Prediction of Career Suitability for Graduates of Teacher Education Programs as Support for Higher Education Decision-Making using Random Forest and XGBoost Algorithms

Abstract

ABSTRACT

The alignment between graduates’ educational backgrounds and their employment fields is an important indicator for evaluating the relevance of higher education programs to labor market needs. This study aims to classify the career suitability of graduates from teacher education programs using Random Forest and XGBoost algorithms to support higher education decision-making. The dataset consists of 730 graduates from 2025 in the Stabat area. The predictor variables include gender, study program, Grade Point Average (GPA), and employment status, while career suitability serves as the target variable. The research process involves data preprocessing, categorical variable encoding, stratified data splitting using an 80:20 ratio, model development, and performance evaluation based on accuracy, precision, recall, and F1-score. The results show that 454 graduates (62.19%) were classified as having careers suitable for their fields of study, while 276 graduates (37.81%) were classified as having unsuitable careers. According to the model performance comparison table, Random Forest achieved an accuracy of 66.44%, precision of 81.82%, recall of 59.34%, and F1-score of 68.79%. Meanwhile, XGBoost achieved an accuracy of 65.75%, precision of 75.95%, recall of 65.93%, and F1-score of 70.59%. Random Forest achieved higher accuracy and precision, whereas XGBoost obtained higher recall and F1-score. These findings suggest that both algorithms have the potential to provide preliminary information for evaluating graduate career suitability, reviewing curriculum relevance, and improving graduate competencies and career guidance services. However, the predictive performance of both models remains limited, and further testing and validation are required before practical implementation in decision-making.