IMPLEMENTATION OF THE DECISION TREE MODEL ON MACHINE LEARNING TO PREDICT POTENTIAL NEW STUDENTS

Abstract

According to a previous study, “Implementation of Naïve Bayes Classifier-based Machine Learning to Predict and Classify New Students at Matana University” has an accuracy of 0.73 or 73%. This is not optimal, the accuracy needs to be improved. In this research, to increase accuracy by using a different model, namely the Decision Tree model. The reason for choosing the Decision Tree is that there are not many predictors used (4 predictors) and can be used for classification or prediction. The 4 predictors are frequency, position, majors of students in SMA/K, and research programs of interest. The target is the entry status of prospective students. The research procedures that were tried were information gathering, pre-processing, machine learning processes with the Decision Tree model and visualization of the results. The programming language used is Python. The results of this Decision Tree show changes, through 10 executions the average accuracy of the ratio of training information and test information is 70: 30 of 0.727 or 72.7% (lowest accuracy is 47% and highest is 87%), for a ratio 80: 20 of 0, 73 or 73% (the lowest accuracy is 60% and the highest is 90%). Thus, the results of the Decision Tree model on average have not exceeded the results of the Naïve Bayes Classifier model. Further research, increase the amount and alteration of information, reduce the difference in results each time the model is executed, try other models, and improve the application ready to use.