A Comparative Review of Fake Review Detection Techniques Using Machine Learning and Transformer Models

– Online review systems play a crucial role in shaping consumer decisions in modern e-commerce environments. However, the increasing prevalence of deceptive or fake reviews has raised serious concerns regarding the reliability of such platforms. Over the years, a wide range of techniques have been proposed to address this issue, spanning traditional machine learning methods, deep learning architectures, and transformer-based models. This paper presents a comprehensive comparative review of major approaches used for fake review detection. The analysis covers feature-based classification methods, network-oriented models, neural architectures such as CNN and LSTM, and advanced transformer models including BERT and RoBERTa. Each approach is evaluated in terms of model design, dataset usage, performance metrics, advantages, and limitations. The study highlights a clear shift from manually engineered feature-based systems to context-aware deep learning frameworks. Although recent transformer-based models demonstrate strong performance, challenges such as cross-domain adaptability, computational complexity, interpretability, and evolving spam tactics remain unresolved. This review aims to provide a structured understanding of existing techniques and identify future research directions for building efficient and scalable fake review detection systems.