Open AccessMulti-Model Stacking Ensemble Learning for Dropout Prediction in MOOCs
Author(s) -
Ya Zhou,
Zhixiang Xu
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1607/1/012004
Subject(s) - dropout (neural networks) , computer science , machine learning , ensemble learning , artificial intelligence , preprocessor , data pre processing , set (abstract data type) , layer (electronics) , online learning , data set , data mining , multimedia , chemistry , organic chemistry , programming language
In recent years, with the rapid development of streaming media technology, massive open online courses (MOOC) have attracted unprecedented attention. Compared with the traditional offline teaching mode, the MOOC of online teaching has a higher degree of openness. Users can arbitrarily interrupt the study of a course according to their own interests, and have a higher dropout rate. The existing single machine learning model needs to be improved in terms of prediction accuracy. In this paper, a dropout prediction model based on multi-model stacking ensemble learning (MMSE) is proposed to further improve the accuracy of MOOC dropout prediction. The model mainly includes two parts: data preprocessing and model building. In the data preprocessing part, the student’s log records are used to design data features in weeks. In the model building part, a two-layer ensemble learning model is established. In the first layer, 5-fold cross-validation is used to train five different base classifiers. The second layer uses the XGBoost algorithm to combine the prediction results of the first layer to predict MOOC dropout. The experimental results on the KDDcup2015 real data set show that the MMSE model has achieved better results than the single model.