Classification of Consumer Sentiment towards Muslim Fashion Products Using TF-IDF and Machine Learning

Abstract

The growth of e-commerce has driven an increase in transactions for Muslim fashion products, generating a vast number of consumer reviews containing information about product experiences and perceptions. Analyzing these reviews is crucial, as sentiment information allows for a more systematic understanding of consumer satisfaction trends compared to manual analysis. This study aims to classify consumer sentiment regarding Muslim fashion products and compare the performance of several machine learning algorithms in this task. The research dataset was obtained from PRDECT-ID, initially comprising 5,400 reviews; filtering for the Muslim fashion category yielded 200 reviews. Sentiments were determined based on review content, and "Neutral" labeled data were excluded, resulting in a final set of 135 reviews (97 negative and 38 positive). The research process involved text preprocessing, feature weighting using Term Frequency-Inverse Document Frequency (TF-IDF), classification using Multinomial Naive Bayes (MNB), Logistic Regression (LR), and Support Vector Machine (SVM), and evaluation using Stratified 5-Fold Cross-Validation. This study contributes an empirical comparison of the three algorithms applied to Muslim fashion product reviews, accounting for class imbalance and multiple evaluation metrics. Experimental results indicate that SVM achieved the best performance, with an accuracy of 84.44%, precision of 81.43%, recall of 60.36%, and an F1-score of 68.00%. These findings demonstrate that the combination of TF-IDF and SVM is more effective than MNB and LR for sentiment classification on the study's dataset.