N045038690

Neha Patil et al Int. Journal of Engineering Research and Applications www.ijera.com
ISSN : 2248-9622, Vol. 4, Issue 5( Version 3), May 2014, pp.86-90
www.ijera.com 86 | P a g e
Text Categorization Optimization By A Hybrid Approach Using
Multiple Feature Selection And Feature Extraction Methods
K. Rajeswari*, Sneha Nakil**, Neha Patil**, Sebatini Pereira**, Neha
Ramdasi**
*Assistant Professor, Department of Computer Engineering, Pimpri-Chinchwad College of Engineering, Pune-
411044
**Student-BE, Department of Computer Engineering, Pimpri-Chinchwad College of Engineering, Pune-411044
ABSTRACT
Text categorization is the task of automatically sorting a set of documents into categories from a pre-deﬁned set.
This means it assigns predefines categories to free-text documents. In this paper we are proposing a unique two
stage feature selection method for text categorization by using information gain, principle component analysis
and genetic algorithm. In the methodology, every term inside the document is ranked depending on their
importance for classification using the information gain (IG) methodology. In the second stage, genetic
algorithm (GA) and principal component analysis (PCA) feature selection and feature extraction methods are
applied individually to the terms which are ranked in decreasing order of importance, and a dimension reduction
is carried out. Therefore, throughout the text categorization terms of less importance are ignored, and feature
selection and extraction methods are applied to the terms of highest importance; so the computational time and
complexity of categorization is reduced. To analyze the dimension reduction in our proposed model,
experiments area unit conducted using the k-nearest neighbor(KNN) and C4.5 decision tree algorithmic rule on
selected data set.
Keywords - Information Gain, Principal Component Analysis, Genetic Algorithm, K-Nearest Neighbor
classifier, C4.5 tree classifier.
I. INTRODUCTION
Text categorization is one of the challenging
research topics due to the necessity to organize and
categorize growing number of electronic documents
worldwide. So far, text classification has been
successfully applied to various domains such as topic
detection [15], spam e-mail filtering [16], SMS spam
filtering [17], author identification[18], web page
classification [19]and sentiment analysis. A
conventional text categorization framework consists
of preprocessing, feature extraction, feature selection,
and classification stages. It solves the problem of
assigning text content to predefined categories. Text
categorization has vital importance in applications
used in the real world. For example, news stories are
typically organized by subject categories (topics) or
geographical codes; academic papers are often
classified by technical domains and sub-domains;
patient reports in health-care organizations are often
indexed from multiple aspects, using taxonomies of
disease categories, types of surgical procedures,
insurance reimbursement codes and so on.
II. RELATED WORK
Text categorization is defined as assigning
pre-defined categories to a given set of documents
based on v classification patterns. Although many
information retrieval applications such as filtering
and searching for relevant information can benefit
from text categorization research, a major problem of
text categorization is the high dimensionality of the
feature space due to a large number of terms. Some
feature extractions have been successfully used in
text categorization such as principal component
analysis, latent semantic indexing, clustering methods
etc. Among the many methods that are used for
feature extraction, PCA has attracted a lot of
attention. PCA is a statistical technique for reduction
of dimensionality that aims at minimizing loss in
variance in original data. Text categorization is the
task of classifying a document into predefined
categories based on the contents of the document. In
recent years, more and more methods have been
applied to the text categorization tasks based on
statistical theories and machine learning, such as
KNN, Naive bayes, Rocchio, Decision tree Support
vector machine.
There are several techniques have been
proposed for the text categorization such as:
2.1 Multi-label text categorization based on a new
linear classifier learning method and a category-
sensitive refinement method [2].
A new approach for dealing with multi-label
text categorization based on a new linear classifier
RESEARCH ARTICLE OPEN ACCESS

