Extended topology based recommendation system for unidirectional social networks

The power and importance of social networks increases day by day and many social networks such as “Facebook, Twitter, Weibo and others” have more than millions of users who communicate with each other. This opportunity is triggering researchers to do studies on the social network area and supports them to do improvements for recommendation systems (RS). In this study, we propose an extension to the topology based and Friends of Friends (FoF) recommendation systems by taking into account the user actions. The proposed approach (PA) firstly classifying the data has been set into four classes and secondly an equation was computed by using the relationship of users. Our model utilizes not only the relationship of the users but also many actions and many mentions of the users to generate the recommendation to users. We evaluate the performance over precision-recall graphs and receiver operating characteristic (ROC) curves. PA extended topology based and FoF algorithm results compared with the other alternative RSs. The benchmarking study show that recommendations of extending topology based RS performs better than the extended FoF and other well-known algorithms such as graph-based and Conceptual Fuzzy Set based algorithms.