The object of this research is the performance of air transportation vocational education. The problem in this research that must be solved is the complexity of the model in machine learning which requires a long processing time and requires high resources, so the knowledge transfer process in knowledge distillation must be carried out carefully so that the student model can capture and reproduce knowledge from the teacher's model. without loss of accuracy and problems such as Good Corporate Governance, Organizational Flexibility, and Strategic Change Management variables, which are interrelated and difficult to model accurately. The results obtained are in the form of a model that can predict vocational education performance by utilizing machine learning and knowledge distillation. The interpretation of this research is to apply the XGBoost machine learning algorithm and knowledge distillation. The characteristics and characteristics obtained are that the teacher model has the best performance in terms of loss, while the student model with distillation shows a significant reduction in loss compared to training without distillation. Thus, the distillation process is proven to help student models capture knowledge from teacher models, producing prediction accuracy of up to 90 % and being an efficient alternative in predicting the influence of main factors on the performance of air transportation vocational education. These findings are expected to provide a significant contribution to the development of more efficient and effective prediction models in the context of vocational education, especially in the field of air transportation
Read full abstract