Ensemble of feature ranking methods using hesitant fuzzy sets for sentiment classification

Abstract

The increase in the volume of opinion posted onsocial media sites has led to a tremendous increase in thedimensionality of data used for the sentiment analysis. Theselection of informative features from textual data can improvethe performance of supervised learning methods. In this article,we propose a novel and efficient method for integratingdifferent filter-based feature selection methods for sentimentclassification. The ensemble method utilizes hesitant fuzzy setsfor representing opinions of different filter-based featureselection methods in order to optimize the relevancy scoreamong features and class labels. Based on this relevancy score,top-k ranked features are selected for sentiment classification.The proposed feature selection method with Naïve Bayes andSupport Vector Machine classifiers was evaluated on three mostwidely used datasets for sentiment analysis using Unigram andParts-of-Speech based text representation schemes. Theperformance is evaluated using five-fold cross validationtechnique and the results show that the proposed method canachieve greater value of accuracy with only 10-25% of totalextracted features. The outcomes of comparison carried out viastatistical tests confirm that the aggregation using hesitantfuzzy sets is more effective than baseline feature selectionmethods on Parts-of-Speech features in terms of performancemetrics.