The Research On Personalized Recommendation Technology Based On Collaborative Filtering In E-learning

With the rapid development of computer network technology, the amount of information on the network is various and increasing.Users often feel at a loss, it is more and more difficult to find their own demand from learning materialson the Internet.So that many scholars and experts start to study the problem and personalized recommendation system arises at the historic moment.Collaborative filtering recommendation technology is the most successfulin personalized recommendation system. Ittakes the similar sources to user with the way of collaboration between users or projects, based on the data that users give the mark to source.Due to the constant expansion of the information, there are also many problems in the collaborative filtering recommendation technology, such as sparse and accuracy of the score data.Because the scholars only look at the user rating and ignore the behavior habit of browsing the web, that is, some implicit information.For example,attributes of the website,duration and times to the web,click rate and so on. This paper mainly studies on sparse and accuracy of the data. The concrete content includes that:(1)In order to improve the similarity between users, the similarity measure based on weighted information entropy is proposed in this paper to increase the accuracy of the results. This method mainly introduces the information entropy to the field of collaborative filtering similarity measures.The method uses the user’s access time as the basis of implicit rating.Then calculate the entropy between users combining the information entropy formula and put the user’s score difference square and intersection as the weight. Because entropy values range is from 0 to infinity, this article uses gauss formula on the normalization,and controls the range between 0 and 1.So that will make the result concentrate and ease of analysis.(2) In order to solve the data sparse problem, this paper puts forward the collaborative filtering algorithm based on user implicit characteristics.The algorithm firstly extracts the basic information of the user who accesses the website from web log file. And classify users’ interests taking the type of web site as a basis.In each category, put the user’s access time and visits as implicit rating.According to the grading standard, the implicit rating translates intoexplicit score and forms rating matrix of the users.So don’t have to consider the projectswhich are not scored by user. There is not the sparseness problem of data.And then in the same way with other categories.Finally, calculate the nearest neighbors of target user withthe similarity measure method proposedin the paper, and forecast the score of unknown resources to recommend.