Abstract
AbstractIn recent times, comparative opinion mining applications have attracted both individuals and business organizations to compare the strengths and weakness of products. Prior works on comparative opinion mining have focused on applying a single classifier, limited comparative opinion labels, and limited dataset of product reviews, resulting in degraded performance for classifying comparative reviews. In this work, we perform multi-class comparative opinion mining by applying multiple machine learning classifiers using an increased number of comparative opinion labels (9 classes) on 4 datasets of comparative product reviews. The experimental results show that Random Forest classifier has outperformed the comparing algorithms in terms of improved accuracy, precision, recall and f-measure.
Highlights
A massive volume of user-generated content is created every day on social media sites like Facebook, YouTube, and Twitter [1] Opinion mining and sentiment analysis (SA) aims at investigating the public’s opinions, sentiment, evaluations, and emotions
We aim to investigate that the performance difference between the Random Forest and the Decision Tree, Random Forest vs Support Vector Machine classifiers, is statistically significant and does not occur randomly
This work deals with the performance-based comparison of different machine learning classifiers with respect to comparative opinion mining
Summary
A massive volume of user-generated content is created every day on social media sites like Facebook, YouTube, and Twitter [1] Opinion mining and sentiment analysis (SA) aims at investigating the public’s opinions, sentiment, evaluations, and emotions. Comparative opinion mining assists business organizations to analyze customer feedback about their manufactured goods in terms of both strengths and weaknesses. Collection and analysis of such comparative opinions is beneficial for business organizations to enhance the quality of their manufactured goods and individuals when buying products. The early studies on comparative opinion mining have used limited datasets (400 reviews dataset) with a limited number of machine learning classifiers [6,7,8]. There is a need to explore multiple machine learning classifiers with datasets of increased sizes for performing comparative opinion mining based on comparative reviews for different products. This study aims at conducting and applying different machine learning classifiers for performing comparative opinion mining and recommending the classifier with best results
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
