AI-Framework to Detect eCommerce Fake Reviews: A Hybrid Neural Network Machine Learning Model

The popularity of Web 4.0 technology has catalysed a substantial increase in online commerce, with eCommerce platforms emerging as primary venues for consumer transactions. Integral to the eCommerce experience are customer reviews, which serve as invaluable sources of feedback and guidance for potential buyers. Indeed, the proliferation of fraudulent reviews, known as fake reviews, threatens the integrity of this feedback mechanism and undermines consumer trust. This research paper focuses on addressing this challenge by proposing a hybrid neural network (HNN) approach tailored to predict and identify fake reviews in eCommerce environments. Drawing from an extensive review of existing literature and research studies, this paper develops a comprehensive framework for modelling the detection of fake reviews using HNN. Employing Agile methodology, the research outlines the development process and evaluates various neural network (NN) machine learning algorithms using diverse datasets. Key preprocessing techniques, including stop word removal, punctuation deletion, series padding, and tokenization, are explored to ensure the consistency and effectiveness of input sequences. Experimental results demonstrated the efficacy of the proposed HNN model with a 99% average training accuracy and a 64% average testing accuracy. Also, the study highlights the potential of convolutional neural network (CNN) methods for feature extraction and dimensionality reduction in detecting fake reviews. This research paper contributes to the ongoing discourse on combating fraudulent activities in the eCommerce domain by offering practical insights and procedures for stakeholders to safeguard the integrity of online reviews and enhance consumer trust in digital marketplaces.