A Weighted Hybrid Recommendation Algorithm Based On Kalman Filter

With the rapid development of technology and application,the information overload problem has already been inevitable under ubiquitous networks in recent years.How to effectively help users to search and filter the information has become the hot topic for many scholars and engineers.Personalized recommendation techniques emerged in response to the right moment.The personalized recommendation is such a technique to help users quickly find useful information.It is not necessary for people to take a clear demand.Through the analysis of the history of user behavior,the technique can provide the recommendations which meet their needs and interests of information for users.Researchers have come up with a variety of recommendation technologies,but because each of them has its limitation and shortcomings,it is difficult to meet the different needs of many users.In order to obtain better recommendation,hybrid personalized recommendation algorithms are usually used in the practical application environment.First,this paper discusses the related concepts and applications of personalized recommendation algorithms,and takes an analysis of the theories of the content-based recommendation algorithm,the collaborative filtering recommendation and the hybrid personalized recommendation algorithm.Hybrid recommender systems are that combine two or more of recommendation methods together to produce its output.They can choose the different strategy according to the particular case,and avoid the defects of the recommendation methods to improve the recommendation performance and provide more valuable recommendations.Furthermore this paper introduces and studies the Kalman filter model and processes,and illustrates its characteristic and adaptability.Discrete kalman filter algorithm is a quadratic linear estimate algorithm,which can make an optimal estimation for the control of discrete dynamic system state.Then,we proposed a novel weighted hybrid recommendation algorithm based on the Kalman Filter model which not only can predict the weights,but also optimize and revise the weights with the help of feedback by taking advantage of the optimum regression feature of Kalman Filter.The novel algorithm implements the optimal estimation of the dynamic weights in the process.Finally,in order to verify the performance of the algorithm proposed in this paper,we have made some tests on the datasets MovieLens and Githubs,and made an analysis of the experiment results.The results show that the algorithm can effectively improve the accuracy and recall rate of recommendation system,and improve the recommendation quality.