An empirical study on detecting fake reviews using machine learning techniques

Reputation systems in E-commerce (EC) play a substantial role that allows various parties to achieve mutual benefits by establishing relationships. The reputation systems aim at helping consumers in deciding whether to negotiate with a given party. Many factors negatively influence the sight of the customers and the vendors in terms of the reputation system. For instance, lack of honesty or effort in providing the feedback reviews, by which users might create phantom feedback from fake reviews to support their reputation. Moreover, the opinions obtained from users can be classified into positive or negative which can be used by a consumer to select a product. In this paper, we study online movie reviews using Sentiment Analysis (SA) methods in order to detect fake reviews. Text classification and SA methods are applied on a real conducted dataset of movie reviews. Specifically, we compare four supervised machine learning algorithms: Naïve Bayes (NB), Support Vector Machine (SVM), K-Nearest Neighbors (KNN-IBK), and Decision Tree (DT-J48) for sentiment classification of reviews in two different situations without stopwords and with stopwords methods are employed. The measured results show that for both methods the SVM algorithm outperforms other algorithms, and it reaches the highest accuracy not only in text classification but also to detect fake reviews.