Abstract:

In recent years, massive open online courses (MOOCs) have received extensive attention. Due to the unreasonable learning styles of learners, their interest in learning is declining and some learning effect is not good, so the dropout rate of MOOCs is very high. In order to solve this problem, we automatically extract continuous features over a period of time from learners’ learning activity logs, and establish a MOOCs dropout prediction model by taking learners’ behavior features as independent variables. Experiments on the KDD Cup 2015 dataset show that the dropout rate prediction model in the  long shortterm memory in the convolutional neural network (CNN_LSTM) can help MOOCs curriculum teachers and designers track the learning states of course learners at different phases, and dynamically monitor the dropout behavior of different stages. The prediction accuracy of the model is high, so it can provide teachers with more reasonable guidance and advice on improving their teaching methods.

Key words: massive open online courses (MOOCs), dropout prediction, time series prediction, long short-term memory, convolutional neural network(CNN)