learning methodology and a category-sensitive
refinement methodology. Here used is a replacement
weighted classification technique to construct a
multi-label linear classifier. We tend to use the
degrees of similarity between classes to regulate the
connection voluminous classes with relevancy a
testing document. The testing document are often
properly classified into multiple classes by
employing a predefined threshold worth.
2.2 Text feature selection using ant colony
optimization [1].
Feature selection and feature extraction are the
most vital steps in classification systems. Feature
selection is usually used to reduce spatiality of
datasets with tens or many thousands of features
which might be not possible to process more. One of
the issues in which feature selection is crucial is text
categorization. A significant drawback of text
categorization is that the high spatiality of the feature
space; thus, feature selection is the most vital step in
text categorization. At the present there are several
ways to affect text feature choice. To boost the
performance of text categorization, we tend to gift a
unique feature choice algorithmic program that’s
supported ant colony optimization. Ant colony
optimization algorithm is impressed by observation
on real ants in their search for the shortest paths to
food sources. Planned algorithmic program is well
enforced and since of use of a straightforward
classifier there in, its process quality is incredibly
low.
III. METHODOLOGY
A two stage feature selection and feature
extraction is used to reduce the high dimensionality
of a feature space composed of a large range of
terms, remove redundant and irrelevant features from
the feature space and thereby decrease the
computational complexity of the machine learning
algorithms used in the text categorization and
increase performances thereof. In the first stage,
every term within the text is ranked depending on
their importance for the classification in decreasing
order using the Information gain method. Therefore,
terms of high importance are assigned to the first
ranks and terms of less importance are assigned to the
subsequent ranks. Within the second stage, the PCA
method selected for feature selection and the GA
method selected for feature extraction are applied
separately to the terms of highest importance, in
accordance with IG strategies, and a dimension
reduction is carried out. Thus, during text
categorization, terms of less importance are ignored.
The terms with highest importance undergo feature
selection and feature extraction method. And
therefore the computational time and complexity of
the category are reduced. Datasets assortment that
can be used for this could be Classic4 data set for
analyzing the dimensionality reduction.
Fig 1 General architecture for two stage text
categorization.
The proposed system is divided into
following five modules:
3.1 Preprocessing:
Preprocessing is one of the key components
in a typical text categorization framework [3].
Preprocessing impacts text categorization in various
aspects such as categorization accuracy, text domain
and dimension reduction and also promotes
efficiency of text categorization [20].This process
includes four parts:
3.1.1 Removal of stop-words: Words such as
conjunctions and pronouns that are not related to the
concept of the text are called stop-words. This
process involves removing certain common words
such as `a', `an', `the', etc., that occur commonly in all
documents.
3.1.2 Stemming: The stemming process gives us the
root form of all the terms. eg. For “stemming” the
root word is “stem”. Thus stemming process removes
suffixes or prefixes like ing, ed, ied, tion, ly , pre- etc.
Porters algorithm is the most commonly used
algorithm for stemming purpose in textual data.
3.1.3 Term weighting: After the words are
transformed into terms, each term is assigned a
weight based on its frequency of appearance in the
document. This process is called term weighting.
The basic formula used for term weighting include
term frequency (tfi.) and inverse document frequency
(idfi ) as follows :
Wi = tfi.idfi [27]
Where, tfi=term frequency

