A Personalized Recommendation System Of Educational Resources Based On Odds Ratio Matching Degree

With the rapid development of Internet,the amount of data in the network has a great growth.In order to enable users to find suitable resources in the massive data,personalized recommendation technologies are introduced.The field of education is no exception.In recent years,educational resources become more and more,and many educational resources platforms are built.How to recommend educational resources to students which have different character- istics and needs has become an important part of educational resources platform.Also, from the perspective of information processing,computing power of CPU fails to meet the demand of massive educational resources data processing.GPU(Graphic Processing Unit),which can increase the computational speed greatly, has gradually been used in general purpose hererogeneous parallel computing because of its large number of computing units.In this paper,several common personalized recommendation technologies, GPU,general purpose parallel computing model and Open CL heterogeneous platform are outlined;Then,we establish model of student and model of educational resource,and present a personalized reco- mmendation scheme,which includes data preprocessing,matching of basic information, matc- hing based on odds ratio and so on;After that,we analyze the personalized recommendation scheme achieved in Open CL platform in detail,and the parallel modes of calculation of odds ratio, matching of basic information, matching based on odds ratio are designed and research- ed respectively;Finally,we set up the educational resources personalized recommendation system and achieve the personalized recommendation service.In order to solve performance problems of processing massive educational resources,a parallel algorithm has been achieved in Open CL platform in the paper,and we have deeply analyzed its acceleration performance.The proposed educational resources personalized recommendation mechanism can provi- de personalized recommendation service for Web educational resources platforms,and GPU- based personalized recommendation parallel algorithm designed in this paper has a faster execution speed than serial algorithm,which can improve the efficiency effectively.