Hot Set Identification for Social Network Applications

Several operations of Web-based applications are optimized with respect to the set of resources that will receive the majority of requests in the near future, namely the hot set. Unfortunately, the existing algorithms for the hot set identification do not work well for the emerging social network applications, that are characterized by quite novel features with respect to the traditional Web: highly interactive user accesses, upload and download operations, short lifespan of the resources, social interactions among the members of the online communities. We propose and evaluate innovative combinations of predictive models and social-aware solutions for the identification of the hot set. Experimental results demonstrate that some of the considered algorithms improve the accuracy of the hot set identification up to 30% if compared to existing models, and they guarantee stable and robust results even in the context of social network applications characterized by high variability.