Long Short-Term Memory and Bidirectional Long Short-Term Memory Algorithms for Sentiment Analysis of Skintific Product Reviews

Abstract

In the era of ever-evolving digital technology, conducting customer sentiment analysis through product reviews has become crucial for businesses to improve their offerings and increase customer satisfaction. This research aims to analyze the sentiment of SKINTIFIC skincare products on the Shopee online store platform using advanced deep learning models: Long Short-Term Memory (LSTM) and Bidirectional Long Short-Term Memory (Bi-LSTM). These models were evaluated using learning rate, number of units, and dropout rate. The dataset consists of 9,184 product reviews extracted through the Shopee API. The reviews were pre-processed using stemming, normalization, and stopword removal techniques. The Bi-LSTM model showed superior performance, achieving an average accuracy of 95.91% and an average F1 score of 95.82%, compared to the standard LSTM model. The optimal configuration for Bi-LSTM included a learning rate 0.01, 64 units, and a dropout rate 0.2. These findings underscore the effectiveness of Bi-LSTM in understanding and classifying consumer sentiment toward specific products.