See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/319975138

Sentiment Analysis using Naïve bayes Algorithm

Article  in  INTERNATIONAL JOURNAL OF COMPUTER SCIENCES AND ENGINEERING · July 2017

DOI: 10.26438/ijcse/v5i7.7577

CITATIONS READS
4 992

3 authors, including:

Arati shivanand Shahapurkar Kavita Hanabaratti

Gogte Institute of Technology Gogte Institute of Technology
5 PUBLICATIONS   5 CITATIONS    6 PUBLICATIONS   11 CITATIONS   

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Kavita Hanabaratti on 22 March 2022.

The user has requested enhancement of the downloaded file.

 International Journal of Computer Sciences and Engineering Open Access
Review Paper Volume-5, Issue-7 E-ISSN: 2347-2693

Sentiment Analysis using Naïve bayes Algorithm

Pramod M. Mathapati1*, A.S. Shahapurkar2, K.D.Hanabaratti3
1*
Dept. of Computer Science, Gogte Institute of Technology, VTU, Belagavi, India
2
Dept of Computer Science, Gogte Institute of Technology, VTU, Belagavi, India
3
,Dept of Computer Science, Gogte Institute of Technology, VTU, Belagavi, India
*Corresponding Author: pramodkle064@gmail.com
Available online at: www.ijcseonline.org

Received: 26/Jun/2017, Revised: 07/Jul/2017, Accepted: 20/Jul/2017, Published: 30/Jul/2017

Abstract— Sentiment analysis is trending topic of research which works on data which is got from review websites, social
networks. Today users having common platforms like Blogs, micro blogs, review sites, twitter and other social networks
through which they can post their feedbacks. Organizations use Sentiment Analysis to understand user’s reviews and feedbacks
about the product which they have released. In this project development of a Sentiment analysis using a generic method which
can be applied for sentiment analysis as well as Emotional Analysis, product reviews is done based on Naïve Bayes classifier
method. Naive bayes Classifier is the better choice for Sentiment Analysis as it is more efficient and gives Quick results
compared to other techniques such as Support Vector Machine and Maximum Entropy.

Keywords—Sentiment Analysis, Modified k means, NLP, Opinion Mining,

I. INTRODUCTION and services they are using through websites which are
present on the internet [4,7,]. Decision making for new user
People and Organizations express their views and opinions
on internet as Sentiments influence human actions, activities. will be decided by these data. Apart from Twitter and face
Sentiments are expressed by people every day through book we have different social networks as well which also
works as data source for sentiment analysis. There are many
internet as people so much dependent on it. There is need of
websites available where they can provide data for Sentiment
analyzing these sentiments hence opinions or views of user
Analysis such as www.amazon.com (product reviews),
have got more importance. Today users having common
www.reviewcentre.com (product reviews),
platforms like Face book, Instagram, review sites, twitter and
other social networks through which they can post their www.fonearena.com (mobile reviews), where millions of
feedback about the product they are using and they can even reviews are posted from the customers [1,8,9,10]. Even we
can have movie review dataset which is used to identify
post their comments whether they are happy with the service
distinguished features, method of selection of feature and
which they got or not [1,2,3]. Likewise even organizations
best supervised learning algorithm. Cornell Movie Review
can also share their experiences. Sentiment analysis is
trending topic of research which works on data which is got Dataset is the classical dataset which consists of thousand
from review websites, social networks [4,5,6]. negative reviews and thousands of positive reviews. People
can easily access this dataset.
In this Work development of a Sentiment analysis using a
generic method which can be applied for sentiment analysis
as well as Emotional Analysis, product reviews is done based
on Naïve Bayes classifier method. Researchers have got new
opportunities as well as challenges while working with
Sentiment Analysis which can also be used for natural
language processing [2].

II. RELATED WORK

Today people are using Twitter and Face book as the main
social networks and even they are known as Micro Blogs as
they allow user to broadcast short text messages, images,
audio or video to other users. Messages which are exchanged
by these networks can be used as data for sentiment analysis. Fig1. Proposed Architecture
Today users are generating thousands of reviews for product

© 2017, IJCSE All Rights Reserved 75

 International Journal of Computer Sciences and Engineering Vol.5(7), Jul 2017, E-ISSN: 2347-2693

III. METHODOLOGY 5) Naïve Bayes Classification.

Document can be classified by applying Naïve Bayes
In this Paper five modules have been used. theorem

1) User interface Module Naive Bayes (NB)

