Design And Implementation Of Recommendation System Based On Services Around

With the rapid development of Internet technology, network information also showed explosive growth trend. In the face of these overloaded information, from which it is difficult to quickly find the information they need. Traditional search engines can only be based on keywords entered to all users the same kind of results, and it can not be given personalized result based on the user's own preferences. As the discovery of personalized recommendation technology, it becomes possible to quickly identify their needs when they face the large amount of information. Compare to traditional search engines, personalized recommendations can recommend the information they might be interested to the users based on the characteristics of each person's interest, and enable users to locate information they needed in minute.Recommend system not only enables users to quickly navigate to their own needs, for the information providers, it can make the information they released to stand out from the mass of information, which is concerned about by vast numbers of users. Feasibility and excellent results, which greatly promoted the application and development of recommendation system in commercial practice. Recommendation system now widely used in e-commerce recommendation, advertising recommendation and news recommendation, services around our lives are far more than these, but very few recommendation system for these services.In this paper, we design a recommendation system for the services around us based on th current recommendation technology.The major contributions are outlined as follows:1. A user-service interestingness calculation method based on user behavior data. User interestingness is draw by three kind of user behavior, which are interestingness based on user operation, interestingness based on visit frequence and interestingness based on residence time. This interestingness calculation method reflects the characteristics of the user's interest more accurately.2. A weighted hybrid recommendation policy which is applicable to the sence described herein is proposed. Hybrid recommendation algorithm gets the final result by weighting result produced by user-based collaborative and content-based recommendation, and adjust the two recommendation algorithm in real time according to the user feedback. This hybrid algorithm combines the advantages of two recommendation algorithms, solves the cold start problems of newly added service, and get a better recommendation result.3. A weight adjust method in accordance with the user feedback is proposed. Two recommendation results are respectively counted on the user click time, then adjust the two algorithm weight based on the count result in a certain period.4. A recommendation report module is added for user. This module allows the users to clear the modeling process and its own model in the system, which enhances the user experience.