Machine Learning–Driven Classification Model to Support Study Assignment Decisions under the Study-on-Mission Program at PT X
TL;DR
This study evaluates the efficacy of machine learning as a decision support instrument for mission assignment at PT X, grounded in empirical model performance and data attributes, indicating that machine learning models are more efficacious when utilized as decision support systems instead of automated decision-makers.
Abstract
This study evaluates the efficacy of machine learning as a decision support instrument for mission assignment at PT X, grounded in empirical model performance and data attributes. Analysis of 98 employee records indicates that previous study assignment patterns can be partially inferred from employee profile features, however the prediction performance is moderate. Of the assessed models, the random forest classifier exhibits the most dependable outcomes, attaining the highest accuracy (60%) and weighted F1-score (53%) in comparison to the decision tree and logistic regression models. Nevertheless, an examination of precision, F1-scores, and confusion matrices uncovers persistent deficiencies in forecasting minority study categories attributable to class imbalance and restricted feature variety. This suggests that previous study assignment practices have focused on a limited number of predominant fields, restricting the model's capacity to generalize across less common categories. The results indicate that machine learning models are more efficacious when utilized as decision support systems instead of automated decision-makers. The findings underscore the significance of people analytics in enhancing consistency and transparency in study assignment decisions, while stressing the necessity for more comprehensive data, strategies for mitigating imbalances, and human-in-the-loop governance to guarantee equitable and contextually informed competency development policies.