A New Model for Recommender Systems based on Data Sources Integration

摘要

The goal of Web recommender system is the process of selecting web pages shown to user based on his navigation patterns and interests. In this paper, a new model for recommender system is proposed to increase the accuracy of recommendations. In this model, some effective data sources are integrated to know the user interestingness. The sources used the proposed model are user spent times on pages, the count of each page views per session, user's location and data referred extracted from search engines. This data sources, combined through proposed model and then clustering operation is performed on it and recommendations are presented to the user through classification operation. In this paper some algorithms are proposed to extract user's interest from each of data sources. The approach is implemented as an experimental system, and its accuracy is evaluated based on F1 criterion.