Research And Application Of Health Knowledge Recommendation System Based On Collaborative Filtering

With the rapid rise of today’s internet applications, users conveniently receive a large amount of information from the various application channels and users’ need for information has been greatly satisfied. Information overload problem caused by the surge in the number of internet information has led to a reduction in the utilization of internet information. Users are increasingly difficult to find the information they want quickly, and the recommendation system has been proposed to solve this problem. Collaborative filtering recommendation algorithm is the most widely used recommendation algorithm in the recommendation industry now. The recommendation algorithm can analyze the characteristics of the user’s interest according to the behavioral data that user has produced in the system, and then generate a personalized recommendation for the user.This thesis is based on the project requirements of "Intelligent Medical" in my laboratory and collaborative filtering recommendation technology will be applied to the field of health knowledge. Each user has different own health status and different attention on health categories so that each user’s need and interest of health knowledge are also different. Therefore, it is very necessary and meaningful to find out the health knowledge that users like from a large number of health knowledge. Based on the above project background and technical research, the following work has been carried out in this thesis:(1) Understand some relevant theoretical knowledge of the recommendation system and recommendation algorithm in detail; focus on the study of collaborative filtering algorithm working principle, components and some related realization modules of an open source framework of collaborative filtering algorithm—mahout; analyze the function model and the existing problems of health knowledge recommendation system based on collaborative filtering, and propose the two important work of this thesis.(2) Research the key problems of traditional collaborative filtering recommendation algorithm, propose some improved method for cold start and sparse data; focus on the introduction of the user interest feature model, health knowledge attribute model and user’s interest degree model of health knowledge, an improved collaborative filtering algorithm based on user’s interest degree is proposed, which is the innovation of this thesis. It analyzes the characteristic data of user interest and attribute data of health knowledge, combines the historical data of user behavior to comprehensively analyze the user’s interest degree of health knowledge, improves defects of the traditional collaborative filtering algorithm and obtains higher quality recommendation effect.(3) Elaborate the design goal, requirement analysis and design for the system architecture. The system consists of four modules: user interface module, log collection module, recommendation engine module and data storage module. Log collection module obtains user’s behavior records generated by the user interface module, and provides initial data source for the recommended engine module; recommendation engine module extracts the log record, uses the recommendation process combining offline data calculation and online data processing, real-time responses to user’s demand, and recommends the health knowledge of interest to the user.(4) Combined with Mahout, the health knowledge recommendation system and experiments of improved algorithm have been completed, and the experimental results of the improved algorithm have been evaluated by using MAE and coverage evaluation criteria.