Sentiment Analysis and Word Association Patterns in Skincare Product Customer Reviews

Abstract

The growth of the skincare industry and the increasing activity of consumer reviews on e-commerce platforms have generated large text data containing customer opinions, experiences, and perceptions. This study aims to analyze sentiment and identify word association patterns in Indonesian-language customer reviews of skincare products. The literature review covers sentiment analysis, Natural Language Processing, the IndoBERT language model, Data Mining, Knowledge Discovery in Databases, as well as Association Rule Mining using the Apriori algorithm. The research method uses a quantitative approach based on KDD, which includes Data Selection, Preprocessing, Data Transformation, Data Mining, and interpretation and evaluation. Data was obtained through web scraping of skincare product reviews on the Shopee platform, resulting in 7,320 clean reviews. Sentiment analysis was conducted using IndoBERT with a Hybrid Linguistic approach to handle neutral rating ambiguities. The results of the sentiment classification were then used as the basis for analyzing word association patterns using the Apriori algorithm for each sentiment category. The findings indicate that IndoBERT is capable of classifying sentiment contextually, while Apriori successfully uncovers word patterns that represent product aspects such as quality, effectiveness, and user experience. This study concludes that the integration of sentiment analysis and word association patterns provides a more comprehensive understanding of consumer perceptions and can be utilized as a basis for strategic decision-making in the skincare industry.