A Hybrid Recommendation Algorithm Based On Collaborative Filtering And Information Timeliness

Nowadays,the Internet is growing rapidly.People's dependence on the network makes the Internet everywhere,whether in life,work or study.This rapid develop-ment leads to all kinds of information inflated,the problem of information overload is increasing.And people are facing with the problem of how to find the information that they are really interested in or valuable in this huge information group,so there has been a personalized recommendation system.The core of recommendation system is recommendation algorithm,a good algorithm enables the user to generate a good user experience.So for users,a technology is needed to understand their preferences and to offer recommendations accurately.In view of the above demand,based on the traditional collaborative filtering algorithm based on project,this paper improves the algorithm for the problems of the data sparsity,scoring matrix simplification and recommendation accuracy.Firstly,considering that the relationship between different projects will affect user's choice,the influence matrix of project is constructed to adjust original user rating matrix,and it reduces the degree of matrix simplification.Then for the sparse matrix,the BP neural network prediction model is established to do the training of the score prediction,and the best prediction value is obtained to fill in the vacancy score.Secondly,the theory of information aging in information metrology is introduced to adjust the traditional Pears-on similarity formula by integrating user's interest and time-effectiveness information,and to tune the initial recommendation list combined with information aging,to filter out outdated information that interfere with your choices and get highly recommended results.Finally,a hybrid recommendation algorithm is proposed based on the improve-ment of the above two algorithms,which can better blend the advantages of the two alg-orithms,reduce parsity of matrix and accurately calculate the similarity between items.At last,it generates more timely information to meet the needs of users recommended to to the user,so that the quality of the recommended system will can be further enhanced..