Research On Personalized Recommendation System Oriented Adaptive Algorithms And Implementation

With the rapid development of the Internet and information technology, the problem of information overloading and information amazing, which we are having been frustrated, has became more an more worse. And all these boost the flourishing development of personalized recommendation system. Present personalized recommendation technology has eased the pressure and cost for people finding their interested information. This paper shows us the characteristic and limitation of present recommendation algorithms, which are not able to satisfy the requirement of recommendation precise and scalability, through deep research and analysis into the existing recommendation technology. Further more, this paper summarize the framework of recommendation system and explains the bottleneck of present recommendation technology: imprecise expressing for user interest model and low performance of recommendation algorithms.In next step, this paper emphasizes the modeling of user interest model and recommendation algorithms, and proposes the improved methods. For user modeling, we introduce the methods and approach using machine learning theory. For recommendation algorithm, we put forward ICBD(Improved City-Block-Distance) for computing similarity between users and CFUPS(collaborative filtering based on users' partial similarity) for recommendation. Then we propose the ARA(adaptive recommendation algorithm) based on UAM interest model, ICBD and CFUPS, which focusing on target user's interest, adopting the idea of collaborative filtering, based on UAM interest model and given user personalized and interesting information through adaptively adjusting the recommendation parameter for each user.In the end, we validate the validity of our proposed algorithms through experiments, and compare with present recommendation algorithms. The results show us that our methods outperform the existing methods in recommendation precise and time cost, which contribute the excellent performance of personalized recommendation system.