Research On Collaborative Filtering Algorithm Based On Users’ Interest Changes

With the rapid development of computer technology and the Internet, information showsexponential growth trend constantly.On the one hand a lot of information providepeople abundant resources; on the other hand it also increased difficulty for people to find useful information quickly. Recommendation system appears in such an environment. Recommendation system based on the user’s explicit and implicit behavior, analyze the user’s interests and then provide for users items that they may be interested in. Not only recommended system can provide users with personalized service, but also can bring economic benefits for businesses. At present, whether in academia or applications, recommender systems have become the focus of researches and concerns.Collaborative filtering recommendation algorithm has good thoughts and recommended results, so it is used in e-commerce widely. Traditional collaborative filtering algorithm is based on the user’s past behavior data to predict the user’s current interests. This recommendation algorithm assumes that the user’s interests are not going to change, but in the fact they are often changing, so the traditional recommendation algorithm has certain irrationality. In this paper, the author put forward a new collaborative filtering recommendation algorithm.It analyzes the factors that influence the changes of user’s interests. The factors include time, the user’s degree of interest and the project similarity. The new collaborative filtering recommendation algorithm sets up different data weights for the above three factors, through a series of experiments analyze the effects of different weights on the accuracy of recommendation results, then select the appropriate data weighting to improve the quality of recommendations.In this paper, the experiment use MovieLens data sets, the MovieLens data sets is collected by the GroupLens Research project group and its data is film’s score that user marked. Through using different value oftime, user interest and project similarity scaling to determine its impact on the recommendation accuracy, select the appropriate value, and determine the optimal weighting scale of three factors to improve the quality of recommendations, and compared with other collaborative filtering algorithms of changing in user interests. The results show that the algorithm in this paper can effectively improve the accuracy of the results that recommended.