Identifying Indicators of Fake Reviews Based on Spammer's Behavior Features
Pan Liu, Zhenning “Jimmy” Xu, Jun Ai, Fei Wang · 2017
In the enterprise marketing process, on-line data plays more and more important role. As a type of data, spam (or fake) reviews of the products, however, have been seriously affecting the reliability of both decision making and data analysis of the enterprise. To detect spam reviews, the paper presents a set of opinion spam detection's identification indicators based on behavior features of the spammer. Two algorithms are then proposed to recognize similar reviews and relevant reviews from all reviews. Compared to the traditional algorithm, our review identification algorithm achieves shorter execution time. More importantly, the proposed algorithm for recognizing relevant reviews can be used to analyze the relevancy between the review content and the given review topic by using the automatic word segmentation technique. The Experimental results show that the number of fake reviews by our algorithms is higher than that of the traditional algorithm. Moreover, we found that about 46% of the mobile phone reviews on the Amazon website were irrelevant to the product's topic, and 54.7% of the reviews were similar to other reviews.