International Journal of Computer Science Issues

More than a traditional journal...

In order to solve the problem that the recommendation quality is poor, which results from the fact that the traditional algorithm cannot dynamically generate recommendation for users and collect information, the paper put forward a friend recommendation algorithm which is based on the fragmentation of time and the transmission of interest. Firstly, the paper divides the users time into several time periods, and calculates the best push time. Secondly, the calculation method of user similarity is redefined based on the Ebbinghaus memory curve and the theory of interest transfer with taking the user browsing time as the standard. Finally, we make experiments by using Gooseeker crawler tools and MATLAB, and the results show that: the calculated similarity mean value and the stability have increased by 38.5% and 33.8%, 2% and 17.1% respectively comparing with the traditional collaborative filtering algorithm and collaborative filtering algorithm based on user similarity.