idfi = Inverse Document Frequency
3.1.4 Pruning : The pruning process basically filters
less frequent features in a document collection. For
example words appearing just twice thrice in a
document are not useful further hence are eliminated
by setting a appropriate threshold value.
3.2 Information gain:
Information gain is one of the popular
approaches employed as a term importance criterion
in the text document data.
where ci represents the ith category, P(ci) is the
probability of the ith category, P(t) and P(t) are the
probabilities that the term t appears or not in the
documents, respectively, P(ci|t) is the conditional
probability of the ith category given that term t
appeared, and P(ci| ) is the conditional probability of
the ith category given that term t does not
appeared[27].
3.3 Principal component analysis:
PCA is a statistical technique used for
extracting information from a multi-variety dataset.
This process is performed via having principal
components of original variables with linear
combinations identified. While the original dataset
with the maximum variability is represented with the
first principal component, the dataset from the
remaining with the maximum variability is
represented with the second principal component
[28,29]. The process goes on like this one after other
as such, with the remaining dataset having the
maximum variability being represented with the next
principal component. Here m represents the number
of all principal components, and p represents the
number of the significant principal components
among all principal components. Further p may be
defined as the number of those principal components
of the m dimensional dataset with the highest
variance values. It is clear therein that p ≤ m.
Therefore, PCA may be defined as a data-reducing
technique. In other words, PCA is a technique used
for producing the lower-dimensional version of the
original dataset.
3.4 Genetic algorithm:
Genetic Algorithms [23,24] are based on the
principles of biological inheritance and evolution.
Every solution which is potential is called an
individual (i.e. a chromosome) in a population. GAs
work iteratively, applying the genetic operations such
as selection, crossover and mutation to a population
of individuals, aimed at creating better adapted
individuals in subsequent generations. For each
category of problems solved by a GA, a fitness
function must be provided. This choice is crucial to
maximize the GA's performance. The fitness function
assigns a fitness value for each individual called the
individual score, and it is well-known to play an
essential role in the genetic evolution [25]. This is
because this score is used in the selection processes
of parents for crossover and of survivals for each next
generation. Thus, the highest probabilities to
reproduce and survive must be given to the best
adapted individuals, that is, the ones who provide the
best features for an accurate query process. Due to its
importance to GA, the fitness function must be
tailored to the problem at hand. Thus, a fitness
function should model the solution space in such a
way that, when solving a maximization problem,
better solutions receive higher scores. Associated
with the characteristics of exploitation and
exploration, GAs can efficiently deal with large
search spaces, and hence are less prone to get stuck
into a local optimum solution when compared to
other algorithms. This derives from the GAs ability to
handle multiple concurrent solutions (individuals) in
the search space and apply probabilistic genetic
operators [23,24]. Genetic algorithm has been
successfully used as an optimization technique
[4][5].The various stages in genetic algorithm [26]
are as follows :
3.4.1 Individual Encoding : The candidate feature set
can be represented by a binary string called
chromosome. In the chromosome the ith bit
represents the presence of the ith feature. If the value
of the gene if coded in binary system as 1, it means
that the corresponding feature is selected and vice
versa. The length of the chromosome is equal the
number of features.
3.4.2 Fitness Function : Fitness function is used to
decide which individuals are to optimum solution.
Every individual has its own fitness value. A higher
value of fitness means that the individual is more
appropriate as a problem solution, on the other hand,
a lower value of fitness means that the individual is
less appropriate as a problem solution.
3.4.3 Selection : The object of the selection process is
to choose the individuals of the next generation
according to the selected fitness function and
selection method among the existing population. In
the selection process, the transfer possibility of the
fittest individuals chromosome to the next generation
is higher than others. The decision of the individuals
characteristic which will be transferred to the next
generation is based on the values evaluated from the

