PENERAPAN FEATURE SELECTION UNTUK MENINGKATKAN KINERJA METODE NAIVE BAYES DALAM ANALISIS SENTIMEN KONSUMEN

Abstrak

The rapid growth of social commerce platforms such as TikTok Shop has significantly

increased consumer interactions, generating a vast number of opinions and product reviews

in real time. These reviews serve as valuable sources of information for companies to better

understand consumer perceptions of their products. This study aims to analyze consumer

sentiment toward Wardah Cushion products on the TikTok Shop platform by employing the

Naive Bayes method and applying Feature Selection techniques. The analysis was conducted

under two scenarios: without feature selection and with feature selection using the Chi-

Square method. The data preprocessing stage involved cleaning, case folding, stopword

removal, stemming, and tokenization to prepare the text before classification. Model

performance was evaluated using accuracy, precision, recall, and f1-score metrics. The

experimental results show that the application of feature selection improved model accuracy

from 88.84% to 92.15%. Based on these findings, it can be concluded that selecting relevant

features has a positive impact on the performance of the Naive Bayes model in classifying

consumer sentiment more accurately. This research contributes to the utilization of text

mining and machine learning in sentiment analysis for social media-based e-commerce

platforms. For future work, it is recommended to employ larger datasets and compare

different classification algorithms to obtain more comprehensive results.