Comparative Analysis of Machine Learning Models for Detecting Fake Reviews on Amazon
Patel, Krishna (2025)
2025
All rights reserved. This publication is copyrighted. You may download, display and print it for Your own personal use. Commercial use is prohibited.
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:amk-2025061022087
https://urn.fi/URN:NBN:fi:amk-2025061022087
Tiivistelmä
This research is concerned with the efficiency of machine learning models when it comes to detecting fake reviews on Amazon. Since or because of the rapid growth of e-commerce, online reviews have become important in determinations of consumer decisions. But the growing trend of fake reviews erodes the trust of the consumers and alters the behavior of the market. There are various evaluations of machine learning algorithms: Logistic Regression, Support Vector Machines (SVM), Random Forest, and Gradient Boosting among others to establish the most effective and reliable model for detection of fake reviews.
This study uses a publicly available dataset of Amazon product reviews identified as either genuine or fake, which contains text data as well as metadata on reviewers. Before training the models, methods of data preprocessing are applied, including the text cleaning, tokenization, and feature extraction. Performance evaluation is done based on the metrics of accuracy, precision, recall, and F1 score. Results reveal that the ensemble methods such as Random Forest, and Gradient Boosting classifiers perform better than other models in terms of recall as well as overall classification performance. The study identifies the issues with processing imbalanced datasets and points to its importance to pay attention to model transparency and interpretability. Lastly, the research offers recommendations to e-commerce platforms in order to increase the review credibility and safeguard consumer trust.
This study uses a publicly available dataset of Amazon product reviews identified as either genuine or fake, which contains text data as well as metadata on reviewers. Before training the models, methods of data preprocessing are applied, including the text cleaning, tokenization, and feature extraction. Performance evaluation is done based on the metrics of accuracy, precision, recall, and F1 score. Results reveal that the ensemble methods such as Random Forest, and Gradient Boosting classifiers perform better than other models in terms of recall as well as overall classification performance. The study identifies the issues with processing imbalanced datasets and points to its importance to pay attention to model transparency and interpretability. Lastly, the research offers recommendations to e-commerce platforms in order to increase the review credibility and safeguard consumer trust.