Sentiment Analysis of Skincare Product Reviews on the Female Daily Platform Using TF-IDF and Multinomial Naïve Bayes

The rapid growth of the skincare industry has led to an increasing number of user reviews on digital platforms, including Female Daily. These reviews provided valuable insights into consumers' perceptions of skincare products. However, the large volume of review data made manual analysis inefficient, creating the need for an automated sentiment analysis approach. This study aimed to analyze the sentiment of skincare product reviews on the Female Daily platform using Natural Language Processing (NLP), Term Frequency–Inverse Document Frequency (TF-IDF) weighting, and the Multinomial Naïve Bayes algorithm. The dataset was collected through web scraping using the Instant Data Scraper extension, resulting in 1.200 reviews. After preprocessing, including case folding, cleaning, normalization, stopword removal, and stemming, 1.198 reviews met the predefined criteria and were retained for analysis. The dataset was divided into training and testing sets using an 80:20 ratio. TF-IDF was applied to extract numerical features, and the resulting vectors were classified using the Multinomial Naïve Bayes algorithm. The experimental results showed that the proposed model achieved an accuracy of 74.58%, a precision of 72.03%, a recall of 74.58%, and an F1-score of 71.96%. The model performed well in classifying positive sentiments but showed lower performance in identifying neutral sentiments. Overall, the combination of TF-IDF and Multinomial Naïve Bayes provided satisfactory performance for sentiment analysis and demonstrated its effectiveness in understanding users' opinions toward skincare products.