Sentiment analysis of online reviews for electric vehicles using the SMAA-2 method and interval type-2 fuzzy sets

Abstract

Numerous electric vehicles (EVs) consumers express their opinions about EVs on automotive websites. These online reviews can be mined and analyzed to understand consumer sentiment and preferences. Although the existing sentiment analysis (SA) has made insight into sentiment polarity of EVs, the SA results on different EV attributes do not consider the accuracy of sentiment classification and the difficulty of accurately obtaining attribute weights in the process of all attribute information fusion. To address these issues, we provide an integrated framework for the EV SA problem using the stochastic multi-criteria acceptability analysis (SMAA) method with interval type-2 fuzzy sets (IT2FSs). We first use a logistic regression model to classify the sentiment polarity for each online comment and construct IT2FS by sentiment classification accuracy to represent the SA results for different EV attributes. Based on this, we considered all feasible weights in the weight space and utilize the SMAA-IT2FS method to aggregate SA results and achieve EV ranking. In addition, our approach provide decision makers with information to assist in decision-making such as the use of central weight vectors to identify the strengths and weaknesses of EVs and the analysis of potential competitors through dominance relationships. Finally, real data extracted from AutoHome websites are used as experimental data to illustrate the implementation of the proposed method and to demonstrate the effectiveness of our approach.