Hybrid Algorithm for Precise Recommendation from Almost Infinite Set of Websites

Typically, recommendation systems are used for increasing users´ activity and spending. Most of them select a recommendation from a finite set of objects (e.g. News, products etc.) but in some cases the number of objects is so huge that standard approach based on similarity of objects will not work. The paper presents recommendation algorithms which work with an almost infinite set of websites by utilizing inherent characteristic of web pages and services delivered by search engines. Proposed algorithms were validated with the help of data from an Article Feedback Tool used on English Wikipedia for evaluating trustworthiness of articles. As the algorithms were developed to meet the special requirements of the Reconcile plug in, a community based websites´ credibility evaluation system, performance of proposed solutions have been measured not only as precision and recall but also in terms of increasing inflow of credibility assessments for new websites. Gathered results show that is possible to create effective recommendation systems, which offer better scalability and lesser computational complexity than standard recommendation systems.