Sentiment Analysis of Lazada Marketplace User Ratings with Naïve Bayes and Support Vector Machine Methods

Abstract

This research analyzes the sentiment of Lazada Marketplace users using Naïve Bayes and Support Vector Machine (SVM) methods. This sentiment analysis is important to understand customer perceptions, which can help Lazada in developing service improvement strategies and business policies. Data was obtained through web scraping, resulting in 5,261 comments divided into 80% training data and 20% test data. The text preprocessing process includes cleaning, case folding, stopword removal, tokenizing, and stemming to ensure the data is cleaner and more structured. The test results show that SVM performs better with 75% accuracy, 74% precision, 86% recall, and 79% F1-score, compared to Naïve Bayes which has 72% accuracy, 75% precision, 76% recall, and 76% F1-score on positive sentiment. The confusion matrix evaluation shows that SVM is more consistent in classifying positive, neutral, and negative sentiments. Word cloud visualization on negative sentiment revealed that the main customer complaints were related to the app, late delivery, and product mismatch. These findings can be used by Lazada to improve service quality by improving the technical aspects of the app, optimizing logistics, as well as improving quality control of products sold by the store to increase customer satisfaction and user loyalty.