Fake Online Product Review Detection Using Supervised and Semi-Supervised Opinion Mining

Abstract

Review ratings on the online shopping platform play an immensely crucial part in shaping the buying preferences and choices of customers. However, with the rising proliferation of fraudulent as well as tampered online review ratings, there has been an upcoming serious threat against the credibility and reliability of the online review system. Therefore, this paper proposes an online fake review detection framework with the aid of opinion mining techniques with both supervised and semi-supervised learning models. For this, the proposed system specifically emphasizes the need to analyze the review ratings through the determination of crucial characteristic data extracted by the opinion mining technique based on the Text Frequency and Inverse Document Frequency (TF-IDF) value. Both the supervised models, namely Naive Bayes and Support Vector Machines (SVM), and the Expectation and Maximization (EM) technique with semi-supervised learning models have been used effectively as they can handle small-scale trained data effectively. Simulations and experiments have been carried out by using the actual Amazon online customer reviews dataset. From the outcomes and results, there has been demonstrated clear evidence that the classification model based on the SVM classifier outperforms and performs more accurately than the existing models.