A Research On Modeling And Predicting Dropout Rate Of MOOC Learners

With the development of Internet,online education with various forms become more and more popular.As a form of online education,MOOC has been accepted by more and more people for its low cost of learning.But it is faced with a very important problem:the high drop out rate which is very harmful for the development of MOOC.In this paper,a background dataset about 39 courses offered by MOOC website is used to model and analyze the behaviors of the learners.Based on this,the paper eventually build a model to predict the drop-out rate of the MOOC learners.The data set used in this paper has been marked by a good correspondence with the predicted results(drop-out or not).So the drop-out rate modeling can be considered as a typical classification problem,which is fit with classification theory.Feature selection is directly related to the performance of the model.In this thesis,three kind of features are proposed which include description and statistical features.These features can reflect the dataset in many ways.After feature selection,some classification models(logistic regression,support vector machine,random forest,AdaBoost,gradient boosting tree)is used to model the drop-out rate.Connecting with the practical situation,the paper analyzes the shortage of each model:a single model cannot dig the all information of the dataset.In view of the shortcomings of the classification model,the concept of the classifier weight is put forward.Then,the paper proposes a multiple classifier weighted model based on the concept to improve the performance of the model.Multiple classifier weighted model is a combination of multiple classifiers.Then,the paper gives a multiple classifier "and" model,with which,the paper can get a "trusted set".The"trusted set" can be used by the multiple classifier weighted model to make better prediction.