Construction of a Model for Predicting Students' Performance in a Programming Exercise Lecture

抄録

In recent years, the field of Learning Analytics has been experiencing a surge in interest. Learning Analytics involves work such as evaluating a learner's achievement degree and predicting his future ability, by data mining and analyzing learning history using a learning management system or e-portfolio. In this research, we predict the transition of a student's performance in a programming exercise lecture, using Learning Analytics. Using the prediction result, we aim to develop an application that finds students who are failing a class at an early stage and supports their learning. Specifically, we cluster students using the data from a comprehension test, build a regression model by applying multiple regression analysis to each cluster, then predict the week when the students will pass. We implemented the proposed model and evaluated its usefulness by leave-one-out cross validation.