selection Methods, Feature Selection Algorithms,
GABAS. criteria for feature selection methods are the classification
accuracy and the class distribution. The classification
accuracy does not significantly decrease and the resulting
I. INTRODUCTION class distribution, given only the values for selected
features. Feature Selection can support many applications,
In recent years, the need to apply feature selection methods it include the problems involving high dimensional data.
in medical datasets has greatly increased. This is because Figure 1 describes feature selection steps. The four key
most medical datasets have large number of samples of steps in feature selection are-
high dimensional features. a. Subset generation
A "feature" or "attribute" or "variable" refers to an aspect of b. Subset Evaluation
the data. Usually before collecting data, features are c. Stopping criteria
specified or chosen. Features can be discrete, continuous, or d. Result validation
nominal. Generally, features are characterized as: The feature selection is used to select relevant features by
i. Relevant: These are features which have an influence removing irrelevant and redundant features to improve the
on the output and their role cannot be assumed by performance and to speed up the learning
the rest. process.
ii. Irrelevant: Irrelevant features are defined as those
features not having any influence on the output,
and whose values are generated at random for each
example.
iii. Redundant: A redundancy exists whenever a feature
can take the role of another (perhaps the simplest
way to model redundancy).

Shiv Kumar Sharma, M.TECH (CSE) Research Scholar, IEC-CET,

Greater Noida
Ashwani Kumar, Assistant Professor, Department of Information
Technology IEC-CET, Greater Noida Figure 1. The general procedure for feature selection
R. K. Sharma, Assistant Professor, Department of Information
Technology, NIET Greater Noida

1 www.erpublication.org
An Effective Algorithm for Correlation Attribute Subset Selection by Using Genetic Algorithm Based On Naive Bays
Classifier
III. PROPOSED WORK AND METHODOLOGY
Works reported so far in the area of feature subset
selection for dimensionality reduction could not claimed that
the solution provided by them in the most optimal solution its
because correlation attribute feature subset selection is an
optimization problem so the scope of work remains open
further and algorithm likes ACO, GA, co-relation based GA,
Meta heuristic Search and PSO have been applied to subset
selection in the past. In my research paper we are working on,
correlation attribute subset selection done by using genetic
algorithm that based on naive bays classifier. Its aim is to
improve the performance results of classifiers but using a
significantly reduced set of features. Genetic Algorithms as an
optimization tool is proposed to be applied where Nave
Bayes Classifier will be used to compute the Classification
accuracy that will be taken as the fitness value of the Figure 3 List of parameters
individual subset.
After the k fold validation results, a subset of selected
IV. RESULT ANALYSIS attribute is shown in last of output screen. Here k is taken by
In this proposed method a source bank dataset will be the user in term of number. Classification ratio is show the
taken as input in arff file; arff is Attribute relation file format. performance of the program.
After that all the attributes of datasets are encoded. A number
of attributes are select randomly. The classification accuracy
is compute with selected attributes.
a complete detail of implemented tool is discussed along with
the description of results obtained.
A tool is designed in Java to select the subset of
features automatically based on GABASS. Tool has a GUI as
shown in figure 6.1 where three command buttons are
provided named;
a. File,
b. Preprocess
c. Classify
By clicking on the file button an ARFF format dataset file is
browsed and taken as input. An ARFF (Attribute-Relation
File Format) file is an ASCII text file that describes a list of
instances sharing a set of attributes. One such a list of features Figure 4. Selected subset of Attribute
can be seen in figure 2. An attribute value and its quantities of
instances on clicking a particular attribute. In this work, five different methods are used for feature
selection. Forward Selection Multi cross Validation,
Bootstrap backward elimination, Relief, MIFS and proposed
GABASS method are implemented and five different feature
subsets were obtained. Forward Selection Multi cross
Validation and Bootstrap backward elimination are wrapper
based method; Relief and MIFS are filter based method. To
calculate the classification accuracy for above described
methods; SIPINA tool of TANAGRA software is used. The
selected feature subsets by these five methods are detailed in
following table. The k-fold cross validation method was used
to measure the performances.

V. CONCLUSION
In the Feature selection methodology is the first task
of any learning approach to define a relevant set of features.
Several methods are proposed to deal with the problem of
Figure 2. Attribute list feature selection including filter, wrapper and embedded
methods. In this work, I focus on feature subset selection to
By clicking on the Classify button a number of input boxes, select a minimally sized subset of optimal features.
some check boxes and two buttons are shown in figure 3. An Feature Selection is optimization problem; genetic
input boxes are used for take user define values in respect to algorithm based attribute subset selection using nave bayes
different parameters. All check boxes are optional and used to classifier is used for this purpose. GABASS are found to be
depend on the user. the best technique for selection purpose when there is very

2 www.erpublication.org
 International Journal of Engineering and Technical Research (IJETR)
ISSN: 2321-0869 (O) 2454-4698 (P) Volume-7, Issue-6, June 2017
large population. The GABASS provides good results and
their power lies in the good adaptation to the various and fast
changing environments.

VI. FUTURE WORK

Future work will involve experiments on the datasets
from different domains. The GABASS algorithm tested on
different domains previously. The difference in performance
and accuracy of different ensemble approaches will be
evaluated.

GABASS can give more efficient results and the

optimization process can become much easier and faster. The
one more important aspect of future work is of finding more
factors that can compare two test suits for their goodness, so
that efficiency of selection process can be improved.

REFERENCES
[1] I. Inza, P. Larranaga, R. Etxeberria, B. Sierra, Feature Subset
Selection by Bayesian network-based optimization Artificial
Intelligence 123, pp. 157184, 2000
[2] Isabelle Guyon, Andre Elisseeff, An Introduction to Variable and
Feature Selection, Journal of Machine Learning Research 3, pp.
1157-1182, 2003.
[3] Lei Yu, Huan Liu, Efficient Feature Selection via Analysis of
Relevance and Redundancy, Journal of Machine Learning Research
5 , pp. 12051224, 2004
[4] Felix Garca Lopez, Miguel Garca Torres, Belen Melian Batista, Jose
A. Moreno Perez , J. Marcos Moreno-Vega, Solving feature subset
selection problem by a Parallel Scatter Search European Journal of
Operational Research 169, pp. 477489, 2006.
[5] Dunja Mladeni, Feature Selection for Dimensionality Reduction
LNCS 3940, pp. 84102, 2006.
[6] C.-R. Jiang, C.-C. Liu, X. J. Zhou, and H. Huang, Optimal ranking in
multi-label classification using local precision rates, Statistica
Sinica, vol. 24, no. 4, pp. 15471570, 2014.
[7] M. S. Mohamad, S. Deris, S. M. Yatim, and M. R. Othman, Feature
selection method using genetic algorithm for the classification of
small and high dimension data, in Proceedings of the 1st
International Symposium on Information and Communication
Technology, pp. 14, 2004.

3 www.erpublication.org