Enhanced Sentiment Analysis for Amazon Reviews Using Butterfly Optimization and Random Forest

Abstract

In today's digital age, online shopping significantly impacts consumer purchasing behaviour. Customer reviews play a crucial role in influencing purchasing decisions, with positive feedback driving others to make similar purchases, while negative reviews deter them. Sentiment Analysis on Amazon product review data helps consumers evaluate products effectively. However, traditional machine learning methods have limitations in accuracy and feature selection, affecting sentiment classification performance. This paper introduces an Advanced Butterfly Optimization (ABO) algorithm to address these challenges, enabling the selection of the best features from the dataset and improving classification accuracy. This approach combines the Term Frequency-Inverse Document Frequency (TF-IDF) weighting scheme, the ABO algorithm, and the Random Forest (RF) algorithm. The ABO and TF-IDF algorithms extract relevant attributes from the input data, and the RF algorithm categorizes the data into positive or negative. The ABO-RF method proposed in this paper demonstrates 89.9% higher accuracy than existing methods such as Naive Bayes with GA, SVM, and Logistic Regression. The results of the simulation experiments show that the proposed approach outperforms traditional methods like Naive Bayes with GA, SVM, and Logistic Regression. Therefore, the ABO-RF algorithm is more effective in constructing a classification model for Amazon user reviews.