Unmasking Spurious Online Reviews Across Digital Platforms: A Comparative Performance Analysis Of Machine Learning And Deep Learning Paradigms

Abstract

The widespread dissemination of spurious online reviews poses significant challenges to consumer trust and market transparency across digital commerce platforms. This study presents a comparative performance analysis of machine learning and deep learning techniques for detecting fraudulent reviews. A comprehensive dataset comprising authentic and deceptive reviews from multiple e-commerce and service platforms was compiled to evaluate the efficacy of various approaches. Traditional machine learning classifiers—including Support Vector Machines, Random Forests, and Gradient Boosting—were benchmarked against advanced deep learning models such as Convolutional Neural Networks (CNNs), Long Short-Term Memory (LSTM) networks, and Bidirectional LSTM architectures. Experimental results demonstrate that deep learning models, particularly BiLSTM networks, outperform traditional classifiers in terms of detection accuracy, precision, and recall while exhibiting robust generalization across platforms. Additionally, the study highlights critical trade-offs between interpretability and predictive performance. These findings underscore the potential of deep learning frameworks to strengthen fraud mitigation strategies and improve content authenticity verification in digital marketplaces.