Detecting Fake Reviews Through the Integration of Machine Learning and NLP Techniques

Abstract

The proliferation of online shopping platforms has brought about a surge in user-generated product reviews, making it susceptible to the infiltration of fake reviews. For these platforms to continue to be dependable and reliable, it is necessary to identify and mitigate the impact of fake reviews. When depending on reviews for the product present on various web pages and applications, the rate of false reviews has been growing in the e-commerce sector. The goal is to anticipate and identify fraudulent reviews on e-commerce sites, namely Amazon, by using a hybrid model that combines classic machine learning (ML) with natural language processing (NLP). The proposed hybrid approach that has been suggested seeks to improve detection accuracy and interpretability by utilizing the combined abilities of ML and NLP technologies. Our method combines the power of Bidirectional Encoder Representations from Transformers (BERT), a state-of-the-art language model and Bag of Words (BoW) with traditional ML algorithms like Random Forest and XGBoost. To enhance model performance, we employ stacking ensemble method with logistic regression as the meta learner. The machine learns complex linguistic patterns, contextual information, and cooperative behaviors suggestive of fake reviews through training on many datasets. The outcomes of the experimental evaluations demonstrate the effectiveness of the hybrid model, surpassing existing methods in accuracy and robustness. This research contributes to a reliable solution, poised to enhance the trustworthiness of online product reviews and fortify consumer decision-making processes which guarantees continued safety and assurance in online shopping environments