Intelligent Review Classification for Trustworthy E-Commerce: Improving Product Recommendations with Sentiment Analysis and Fraud Detection

Internet business has become a fast-growing industry, and though online buying has not always kept pace with this expansion in the past, its potential is finally being realized. Sentiment analysis has been a prominent topic of study within text mining. Opinion and sentiment extraction from natural language is a challenging task. Yet, sentiment analysis is a good solution, offering useful insights for decision-making in many fields. There are many techniques for sentiment detection, each impacting the quality of the results. This paper proposes to evaluate consumers' sentiments on e-shopping services through reviews, ratings, and emoticons. The main purpose is to identify fake reviews based on user and product IDs to aid in making a well-informed purchase decision and to find the best products. For this, we use a hybrid learning algorithm that analyzes diverse feedback about services. A word-level score calculation is done using a text mining algorithm, and the sentiments are then classified as negative, positive, or neutral. This method helps identify fraudulent reviews on the website. Finally, we aim to predict fraudulent reviews entered by users and ensure that only verified purchaser reviews are taken into account, while duplicates are also checked using user and booking IDs. Genuine reviews are given higher priority for product recommendations.