Smart Detection of Deceptive Reviews Using Hybrid Machine Learning Techniques

Abstract

In the age of online shopping, customer reviews greatly influence buying decisions. However, the rise of fake or manipulated reviews misleads consumers and undermines trust. This project proposes a semi-supervised machine learning approach to detect deceptive online reviews effectively. By leveraging both labeled and unlabeled data, classifiers such as Naïve Bayes, Logistic Regression, SVM, and Decision Trees are trained and tested on review datasets. The model extracts features like review length, sentiment polarity, and word frequency to enhance prediction accuracy. Results show significant improvement in identifying fake reviews using the proposed hybrid model. This system can benefit e-commerce platforms by promoting genuine user feedback, helping businesses maintain credibility, and aiding consumers in making informed decisions.