A Systematic Analysis of Machine Learning and Deep Learning Strategies for Identifying Fake Reviews

Fake reviews have become a growing concern in various online platforms, as they mislead consumers and damage the credibility of businesses. These fraudulent reviews are often used to manipulate public perception, making it essential to develop robust detection methods. Given the increasing complexity and sophistication of fake reviews, there is a pressing need to explore advanced techniques that can effectively address this issue. This paper offers a systematic literature review that compares traditional machine learning methods and deep learning approaches for identifying fake reviews in multiple domains. The review traces the evolution from machine learning methods like Support Vector Machines and Naive Bayes to deep learning approaches such as LSTM, CNN, and BERT. While machine learning methods have been effective in identifying basic patterns, deep learning models have demonstrated superior performance, particularly in detecting more complex and adaptive fraudulent reviews. These deep learning models have shown notable improvements in detecting complex and adaptive fraudulent reviews. The study also examines the progression in feature extraction techniques, highlighting a shift towards experimental and hybrid approaches that combine methods such as TF-IDF, Word Embedding, and attention mechanisms, tailored to specific domain challenges. By providing a comprehensive comparison of these methodologies and their application across various review domains, this review offers valuable insights into the strengths, limitations, and rapid advancements in fake review detection. This research can help improve the reliability of online review systems, providing businesses and consumers with more trustworthy information for decision-making.