Prediction Of Information And Content Recommendation In Social Leraning Network

With the development of Internet technology, the society has entered the era of information explosion, and people's increasing demand for knowledge has been beyond the scope of traditional education mode.How to make full use of information resources in the information world,and create a personalized learning environment to meet the needs of learning at anytime and anywhere, has become a top priority. With online education come across, more and more people are attracted. Thesis name the composition of online education including characters, courses,contents, platforms as social learning network(SLN). Social learning network can meet people's demand for learning in the era of information explosion, but because it is in the initial stage of development, there are a lot of deficiencies such as high drop-off rate, unreasonable distribution of education resource and so on. Nowdays the researches on the problems may not solve them in a proper ways. This paper do some researches on predicting action of drop-off and recommendations for users:1) Thesis focus on the video of SLN to solve the high drop-off rate .The action of click will be recorded and we take a deep understand ,analysis the structure,extract the important features. finally we make a useful transformation from stream to structure data, which is easier for us to deal with and compute.2) A learning weight naive bayes take place of the classical method naive bayes by assigning the weight to all features On predicting. Our innovation is to make the loss function and solve the set of weight values for fitting the data better through a learning methods which make accuracy raised from 60% to 72%. And the importance of all features is shown as weight in our results.3) Most of traditional recommendation use the feature of course or video the user has chosen, and it may not reflect the interests in a comprehensive way. In order to recommend the course better for users,we focus on the forums data not only the list of courses . we deem that the express of forums will show the hobbies of users .Thus this paper will take machine learning and TF-IDF for extracting the keywords of courses and make a connection between the key words nodes and characters nodes for analyzing the interests of the users ,which make the recommendation tally.Thesis have done some researches on the prediction of drop-off and course recommendation. Although our works can not solve the problems of high drop-off rate and the unbalanced distribution of courses completely, some kinds of method for finding the problems in time are conducted, which make a good sense.