ROBUST FAKE REVIEW DETECTION WITH XGBOOST AND EXPLAINABLE ARTIFICIAL INTELLIGENCE(SHAP)

Abstract

Fake reviews are emerging as a major issue on online commerce websites, misleading customer opinions and negatively affecting brand reputation. In this research, we propose an intelligent system for identifying spurious reviews based on machine learning algorithms, using the XGBoost algorithm as a specific example. The main goal of this research is to create a highly accurate and trustworthy model that can easily differentiate between authentic and bogus reviews. XGBoost was trained using a large dataset of user comments, in which text features were processed by natural language processing (NLP) approaches. Experimental results show that the proposed XGBoost model outperformed conventional machine learning algorithms. In addition, to provide model transparency and credibility, Explainable Artificial Intelligence (XAI) techniques like SHAP were incorporated for interpreting and visualizing the decision-making process of the model. This interpretability facilitates improved comprehension of feature importance and ethical AI practices in review authenticity detection.