Optimasi Prediksi Kelulusan Tepat Waktu: Studi Perbandingan Algoritma Random Forest dan Algoritma K-NN Berbasis PSO

Abstract

The prediction of on-time graduation for students involves various measurement techniques, including criteria such as majors, class types, and semester grade achievements. These factors play a crucial role in determining whether students will complete their studies within the designated timeframe. In line with this, a model has been developed to forecast the probability of timely graduation. This model leverages the Random Forest and k-Nearest Neighbor (K-NN) algorithms as tools to classify students into appropriate groups. Optimization is carried out using the Particle Swarm Optimizer (PSO) algorithm to enhance prediction accuracy. The data used originates from alumni of various Universities in Palembang.This model utilizes multiple attributes, such as majors, university origins, class types, and semester grade records up to the fourth semester. Other attributes encompass the year of graduation and year of enrollment. Data management and processing are conducted using Rapidminer. Validation is performed by splitting the dataset into training and testing groups through the split validation method. Based on research and testing, the Random Forest algorithm achieves an accuracy of 95.79% with an Area Under Curve (AUC) of 0.991. After optimization with PSO, accuracy increases to 97.89% with an AUC of 0.993. Meanwhile, the k-NN algorithm achieves an accuracy of 93.49% with an AUC of 0.975; after optimization with PSO, accuracy rises to 96.74% with an AUC of 0.986.