—Online consumer reviews significantly influence purchase decisions, but the usefulness of reviews is impeded by fake reviews that portray distorted quality signals. Developing automated methods for detecting fake reviews is crucial yet challenging. This project aims to develop machine learning models that can distinguish between computer-generated (CG) fake reviews and authentic human-written reviews (OR). The models will be trained and evaluated on a dataset of 20,000 fake and 20,000 real reviews covering diverse product categories and review lengths. Both classical machine learning algorithms and state-of-the-art deep learning models will be explored. The project evaluation will focus on standard classification performance metrics like AUC, accuracy, precision and recall. This project will provide benchmarks and a fake review dataset to support future academic research and industrial applications in fake review detection. Index Terms—fake review detection, text classification, machine learning, deep learning
I. PROBLEM DEFINITION Online consumer reviews have revolutionized the way individuals make purchasing decisions, offering valuable insights into product quality, user experiences, and overall satisfaction. However, this democratization of opinion comes with its own set of challenges, chief among them being the proliferation of fake or deceptive reviews. These fabricated reviews aim to distort genuine product perceptions, mislead consumers, and manipulate purchasing behaviors for financial gain. The presence of fake reviews undermines consumer trust in the credibility of online reviews. Developing automated methods for fake review detection is therefore crucial yet challenging. This project aims to develop machine learning models that can accurately distinguish between fake deceptive reviewsgenerated by computers and authentic truthful reviews written by real consumers. It will investigate alternative detection approaches beyond existing methods like BERT, explore the synergy between classical and modern machine learning algorithms for improved accuracy, establish an evaluation framework using standard metrics, and provide benchmark datasets to advance future research in fake review detection. Successful models can help e-commerce and review hosting platforms detect and filter out fake reviews, improving the online shopping experience.