Penerapan Algoritma Klasifikasi Dengan Fitur Seleksi Weight By Information Gain Pada Pemodelan Prediksi Kelulusan Mahasiswa

Abstract

The problem faced by institutions of higher education is not exactly of graduation. It is the task of institution of higher education, especially courses in the students academic monitoring. Today, there are many methods or ways to solve various problems of information technology either by machine learning at data mining. This study uses Naive Bayes. Linear Regression, dan Multi Layer Perceptron algorithm also Weight by Information Gain as Features Selection to optimize accuracy in predicting the time of students graduation. The dataset processing with certain attributes including registration line, origin school, origin city, and the semester grade point 1 to 4 in RapidMiner by implementing the three algorithms along with the selection feature produces relatively good accuracy. Naive Bayes produces an accuracy of 81.66% with an execution time of 1.16 seconds, Linear Regression of 80.70% in 2.44 seconds and Multi Layer Perceptron of 82.16% in 1 hour 57 minutes.