AMAZON PRODUCT FAKE REVIEW IDENTIFICATION USING ASPECT BASED MULTI‑LABEL SENTIMENT ANALYSIS

In recent years, online reviews have become increasingly important for both consumers and businesses, as they greatly influence purchasing decisions. However, the presence of fake reviews has emerged as a significant challenge, as they manipulate consumer perceptions and undermine trust in e-commerce platforms. This study aims to address the problem of fake review identification on Amazon using aspect-based multi-label sentiment analysis. By leveraging the fine-grained nature of aspect-based sentiment analysis, our approach captures multiple sentiment polarities associated with different aspects of a product, providing a deeper understanding of review content. We propose a novel framework that combines deep learning techniques, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), with aspect extraction and sentiment classification methods to identify fake reviews. The model is trained on a large dataset of genuine and fake Amazon reviews, which are manually annotated with aspect labels and sentiment polarities. By detecting inconsistencies in sentiment polarities for various aspects of a product, our method is able to identify suspicious reviews with high accuracy. Experimental results demonstrate that our aspect-based multi-label sentiment analysis approach outperforms traditional machine learning methods and other DL techniques in fake review detection. Furthermore, the proposed method is robust against various types of fake reviews, including those generated by review spammers and opinion spam groups. Our findings suggest that the use of aspect-based multi-label sentiment analysis can significantly improve the accuracy and reliability of fake review identification, ultimately fostering trust and transparency in e-commerce platforms