IMPLEMENTATION OF SUPPORT VECTOR MACHINE, PARTICLE SWARM OPTIMIZATION, AND NAÏVE BAYES ALGORITHMS IN SENTIMENT ANALYSIS OF PRODUCT REVIEWS: A CASE STUDY OF E-COMMERCE LAZADA

Sentiment analysis is pivotal in deciphering customer opinions and attitudes towards products on e-commerce platforms such as Lazada. Machine learning algorithms like Support Vector Machine (SVM), SVM with Particle Swarm Optimization (PSO), and Naïve Bayes (NB) are leveraged to automate this process, aiding decision-making in business settings. This study specifically aims to assess the performance of SVM, SVM + PSO, and NB in analyzing sentiment from Lazada product reviews, focusing on key metrics like accuracy and Area Under the Curve (AUC). Using a dataset of Lazada reviews, each algorithm is rigorously trained and evaluated. SVM achieves 72.74% accuracy and an AUC of 0.893, while integrating PSO boosts accuracy significantly to 84.84% with an AUC of 0.898. In contrast, NB achieves 75.34% accuracy and an AUC of 0.663. These results highlight SVM + PSO's superior performance in sentiment classification compared to SVM and NB. The findings suggest that SVM + PSO presents a robust solution for sentiment analysis in e-commerce, surpassing traditional SVM and NB methods in accuracy and AUC metrics. This underscores the potential of optimization techniques like PSO to enhance machine learning algorithms for effective sentiment analysis in practical e-commerce applications.