Opinion-Mining Based Fake Review Detection for Genuine Online Products

The widespread growth of e-commerce platforms has significantly increased the reliance of consumers on online product reviews. Unfortunately, this growth is accompanied by a parallel rise in deceptive practices, particularly the proliferation of fake reviews, misleading consumers and harming genuine businesses. This study presents a system for identifying and classifying fake reviews utilizing techniques from Opinion Mining and Sentiment Analysis. The methodology involves three major stages: Initially, review texts are preprocessed to enhance textual quality through methods such as tokenization, stop-word removal, and stemming. Following preprocessing, sentiment analysis is performed using the VADER (Valence Aware Dictionary and sEntiment Reasoner) model to capture the emotional tone and polarity of reviews. In the final stage, machine learning and deep learning techniques-including Random Forest Classifier (RFC), Support Vector Classifier (SVC), and Long Short-Term Memory (LSTM)-are employed to distinguish fake reviews from genuine ones. Comprehensive experimentation conducted on an Amazon review dataset reveals superior performance by the LSTM model, achieving an impressive accuracy of 94.71%, compared to 83.97% and 88.03% by RFC and SVC, respectively. These results underscore the effectiveness of combining sentiment-driven features with advanced neural network architectures. By accurately identifying deceptive content, this approach not only protects consumers from fraudulent reviews but also helps businesses maintain credibility and integrity in a highly competitive digital marketplace.