Research On Collaboratice Filtering Recommendation Algorithm Based On Item Similarity And User Requirements

The rapid development of computer and the network communication technology is the foundation of the fifth information technology revolution,with the arrival of the information age has brought unprecedented changes and experiences to people's lives.But there is no doubt that the development and the problems are mutually related.The problem of information overload is becoming more and more serious.Meanwhile,the demand for people is changing faster and faster.With the continuous development of information technology,homogenization of services has been difficult to adapt to changes in people's needs,personalized service has become the mainstream of development,that not only improve the user's shopping experience also bring more benefits for businesses.This paper describes the development of recommendation system model.Also introduces five kinds of mainstream recommendation algorithms,such as collaborative filtering,content based,latent semantic,association rules and scene information,collaborative filtering algorithm is the focus.At the same time,this paper analyzes the causes of common problems and the related solutions.As the rapid expansion of user and item data,brings the data sparseness problem inevitably,the collaborative filtering algorithm based on user have bring the inevitable problem of data sparseness greatly reduces the algorithm accuracy of recommendation score and effect.To solve those problem,this paper studies the relationship between the items,and proposes an improved user based collaborative filtering algorithm "ISUCF".The algorithm combines items attribute similarity and relevance similarity to fill score matrix.This paper did a detailed description of the improvement ideas and derivation of the algorithm,and through experimental comparative analysis of "ISUCF" algorithm and the traditional algorithm,the results show it can effectively improve the recommendation accuracy.Although the ISUCF algorithm in this paper is proved to be able to solve the problem of data sparseness,and to a certain extent can solve the data sparseness problem,but the impact on the accuracy of the recommendation result and quality induced by how to dynamically track the user needs and the poor timeliness must not be ignored.And in this fast changing time,the factors that affecting people's shopping needs are constantly changing,only dynamic analyze and tracking user needs is the right way to achieve a better results.So,this paper proposes a time delay and time window model to simulate the change of user requirements algorithm'"ITICF" that based on the "ISUCF" algorithm,the experiment results show that this algorithm can improve the recommendation accuracy and recommendation quality.