fitness function and shows the quality of individual .
The roulette wheel selection method is generally used
for selection purpose.
3.4.4 Crossover : In this phase an individual is
selected for mating . Forming of new generation by
mating is called crossover. The method is of forming
new individuals from two chromosome. Different
type of crossover procedures are there like single
point, double point and multile point crossover.
3.4.5 Mutation : To increase the variety of
chromosomes the which are applied on crossover,
process mutation process can be applied. Mutation
introduces local variations to the individuals for
searching different solution spaces and keeps the
diversity of the population. In our study, the number
of chromosomes that will be mutated is determined
according to the mutation rate and their values are
changed from 1 to 0 or 0 to 1 respectively.
3.5 Text categorization methods:
For categorization of text documents
effectively following two methods have been
considered.
3.5.1 KNN classifier:
KNN is typical methods of example-based
classifiers. They have been called lazy learners due to
the fact that they do not learn before they start to
classify a document. The KNN method classifies a
testing document by the top k nearest training
documents in the vector space. The difference
between traditional linear classifiers and KNN is that
KNN does not divide the document space linearly
and does not have the drawback of Rocchio’s
method. A number of experiments have shown that
the KNN method is quite effective [21][22].The
KNN algorithm is a well-known instance-based
approach that has been widely applied to text
categorization due to its simplicity and accuracy. To
categorize an unknown document, the KNN classifier
ranks the document's neighbors among the training
documents and uses the class labels of the k most
similar neighbors. Similarity between two documents
may be measured by the Euclidean distance, cosine
measure, etc. The similarity score of each nearest
neighbor document to the test document is used as
the weight of the classes of the neighbor document. If
a specific category is shared by more than one of the
k-nearest neighbors, then the sum of the similarity
scores of those neighbors is obtained from the weight
of that particular shared category.
3.5.2 C4.5 decision tree classifier:
C4.5 algorithm is an improved version of
ID3. This algorithm uses a value called Gain Ratio as
a splitting criterion, unlike ID3 algorithm where gain
is used for splitting criteria in tree growth phase.
Hence C4.5 is an evolution of ID3 [8]. This algorithm
handles both continuous and discrete attributes- In
order to handle continuous attributes, in C4.5 a
threshold is created and then the list gets split into
those whose attribute value is above the threshold
and those that are less than or equal to it[9]. Same as
ID3 the data is sorted at every node of the tree in
order to determine the splitting attribute that is best.
The splitting stops when the number of instances to
be split is below a certain threshold value. The main
plus point of C4.5 is when building a decision tree,
C4.5 can deal with datasets having unknown attribute
valued patterns and datasets with attributes having
continuous domains by discretization. This algorithm
can deal with training data with attribute values by
allowing an attribute value to be marked as missing.
A missing attribute value is simply avoided in gain
and entropy calculations. It includes an enhanced
method of tree pruning that decreases the
misclassification errors due to noise or too much
detail in the training data set.
IV. CONCLUSION
In this paper, we have presented a unique
theme called two stage feature selection method for
text categorization. Every term inside the document
is ranked depending on their importance for
classification using the information gain (IG)
methodology. Information Gain methodology is
proved to be best among other feature selection
techniques[27]. In the next stage genetic algorithm
(GA) and principal component analysis (PCA)
feature selection and feature extraction methods are
applied individually to the terms which are ranked in
decreasing order of importance, and after that a
dimension reduction is carried out. This proves to be
better than the other methodologies generally used
which involve single stage feature selection and
extraction. Two stage feature selection improves
performance since filtering for only important
relevant features is done twice ,hence , reducing
noise and dimensionality of feature space. To analyze
the dimension reduction we use the k-nearest
neighbor (KNN)[22] and C4.5 decision tree
algorithmic rule on selected data set. The evaluation
results shows that, the computational time and
complexity of categorization is reduced.
REFERENCES
[1] M.H. Aghdam, N. Ghasem-Aghaee, M.E.
Basiri, Text feature selection using ant colony
optimization, Expert Systems with Applications
36 (2009) 6843–6853.
[2] Yu-Chuan Chang, Shyi-Ming Chen, Churn-
Jung Liau, Multi-label text categorization
based on a new linear classifier learning
method and a category-sensitive refinement

