Abstract
With the advent of Web 2.0 and the enormous amount of information on the internet today, new tools are being developed to help organize all this information. One genre of information is user reviews provided on opinion sharing websites such as Epinions.com. While the customer has a wealth of reviews at his fingertips to help make a decision on a product, sifting through all these reviews and deciding who to trust becomes a problem in and of itself. One solution is using a web of trust to predict for the user who he is likely to trust, providing the user with a larger array of opinions he could trust, as well as aide in several emerging applications including trust-based recommendation and viral marketing. Existing trust prediction studies have been limited to the use of trust propagation. Our approach, however, is to use a learning-based classification technique (specifically, C4.5 or SVM) to predict possible trust relationships in an opinion sharing website. We collected data from Epinions.com, using three categories, PC Laptop, PC Desktop and Mac Laptop. Our empirical analysis suggests that our proposed approach can effectively predict trust relationships with high accuracy rates, with C4.5 marginally outperforming SVM in accuracy and recall, while SVM achieved perfect precision.