Comparison of Naive Bayes and Support Vector Machine (SVM) Methods in Female Daily Skincare Sentiment Analysis

Perbandingan Metode Naive Bayes dan Support Vector Machine (SVM) dalam Analisis Sentimen Skincare Female Daily

Abstract

The development of the beauty industry in Indonesia has increased significantly, along with the high participation of consumers in providing online reviews, especially through the Female Daily platform. These reviews contain opinions that can be analyzed to determine consumer sentiment towards a product. This study seeks to perform sentiment analysis on beauty product reviews by applying two text classification techniques: Support Vector Machine (SVM) and Naive Bayes. The research stages include collecting review data from Female Daily, text preprocessing (such as tokenization, stemming, and stopword removal), and sentiment classification into two categories, namely Yes and No. The evaluation results indicate that the SVM method outperforms Naive Bayes, achieving a higher level of accuracy. SVM is able to capture more complex patterns in text data, while Naive Bayes tends to produce lower performance due to the assumption of independence between features. The evaluation results demonstrate that the SVM method performs better than Naive Bayes, achieving higher accuracy scores. SVM excels at recognizing more intricate patterns in textual data, whereas Naive Bayes often shows lower performance due to its assumption of feature independence. Overall, the majority of user reviews are positive, reflecting satisfaction with certain beauty products. This study shows that the SVM method is more recommended for text-based sentiment analysis of reviews in the beauty industry, especially in the context of diverse and unstructured consumer review data. As for the accuracy results of the Naive Bayes method is 80% and the accuracy results of the SVM method is 87%.