Design And Implementation Of The Personalized Recommendation System Based On Education Cloud Platform

With the combination of cloud computing and education informatization,education cloud platform is favored by more and more education institutions and users for its convenient Resource sharing,efficient centralized management, wide area coverage and the advantages of the education resources in a wide range of high quality.This lead to the rapid growth of courseware resources stored in education cloud platforms.Vast amounts of courseware resources also bring out a problem of information overload when users study in the platform. it is hard to find the learning materials they are interested with only the traditional classification search and keyword retrieval method.For this,personalized push service appeared.Now the personalized recommendation service provided for users is mostly based on their evaluation information for certain items,if the users has no evaluation information,the system will not be able to make recommendations for them,and now most of the popular recommendation algorithms are based on the basic of stand-alone mode,the computing capacity and system scalability has great limitations, and they can’t really adapt to the massive data analysis and processing of the cloud platform.Therefore, this thesis designs and implements a personalized courseware recommendation system based on education cloud platform.This thesis aims at the courseware recommendation service for the education cloud platform users.Firstly,it summarizes the status quo of the researches in the education cloud platform and recommendation technology.Then, it gives a detailed in-depth research on the build process of the recommended system and the design process based on Map Reduce program.According to the actual situation of this issue,this thesis carries on a demand analysis for the system from the users’ interest model, courseware model,recommendation model and the high-performance mass data processing,proposes a overall architecture of the system, gives a detailed design on the interest model of the users and the object model of the course which are constructed by the Map Reduce computation model,and finished the latent factor course recommendation model based on topic,the course recommendation model based on the content and the popularity.Finally,this thesis tests the system from function to performance,and give a display for the important functions.The advancement of this thesis is mainly reflected in two aspects:(1)the whole construction of the recommendation system is based on the calculation model of Map Reduce,taking advantage of the cloud platform powerful data processing capability which can compute the recommend results of the users off line,makeing up for the poor calculation ability of the traditional stand-alone recommendation model and the long time in real-time recommendation.(2)This paper proposes a new BFIUF user interest model. The model draws on the idea of TFIDF and uses a special way to calculate the user’s preference on the coursewares,the model also introduces the time context information to reflect the influence of the past behaviour for today’s preference. Based on this model and the courseware model it can calculate out the user’s interest model about topic,which can better reflect the recent personalized theme preferences for the user and gives a better recommendation results.