Strategy Research And Implementation On Improved Recommendation Algorithm Based On Information Timeliness

Traditional recommendation algorithms usually only focus on the matching degree between items and user interest. The items that have higher matching degree would have higher value for user. The value of the item is regard as the probability of clicking the item. But in the recommender system there may be a lot of old items which may be very match with a user’s interest. So those items might be recommended to the user by traditional recommendation algorithms. But these old items are meaningless for users, which are not use to improve user experience.In order to solve the above problem, this paper combines with information obsolescence theory and proposes high timeliness recommender algorithm. We analyze users’click history to construct the evaluation model of timeliness. It can predict click probabilities of items at present. We combine the evaluation model with item-based collaborative filtering and synthetically consider interest of users and the timeliness of items to find the nearest neighbors. And then we can make high timeliness recommendations. At the same time, we propose to filter the recommendations with the timeliness of information to improve the timeliness of recommendations. The experimental results show that the method can improve the timeliness of recommendations comparing with traditional recommendation algorithms.The existing interest-drift models can not dig out the drifting law of user’s interest based on the user’s limited behaviors. To solve this problem, this paper extends the timeliness model. So it can get the changes of users’interest. We argue that the changing characteristics of utilization ratio of certain resource demonstrate the changing rule of users’interest. To dig out the drifting law of users’interest to a certain resource, we use the method of the timeliness model to construct a new model, resource-timeliness quantitative model, which can measure the changes of user’s interest over time. The experimental results show that the model can reflect the change rule of users’interest, and accurately obtain change rate of users’interest.The first chapter of this article introduces the background, content and significance of algorithm improvement research based on timeliness of information. The second chapter briefly illustrates the related concept, input and output algorithm. The third chapter presents information timeliness of the related concepts and how to establish information timeliness model. The fourth chapter presents the application of information to the timeliness recommended optimization method. The fifth chapter presents the application of the algorithm to improve timeliness parameters recommended method. The sixth chapter introduces the possibility of the application for the information timeliness in the interested offset model. The seventh Chapter summarizes on the work ahead of the next step and job prospect.