Sentiment Analysis with Soft-Voting Method on Customer Reviews for Purchasing Transactions of E-Commerce

Abstract

The accuracy of customer reviews is crucial for an e-commerce platform to assist buyers in selecting high-quality products from a vast array of options. This research aims to develop a sentiment analysis model for evaluating customer opinions expressed in e-commerce product reviews. The proposed approach utilizes the Soft Voting (SV) technique, which demonstrates superior accuracy compared to the conventional Sentiment Selector (SS) method. The sentiment analysis model’s accuracy is determined by gathering probability values from three classifiers (Support Vector Machine (SVM), Random Forest (RF), and Naive Bayes (NB)) for each sentiment category (positive, neutral, negative). The evaluation is conducted using the Tokopedia product review dataset. The findings indicate that the Soft Voting (SV) model outperforms the Sentiment Selector (SS) approach. The proposed SV model achieves an accuracy, precision, recall, and f1-score of 69%, 70%, 69%, and 69%, respectively.

Access to Document

Other files and links

• Link to publication in Scopus

Fingerprint