MOOC Users' Behavior Analysis Based On K-means

With the popularization of "Internet + Education”,MOOC(Massive Open Online Course)develops quite well,which provides a suitable platform for people to accept online education.In China,practical research of MOOC is far earlier than theoretical research and lots of platforms are emerging recently.Users' educational background and learning motivation are diversely distributed.A large amount of behavior data have been produced in the process of learning.In order to explore rules hidden in the data and provide personalized learning environment and learning guidance for different user groups,it is of great importance to analyze and study behavior data.Clustering analysis is a common method used in data mining and is a typical application in unsupervised learning.It can be used not only on multivariate statistical analysis of classified data,but also on preprocessing of other algorithms.Through studying and summarizing multiple methods of user behavior analysis,the paper uses K-Means algorithm for clustering analysis.By optimizing K-Means and building grade prediction model,it successfully makes prediction on grade of MOOC users,and it also provides a visual display of grade prediction values.Main researches of this paper are as follows:(1)Based on behavior data of users,the paper draws some conclusions of MOOC through deeply analyzing and exploring basic information,user types and factors affecting grade of users.(2)The paper proposes a new K-Means feature selection algorithm through feature selection and optimizing initial clustering centers,and an equilibrium discriminant function is used to balance differences in clusters and between clusters.(3)The clustering centers obtained by K-Means feature selection algorithm are taken as centers of neural network,and then it sets up both input and output variables and parameters in prediction model.The paper constructs a prediction model with RBF neural network to get a more accurate grade prediction,which is updated dynamically.In order to verify efficiency of the algorithm,the paper designs a simulation experiment,which shows its performance by comparing K-Means feature selection algorithm and K-Means algorithm based on density.Meanwhile,it also demonstrates high accuracy on predicting grade of MOOC users through grade prediction model with a series of simulation experiments.Finally,the paper designs and constructs a prediction module by using grade prediction model,which can intuitively display users' prediction results and provide advice and guidance and warnings for users with low grade.