The prediction of programming performance using student profiles

Guohua Shen, Sien Yang, Zhiqiu Huang, Yaoshen Yu, Xin Li

Abstract: Due to the growing demand for information technology skills, programming education has received increasing attention. Predicting students’ programming performance helps teachers realize their teaching effect and students’ learning status in time to provide support for students. However, few of the existing researches have taken the code that students wrote into consideration. In fact, code is informative and contains lots of attributes. Student programming performance can be better understood and predicted by adding code information into student profiles. This paper proposed a student profiles model to describe students’ characteristics, which contains the code information and then was used as the input of a deep neural network to predict the programming performance. By comparing different machine learning techniques and different combinations of dimensions of student profiles, the experimental results show that a four-layer deep neural network fed with all available dimensions of student profiles has achieved the best prediction with RMSE 12.68.