Abstract—This project attempts to conduct sentiment analysis of short and long Amazon reviews and report their effects on the supervised learning Support Vector Machines (SVM) model, to bridge for fake reviews classification. Firstly, the SVM model was evaluated by comparing its performance against Naive Bayes, Logistic Regression, and Random Forest models and proved to be superior (second assumption) based on the accuracy (70%), precision (63%), recall (70%), and F1-score (62%). Hyperparameter tuning improved the SVM model for sentiment analysis (accuracy of 93%), then altering the review length affected the model’s performance, which validated that review length affects the classifier (first assumption). Secondly, conducted fake reviews classification on the fake reviews’ dataset yielded 88% accuracy, while the merged subsets of the two datasets yielded 84% accuracy.
 
Keywords—fake reviews detection, sentiment analysis, natural language processing, Machine Learning (ML) supervised learning


Copyright © 2024 by the authors. This is an open access article distributed under the Creative Commons Attribution License (CC BY-NC-ND 4.0), which permits use, distribution and reproduction in any medium, provided that the article is properly cited, the use is non-commercial and no modifications or adaptations are made.