Comparative Analysis of Machine Learning Algorithms for Fake Product Review Detection

The explosive rise in e-commerce has resulted in online product reviews becoming a fundamental component of consumer buying behaviour. The proliferation of faked or incentivised reviews for commercial purposes has undermined the credibility of online reviews as well. This article outlines the design and implementation of a Fake Product Review Detection System, which consists of two core elements that work together: Natural Language Processing (NLP) techniques for preprocessing the review text, coupled with Support Vector Machine (SVM) models for classifying the review. The NLP preprocessing pipeline performs tokenization, stop word removal, lemmatisation, and Term Frequency - Inverse Document Frequency (TF-IDF) Feature Extraction to transform the raw text data to high-dimensional numerical feature vectors. The SVM model was selected as the primary classification algorithm for its suitability to high-dimensional sparse text data. A comprehensive evaluation was conducted comparing the performance of the proposed system using SVM with five popular machine learning algorithms (Logistic Regression, Multinomial Naive Bayes, Random Forest, Decision Tree, and K-Nearest Neighbours) using accuracy, precision, recall, and F1 as evaluation metrics. The results from the experiments and graphical analysis indicate that SVM performed the best overall in terms of achieving the highest accuracy of 94.5% and the best balance of precision and recall, confirming the selection of SVM as the preferred classification method for detecting fake product reviews. The proposed Fake Product Review Detection System is an effective and scalable solution for identifying fake product reviews on e-commerce platforms.