Implementation of Machine Learning to Predict The Timeliness of Graduation of Employees on Study Assignment at Company X

Abstrak

The energy transition requires workers in the energy sector who have relevant skills that can be applied in the future. Company X implements a study assignment program to improve its employees' skills, but delays in completing their studies hinder their readiness to enter the workforce. Identifying the factors that influence graduation timeliness can improve the program's effectiveness. This study aims to develop a predictive model to determine whether employees in Company X's work-study program will graduate on time. The main purpose of this model is to provide early warnings about employees at risk of delays, enabling more targeted interventions to improve human resource management. We applied the CRISP-DM framework and used Machine Learning to analyze data from 317 employees who participated in the study program. Four machine learning algorithms were tested, namely Gradient Boosting, Decision Tree, Random Forest, and Naive Bayes. 17 factors were trained to cover academic, demographic, and administrative aspects to predict timely graduation. Among the algorithms tested, Gradient Boosting showed the best performance with an AUC of 0.956 and an accuracy of 0.909. These results were supported by high ROC and confusion matrix values, indicating the model's excellent predictive ability.