Enhancement of Fake Reviews Classification Using Deep Learning Hybrid Models

Abstract

Fake reviews can significantly affect consumer behaviour and a company's reputation, making it crucial to identify them in today's digital age. This study uses deep learning to detect fake reviews. The proposed approach captures review text contextual data using five deep learning designs: LSTM, Bidirectional LSTM, multi-dense LSTM, GRU, and Bidirectional GRU. The models were trained on a large annotated dataset of reviews and assessed using precision, recall, accuracy, and F1-score. This investigation's primary objective is to establish whether or not a review can be trusted as genuine. According to the findings, deep learning models perform noticeably better than more traditional machine learning-based approaches and provide a reliable solution for the detection of fake reviews. The proposed research proves that Bidirectional LSTM generates 99.9 accuracy results on training data and 85.59 for validation data, among other models. For future research, we will focus on developing text enriches columns by adding a polarity feature to the existing dataset and ensemble modelling for novelty purposes. This study helps researchers to take potential benefits in various application domains, including e-commerce and consumer feedback platforms, where accurately detecting fake reviews is essential for maintaining trust and transparency.