Page 56 - IJOCTA-15-1

An International Journal of Optimization and Control: Theories & Applications
ISSN: 2146-0957 eISSN: 2146-5703
Vol.15, No.1, pp.50-70 (2025)
https://doi.org/10.36922/ijocta.1655

RESEARCH ARTICLE

BSO: Binary Sailfish Optimization for Feature Selection in
Sentiment Analysis

1
1*
Omer Ayana , Deniz Furkan Kanbak , Mumine Kaya Keles 2
1 Department of Software Engineering, University of Adana Alparslan Turkes Science and Technology, Turkey
2 Department of Computer Engineering, University of Adana Alparslan Turkes Science and Technology, Turkey
oayana@atu.edu.tr, dfkanbak@atu.edu.tr, mkaya@atu.edu.tr

ARTICLE INFO ABSTRACT

Article History: Sentiment analysis (SA) plays a critical role in various domains, providing valu-
Received: 1 August 2024 able insights into public opinion regarding brands, products, and events. By
Accepted: 4 January 2025 leveraging this method, companies can enhance customer satisfaction through
Available Online: 23 January 2025 informed adjustments to their products. This study aims to implement sen-
timent analysis on user comments from online sales platforms. We propose
Keywords:
and evaluate four machine learning (ML) algorithms alongside a deep learning
Binary sailfish optimization
(DL) model. Moreover, our dataset contains noise data that is unsuitable for
Deep learning
classification, which negatively impacts performance. To address this issue, fea-
Machine learning
ture selection methods are employed to facilitate the algorithms in identifying
Sentiment analysis
meaningful patterns more effectively, thereby reducing computational time by
Text preprocessing
focusing on the most contributive features within the dataset. In this context,
AMS Classification 2010: we apply the binary variant of the Sailfish Optimization Algorithm (SOA),
68T20; 90C27 referred to as the Binary Sailfish Optimizer (BSO), as a feature selection tech-
nique tailored for our textual dataset, marking its inaugural application in
sentiment analysis. To assess the effectiveness of the BSO, we conduct com-
parative analyses against four other optimization algorithms: Harmony Search
(HS), Bat Algorithm (BA), Atom Search Optimization (ASO), and Whale Op-
timization algorithm (WOA). Our findings indicate that the BSO outperforms
the existing algorithms, achieving an F-score of 0.91 while utilizing nearly half
of the available features.

1. Introduction there is also the possibility that the expansion of
data in the future will be uncontrollable due to
these social networking sites. 2,3 The ability for
Since there are many products in the market, anyone to comment on products at any time has
it is difficult for the customers to choose a de- led to a tremendous and continuous increase in
sired product. Likewise, a situation similar to online data and information. Therefore, it has
the market environment is observed in the Web become quite difficult to extract relevant infor-
environment. With the increasing product and mation from the Internet accurately. With Sen-
comment data on websites, it is getting harder timent Analysis (SA), which is among the main
for customers to find the products they prefer on topics of natural language processing (NLP), not
e-commerce sites. Customers make the decision only manufacturers but also customers will be
they want by focusing on the comments found on able to analyze positive and negative comments
the web. With the popularity of social networks in about each product. By using SA, the mood or
recent years, reviews and comments on products behavior of the person who criticizes or comments
1
can be seen on social networking sites. However, can be determined as positive or negative. 4
*Corresponding Author
50