Combining the Unsupervised Discretization Method and the Statistical Machine Learning on the Students’ Performance

The suitability of the data with the method in the process of data mining is very important to increase the process performance. However, In Educational Data Mining (EDM), not much research has focused on this field. Therefore, this study proposes to combine an unsupervised discretization method called "equal width interval" and logistic regression as statistical machine learning to improve the performance of the model relating to students' performance. The discretization method is performed on student data with several intervals, namely: 3-interval, 4-interval, and 5-interval. Then, these intervals are combined with logistic regression in two regularizations, namely: lasso and ridge. Evaluation is carried out on all combinations. The experimental results indicate that the highest performance, in terms of the accuracy level, is achieved by the model combining a 3-interval and logistic regression on all regularization. This combination can increase the model performance based on the average accuracy level of about 4.08-8.49 on the ridge regularization and about 4.28-8.6 on the lasso regularization.