Modeling individuals and making recommendations using multiple social networks

Web-based platforms, such as social networks, review web-sites, and e-commerce web-sites, commonly use recommendation systems to serve their users. The common practice is to have each platform captures and maintains data related to its own users. Later the data is analyzed to produce user specific recommendations. We argue that recommendations could be enriched by considering data consolidated from multiple sources instead of limiting the analysis to data captured from a single source. Integrating data from multiple sources is analogous to watching the behavior and preferences of each user on multiple platforms instead of a limited one platform based vision. Motivated by this, we developed a recommendation framework which utilizes user specific data collected from multiple platforms. To the best of our knowledge, this is the first work aiming to make recommendations by consulting multiple social networks to produce a rich modeling of user behavior. For this purpose, we collected and anonymized a specific dataset that contains information from BlogCatalog, Twitter and Flickr web-sites. We implemented several different types of recommendation methodologies to observe their performances while using single versus multiple features from a single source versus multiple sources. The conducted experiments showed that using multiple features from multiple social networks produces a wider perspective of user behavior and preferences leading to improved recommendation outcome.