Predicting student's psychomotor domain on the vocational senior high school using linear regression

Abstract

The educational data can be mined to produce the useful knowledge. This paper focuses on the educational data processing to predict student's psychomotor domain. Here, we apply linear regression method to do it. On process stage, we use 4 regularizations, namely: no regularization, ridge regression, lasso regression and elastic net regression. Furthermore, we exploit 2 sampling methods as the evaluation technique, for examples: cross-validation sampling and random sampling. The experimental result indicates that the best regularization on cross-validation and random sampling are an elastic net regression because this regularization achieves the lowest predicting error. On cross-validation, values of MSE, RMSE, and MAE are 40.079, 6.330 and 5.183, respectively. Additionally, for random sampling, respectively, values of MSE, RMSE, and MAE are 86.910, 8.428 and 6.511.