predict student performance based on week-wise clickstream data

We develop the LSTM framework to predict risky students. We are using the public dataset OULAD https://analyse.kmi.open.ac.uk/open_dataset to verify the LSTM model. The prediction accuracy is improved over weeks.

We take the course FFF as the example. We first extract the daily click behavior for students admitted in the course FFF and then aggregate them into week-wise data. Then we run the LSTM model to predict student performance.