Analysis Of User Behavior Data Based On MOOC

Nowadays,the large-scale open-ended online course,MOOC,has gradually emerged at home and abroad.The elite schools offer a large number of free courses on the online education platform,providing more students with independent courses and other related courses.may.At present,the mainstream of MOOC includes Coursera,Udacity,edX three major platforms.MOOC has become the main way of learning in school and so on.MOOC's rapid development,as well as the massive data generated,also makes data analysis increasingly facing challenges.Data analysis technology with each passing day,in different platforms have different applications,but seemingly excellent data analysis technology,but in different platforms have different effects,and more common a shortcomings in the user's cold start on the issue,that is,When a new user into the platform,because the lack of data,it is often difficult for such users to make the right and reasonable analysis.Online education platform in recent years,the amount of data and the type of increasingly large,and the user's behavior data,compared to the previous areas through static data analysis,have better results.Therefore,how to use these data efficiently to help users get a better learning experience is facing great challenges.Most of the researches on data analysis are related to data analysis.The existing data analysis techniques in industry are more versatile,but most of them are simple classification clustering,single classification or clustering algorithm effect.It is not ideal,can not reach the user to provide guidance on how to quickly and easily operate the use of existing classification clustering algorithm for data analysis,is the mainstream of large commercial data development.In this paper,an improved method is proposed for TrAdaboost,an instance-based migration learning algorithm,to adapt to the different data sets for different non-identical experimental data and training data.After the data set is preprocessed,the algorithm can be used to alleviate the cold start problem of the new or popular course in the online education platform to a certain extent.For the integration of classification and clustering algorithms,Boosting algorithm for the integration of this classifier optimization,provides a very good solution.In this paper,we propose a new idea for multi-tag AdaBoost series algorithm,which can reduce the learning error rate of the algorithm.The main idea is to modify the sample distribution strategy of the algorithm to break the uniformity of the sample distribution in the existing AdaBoost algorithm,so that the upper bound estimate of the learning error can be effectively reduced in the process of adding a weak classifier The performance of multi-tag AdaBoost algorithm is improved.