In this module development of a web portal which will allow To get exact and proficient results for linearly separable
any user to sign up and achieve sentimental analysis on any cases Naïve bayes is the best choice and even it can be used
kind of data regardless of any type like twitter data, movies for non-linearly separable cases also. Because of its simple
reviews, product review, etc. Even the portal will keep track interpretation and effective computation naïve bayes is used
of all recent activities the user has done which will kept as in different applications. When probability of instance which
reference for his future work. User has an option to supply has already occurred is known then naïve bayes is used to
input as a text file or just a sentence as an input which will be calculate the probability of current instance. We can write
further processed by the web portal. The web module will like P (D | C) P(C) in Mathematics
also have a admin section where he can get all user list with P (D): Probability of document D occurring
extra options, even a section where he can update the training P(C | D) = Where, P (C | D): Probability of Document D
data set of the sentimental analysis.By managing unrelated being in Class C
data and increases accuracy with Naïve Bayes Classification P(C): Probability of occurrence of Class C,
algorithm. P (D) P (D | C): Probability of generating Document D given
Class C.
2) Training the Classifiers
Decision making is very important for organizations as they Naïve Bayes Classifier
have to understand how people accepting their product when  Simple classification of words based on ‘Bayes
it is newly launched in the market. That can be understood by theorem’.
the reviews, compliments or opinions given by the people so  It is a ‘Bag of words’ (text represented as collection of
that organizations can decide about their future decisions. its words, discarding grammar and order of words but
keeping multiplicity) approach for subjective analysis
3) Pre-processing Module of content.
Preprocess data  Better in terms of CPU and Memory utilization
1) @username - elimination of "@username" via regex
matching or replace it with generic word AT_USER is Probabilistic Analysis of Naïve Bayes
possible. For a document d and class c, By Bayes theorem
2) #hashtag - hash tags can give some useful information, so
it is useful to replace them with the exact same word p(c/d)=p(d/c)p(c) / p(d)
without the hash. E.g. #nike replaced with 'nike'.
3) Lower Case – tweets will be converted to lower case. Naïve Bayes Classifier will be:
4) URLs - No need to follow the short urls and determine the
content of the site, so we can remove all of these URLs via C* = arg maxc p(c/d)
regular expression matching or replace with generic word
URL. Multinomial Naïve Bayes Classifier
5) Punctuations and additional white spaces - remove Accuracy – around 75%
punctuation at the start and ending of the tweets. Ex: ' the Algorithm - :
evening is wonderful! ‘Replaced with 'the evening is  Dictionary Generation
wonderful'. It is also helpful to replace multiple Making a dictionary of some most frequent
whitespaces with a single whitespace. words can be achieved by counting
occurrences of all words in given dataset.
4) Feature Extraction  Generation of Feature Set
Selecting a subset of the words appearing in the training
Database, feature extraction can be achieved. In text Representation of all documents as a feature vector over the
categorization only this subset should be taken as feature. space of dictionary words.
Two main causes for feature extraction:
1. Amount of useful words is decreased so that training and It is required to keep track of dictionary words with their
applying a classifier method can be more capable. number of occurrences in the document. .
2. Classification accuracy can be improved by removing
noise features and which also enhance feature extraction. IV. RESULTS AND DISCUSSION

© 2017, IJCSE All Rights Reserved 76

 International Journal of Computer Sciences and Engineering Vol.5(7), Jul 2017, E-ISSN: 2347-2693

Naive Bayes algorithm has given quick and accurate results [8]. C.P. Robert and G. Casella, “Monte Carlo Statistical Methods,
compared to other methods like SVM(Support Vector second ed. Springer Publisher 2005.
[9]. S. Morinaga, K. Yamanishi, K. Tateishi, and T. Fuku-shinna,
Machine) and Maximum Entropy. Time efficiency is also
“Mining Product Reputations on the Web,” Proc. Eighth ACM
high in case of Naive Bayes as SVM and Maximum Entropy SIGKDDInt’l Conf. Knowledge Discovery and Data Mining
takes more time to classify the input data and process. Naive (SIGKDD ’02), pp. 341-349, 2002.
bayes saves time for computing as it don’t need and [10]. [10] M. Hu and B. Liu, “Mining and Summarizing Customer
preprocessing mechanism to classify the given input data Reviews,” Proc. 10th ACM SIGKDD Int’l Conf. Knowledge
hence it saves time. Discovery and Data Mining (SIGKDD ’04), pp. 168-177, 2004.

V. CONCLUSION AND FUTURE SCOPE

In this Paper, proposal of a method using naïve bayes has

given more quick and accurate results compared to other
techniques of Sentiment Analysis such as Support Vector
Machine and Maximum Entropy. More time is needed to
SVM and Maximum entropy to give optimum results. Naive
bayes classifier can be used for both Sentiment and Emotion
Analysis. High accuracy is achieved for sentiment analysis
and as well as for product reviews, movie reviews. Even for
text based classification and for social interpretation it gives
better results.

In future focus will be on to find out how this method can be

used as more generic method when applied to customer
reviews, product reviews, human sentiments etc. Focus will
be on improvement and more accurate results for sentiment
analysis. Efforts will be made to add some additional features
in Naïve Bayes classifier algorithm so that it can be used to
develop a more generic method. Different Social networks
can use this method as their API.
REFERENCES

[1]. O. Coban and B. O. and G. T. Ozyer, “A Comparison of Similarity

Metrics for Sentiment Analysis on Turkish Twitter Feeds,” 2015
IEEE International Conference on Smart City /SocialCom
/SustainCom (SmartCity), Dec. 2015.
[2]. Pablo Gamallo, CITIUS, Univ. de Santiago de Compostela Citius:
“A Naive-Bayes Strategy for Sentiment Analysis on English
Tweets” Proceedings of the 8th International Workshop on
Semantic Evaluation(SemEval 2014), pages 171–175, Dublin,
Ireland, 2014
[3]. Shakhy P S and Vidya K S, "A Survey on KASR for Big Data
Applications", International Journal of Computer Sciences and
Engineering, Vol.3, Issue.4, pp.85-89, 2015.
[4]. Ashish Shukla, Rahul Mishra, “Sentiment Classification and
Analysis using Modified K-means and naïve Bayes Algorithm”
International journal of Advanced Research in computer science
and Software Engineering. Volume 5, Issue 8, August 2015.
[5]. J. Kaur, S.S. Sehra, S.K. Sehra, "A Systematic Literature Review of
Sentiment Analysis Techniques", International Journal of
Computer Sciences and Engineering, Vol.5, Issue.4, pp.22-28,
2017.
[6]. U. Aggarwal, G. Aggarwal, "Sentiment Analysis : A Survey",
International Journal of Computer Sciences and Engineering,
Vol.5, Issue.5, pp.222-225, 2017.
[7]. D.M. Blei, A.Y. Ng, and M.I. Jordan, “Latent Dirichlet
Allocation,” J. Machine Learning Research, vol. 3, pp. 993-1022,
2003.

© 2017, IJCSE All Rights Reserved 77

View publication stats