Detection of Fake Reviews in E-Commerce Platforms Related to Ayurvedic Cosmetics in Sri Lanka

The Ayurvedic cosmetics industry in Sri Lanka is grappling with the prevalence of fake reviews. To mitigate this problematic situation, this study first utilized Natural Language Processing (NLP) techniques to analyze English-language un-structured text reviews from popular e-commerce platforms. Then, it incorporated three primary approaches: a rule-based algorithm, a machine learning model using logistic regression, and a deep learning model based on Long Short-Term Memory (LSTM) to identify fake news. Initially, rules for the rule-based al-gorithm were crafted with guidance from existing literature. This research investigated various feature engineering techniques and model architectures to enhance prediction accuracy and assessed the model performance through precision, recall, and F1 score metrics. The rule-based approach yielded a 50% accuracy, which was sub-optimal. However, the Logistic Regression model and the LSTM model obtained 89% and 80% accuracies respectively, out of which the LR model shows the highest accuracy. These findings demonstrate the effectiveness of AI-based technologies in detecting fake reviews, thereby fostering a more transparent and ethical market for Ayurvedic cosmetics in Sri Lanka.