Research And Design Of Recommendation Algorithm With Item Attributes And User Interests

With the rapid development of Internet and information technology,we are coming to the time of information explosion.As the main solution to information overload,collaborative filtering is firstly put forward in the 1990 s,becoming the crucial topic soon in the field of man-machine interaction,machine learning and information retrieval.Moreover it has achieved a great success in practical applications represented by e-commerce sites.This article will do some work about collaborative filtering and recommendation technology.After comparing the main technologies of recommender system,this paper will mainly focus on item based collaborative filtering algorithm.The traditional algorithm has three deficiencies.Firstly it ignores the value of item attributes in item similarity measure.Secondly it doesn't consider time factor of user ratings,therefore the algorithm can't reflect the change of user interests through time.Thirdly it's unable to introduce the new item to users.According to the above shortcomings,this paper proposes a new algorithm considering item attributes and user interests.The new algorithm takes item attributes into the consideration of item similarity measure,what's more,adding the time factor of user ratings to reflect change of user interests through time,expecting to enhance recommender results.To improve the advance of the new algorithm,this paper design several experiments based on the open MovieLens dataset.According to results of experiments,the new algorithm can reflect the change of user interests and alleviate the effect of data sparsity.