Recommended System Development And Research Based On Hybrid Filtering

With the advancement of the global process of information and popularity of the Internet, Internet gradually become the main way to get new information, Web resources on the explosive growth trend gradually, the user is more difficult to extract useful information, some users are not concerned information can easily become isolated islands of information.Thus, the emergence of personalized recommendation systems effectively addresses these issues, personalized recommendation system analysis and interest in the user purchasing behavior characteristics of potential interest to the user recommendation and satisfactory information to enhance the user the right decision-making. The prevalence of current e-commerce recommendation system recommended by poor quality, inefficiency and recommended the low degree of automation.This paper describes the main personalized recommendation technology, which mainly describes the content-based filtering technology based on collaborative filtering technology, by comparing their advantages and disadvantages, the use of hybrid technology combines two recommended methods recommended, proposed based on The recommended framework for mixed models. In the recommendation system, the interest model, and user clustering has been a hot discussion is that people, and this particularly concerned about the user interest model. The classic study of existing hybrid filters improves the problem, find their shortcomings, an improved algorithm and model validation. Analysis of hybrid filtering algorithms recommended by the quality of the sparsity of the problem and the recommendation of the integrity of user satisfaction problem, hybrid filtering algorithm is proposed to improve the new hybrid filter Algorithm. Improved similarity measurement method is recommended to improve precision. And on this basis, search for the target user's nearest neighbors, thereby reducing the range of nearest neighbor search and generate recommended result. . Traditional algorithm because of the new users to add lead to the increase in the amount of data, resulting in low efficiency of online data processing algorithm is proposed by using the user's method of clustering users with similar interests will be divided into the same clustering, and propose a collaborative filtering algorithm based, other methods based hybrid filtering algorithms. Web recommendation system of complex distributed systems problems. Build a model of distributed multi-site, loose, heterogeneous information integration, Web aggregation presents a complex distributed recommendation algorithm. Finally, we designed and implemented a complex mixed-mode-based personalized e-commerce recommendation prototype system. The system uses J2EE technology, MVC pattern, SSH architecture modeling techniques. The system is a recommendation system based on multi-model engine model, which provides personalized recommendations, new projects recommended, recommended, and many other popular recommendation, and a combination of user ratings and keyword search functionality, helping users to access different levels from individual Information, to further enhance the user experience.