Accuracy’s Comparison of Machine Learning Models for Predicting State College Admission Selection

Abstract

Seleksi Nasional Masuk Perguruan Tinggi Negeri (SNMPTN) is still one of the favorite admission routes for high school students to continue their education at Perguruan Tinggi Negeri (PTN). SNMPTN uses semester 1 to 5 report card scores for 6 subjects that are inputted in Pangkalan Data Sekolah dan Siswa (PDSS). Prediction of SNMPTN can be done using machine learning models with various methods. This study aims to create a predictive model using the Decision Tree CART, Gaussian Naïve Bayes and Logistic Regression methods, make predictions and compare the level of accuracy of the models made. The methodology used in this research is Knowledge Discovery in Database (KDD). This is to get useful knowledge from data. The dataset used is data on the scores of 6 subjects for 5 semesters from class 2015 to 2022. Model evaluation uses the Split Percentage Method and K-Fold Cross Validation. The results show that the accuracy scores for the 3 models are different. Logistic Regression has a score of 0.82, followed by Decision Tree CART with a score of 0.75 and finally Gaussian Naïve Bayes with a score of 0.70. The hypothesis put forward by the researcher is in accordance with the results obtained, that the Logistic Regression model has a higher accuracy score. Mathematically, Logistic Regression is not too complicated when compared to other models. To get a model that fits with needs must involve iterating through the machine learning process and trying various variations.