Author(s): Dominique Cuadra, Angel Perez, Juan Castillo-Gomez
The proliferation of online reviews has led to an increased risk of fraud, as some reviewers may be incentivized to post fake or biased reviews. In this paper, we examine several machine learning-based methods for the automated detection of fraudulent reviews. We use a gold-standard dataset of reviews from a popular online platform, and use natural language processing techniques to extract features from the text of the reviews. We experiment with several different feature sets. We then train a machine learning model to classify reviews as either fraudulent or not fraudulent. We evaluate the performance of our model using several metrics, such as precision, recall, and F1 score. Our results show that our method is able to achieve reasonable accuracy in detecting fraudulent reviews but is not able to scale at high precision. We also discuss potential limitations and future directions for improving the performance of our model. Overall, our study demonstrates the feasibility of using machine learning for the automated detection of fraudulent reviews, and highlights the importance of continued research in this area.