Title: Intelligent fake review identification e-commerce with multi-attention residual shrinkage-CNN

Authors: Madhavi Samala; Abdul Ahad

Addresses: Department of Artificial Intelligence, School of Engineering, Anurag University, Hyderabad, India ' Department of Artificial Intelligence, School of Engineering, Anurag University, Hyderabad, India

Abstract: With the rise of e-commerce, customer choices are heavily dictated by internet reviews, which also attract spammers who generate fake reviews to manipulate product reputation. Detection of such spam reviews is still a significant challenge since existing methods often suffer from poor feature representation, insufficient capacity to model contextual dependencies, and weak ability to adapt to spammer evolving patterns. To overcome these limitations, this research proposes the intelligent fake reviews detection framework based on Multitask Multi-Attention Residual Shrinkage-CNN in e-commerce websites (IFRD-MMARSCNN-ECW). The framework utilises the Computational Metaphor Processing Model (CMPM) for sophisticated pre-processing, and then the Adaptive Spatiotemporal Transformer (AST) for aspect-based extraction of features. MMARSCNN is used for classification, and optimisation using Black-Winged Kite Algorithm (BWKA) to enhance accuracy. Experimental testing on the Deceptive Opinion Spam Corpus Dataset shows that IFRD-MMARSCNN-ECW attains an accuracy of 99.1%, and ROC of 0.99, significantly outperforming existing models.

Keywords: adaptive spatiotemporal transformer; black-winged kite algorithm; computational metaphor processing model; e-commerce website; intelligent fake reviews detection.

DOI: 10.1504/IJWMC.2026.155350

International Journal of Wireless and Mobile Computing, 2026 Vol.31 No.1, pp.93 - 102

Received: 13 May 2025
Accepted: 14 Jan 2026
Published online: 30 Jul 2026 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article