Collaborative Filtering Of Personalized Recommendation Technology Research And Improvement

Began in the 90 s, the Internet affects the way people access to information,people access to resources from the print way gradually transferred to the Internet.This way can make us more convenient access to information to share and disseminate knowledge and live and work easy. The development of the Internet has brought the prosperity of information, especially mobile Internet in nowadays and everyone's work and life bundled together with Internet, how to find information they are interested in This is information overload problem[1]. Recommendation system as efficient way to find target information tool, you can quickly pick out userful message from the ocean of information resources. From the reality point of view, the demand for personalized recommendation system research is not only urgent but also has very important significance. Many scholars in the field of recommendation research has deep for many years and now has been familiar with recommender systems method are mostly based on collaborative filtering recommendation system. This paper from the Angle of the recommendation algorithm optimize the personalized collaborative filtering recommendation algorithm, make it more accurate and efficient.For personalized collaborative filtering recommendation algorithm,the most important part was to quantitative modeling similarity between users and item.Currently collaborative filtering recommendation algorithm was well know for many scholars, method based on user evaluation score to the item, according to Pearson correlation coefficient, cosine correlation coefficient or Jaccard related coefficient to calculate the similarity. Because traditional similarity calculation method only involves the user rating of item, ignore a lot of important part and it influential to recommend the result accuracy of data, the recommendation accuracy is still not very ideal. In order to make up for the shortage of the traditional recommendation algorithm of similarity calculation, this paper puts forward the calculation of similarity degree method is not only including user rating of the project, also add users judgment factors and age characteristics and other information. In the improved similarity degree calculation method, consider joining some traditional recommendation methods ignored data, in order to improve the recommendations of the recommendation system effectiveness and accuracy.When calculating the similarity and prediction score, this paper also considersthe degree of trust degree between users, improved collaborative filtering recommendation algorithm based on trust degree. Improved recommendation algorithm will trust degree method between users into the traditional recommendation algorithm, make predict score more accurate.In view of the article put forward the improvement of the two kinds of recommendation algorithm, design four groups of contrast experiment, respectively,the improved algorithm and recent the classical literature presents the results of the recommended method were compared. According to the comparison of evaluation index can be seen that, this article put forward the improvement of personalized collaborative filtering algorithm can effectively improve the recommendation accuracy.