method, Expert System with Application
34(2008) 1948-1953.
[3] Jiawei Han, Jaian Pei, Micheline Kamberg,
Data mining concepts and techniques, third
edition.
[4] F.J. Damerau, T. Zhang, S.M. Weiss, N.
Indurkhya, Text categorization for a
comprehensive time-dependent benchmark,
Information Processing and Management 40
(2004) 209–221.
[5] M.E. ElAlami, A filter model for feature
subset selection based on genetic algorithm,
Knowledge-Based Systems 22 (2009) 356–362.
[6] M. Gen, R. Cheng, Genetic Algorithms and
Engineering Optimization, vol. 68, Wiley
Interscience Publication, 2000
[7] Y. Li, D.F. Hsu, S.M. Chung, Combining
multiple feature selection methods for text
categorization by using rank-score
characteristics, in: 21st IEEE International
Conference on Tools with Artificial
Intelligence, 2009, pp. 508–517.
[8] M atthew N. Anyanwu & Sajjan G.
Shiva, “ Comparative Analysis of Serial
Decision Tree Classification Algorithms”,
International Journal of Computer Science
and Security, 2009, (IJCSS)Volume (3): Issue
(3).
[9] J. R. Quinlan, 1996, “Improved use
of continuous attributes in C4.5” , Journal of
Artificial Intelligence Research, Vol. 4, pp. 77-
90
[10] V. Mitra, C.-J. Wang, S. Banerjee, Text
classification: a least square support vector
machine approach, Applied Soft Computing 7
(2007) 908–914
[11] D. E. R. O. Alonso and B. Stewart,
Crowdsourcing for relevance evaluation,"
SIGIR Forum,November 2008.
[12] F. Sebastiani, A tutorial on automated text
categorisation, in: Proceedings of the ASAI-99,
in: 1st Argentinian Symposium on Artificial
Intelligence, Buenos Aires, AR., 1999, pp. 17–
35.
[13] A. McCallum, K. Nigam, A comparison of
event models for Naive Bayes text
classification, in: AAAI’98 Workshop on
Learning for Text Categorization, 1998, pp.
41–48.
[14] W. Lam, Y. Han, Automatic textual document
categorization based on generalized instance
sets and a metamodel, Proceeding of the IEEE
Transactions on Pattern Analysis and Machine
Intelligence 25 (5) (2003) 628–633.
[15] Ghiassi, M., Olschimke, M., Moon, B., &
Arnaudo, P.. Automated text classification
using a dynamic artificial neural network
model. Expert Systems with Applications,
(2012), 39, 10967–10976.
[16] Ergin, S., Gunal, E. S., Yigit, H., & Aydin, R.
Turkish anti-spam filtering using binary and
probabilistic models. AWERProcedia
Information Technology and Computer
Science, (2012),1, 1007–1012.
[17] Uysal, A. K., Gunal, S., Ergin, S., & Gunal,
E.S.. A novel framework for sms spam
filtering. In Proceedings of the IEEE
international symposium on innovations in
intelligent systems and applications. Trabzon,
Turkiye. (2012)
[18] Cheng, N., Chandramouli, R., & Subbalakshmi,
K. P.. Author gender identification from text.
Digital Investigation, (2011) ,8, 78–88.
[19] Ozel, S. A. , A web page classification system
based on a genetic algorithm using tagged-
terms as features. Expert Systems with
Applications, (2011),38,3407–3415.
[20] Alper Kursat Uysal, Serken Gunal , The
impact of preprocessing on text classification,
Information Processing and Management 50
(2014) 104–112
[21] Yang, Y. , An evaluation of statistical
approaches to text categorization.Information
Retrieval, (1999). 69–90.
[22] Tan, S. (2005). Neighbor-weighted K-nearest
neighbor for unbalanced text corpus. Expert
Systems with Applications, 28(4), 667–671.
[23] D.E. Golberg, Genetic algorithms in search,
optimization and machine learning,Addison
Wesley,1989
[24] R.L. Haupt, S.E. Haupt, Practical Genetic
Algorithms, second edition, John Wiley &
Sons, New Jersey, United States, 2004.
[25] C. López-Pujalte, V.P. Guerrero-Bote, F.
Moya-Anegón, Order- based fitness functions
for genetic algorithms applied to relevace
feedback, Journal of the American Society for
Information (2) Science 54(2003) 152–160.
[26] Sergio Francisco da Silva, Marcela Xavier
Ribeiro, João do E.S. Batista Neto , Caetano
Traina-Jr., Agma J.M. Traina, Improving the
ranking quality of medical image retrieval
using a genetic feature selection method ,
Decision Support Systems 51 (2011) 810–820.
[27] Yang, Y., & Pedersen, J. O. (1997). A
comparative study on feature selection in text
categorization. In 14th international
conference on machine learning (pp. 412–
420). Morgan Kaufmann Publishers Inc.
[28] Harun Uguz, “A hybrid approach for text
categorization by using x2 stastics, Principal
Component Analysis and partical swam
optimization”, Academic Journals, vol(8) 37,
PP 1818-1828, 2013.
[29] Kemal Polat, Salih Gunes, “ Prediction of
hepatitis disease based on principal component
analysis and artificial immune recognition
system”, Applied Mathematics and
computation 189(2007) 1282-1291.

N045038690

More Related Content

What's hot (20)

Viewers also liked (20)

Similar to N045038690 (20)

Recently uploaded (20)

N045038690