The Effect of Data Transformation Techniques on Machine Learning Performance: A Case Study on Student Dropout Prediction

This study investigates the effects of data transformation techniques on the performance of various machine learning classification models. Specifically, the standard scaler (or Z-score) and min-max scaler techniques coupled with several machine learning and statistical algorithms, including K-Nearest Neighbors, Naive Bayes, Support Vector Machine, and Logistic Regression (LR), are examined systematically. Based on a set of student performance data from an institution, application of the resulting model to predicting student dropout rate is demonstrated. The empirical results indicate that LR with a standard scaler yields the highest AUC (area under the receiver operating characteristic curve) score of 89.95%.