Early Prediction of Student Suspension and Dropout for Counseling Resource Optimization Using Multilayer Perceptron

The sub-replacement fertility rate in Taiwan has caused many educational institutions to face significant challenges, making student suspension and dropout critical issues for sustainable operations. Addressing these challenges requires effective prediction tools to identify at-risk students and allow for timely intervention. This study applies a Multilayer Perceptron to predict student suspension and dropout, with the goal of assisting schools in better monitoring these situations and proactively deploying counseling resources. By enhancing the precision of resource allocation, schools can improve the effectiveness of their interventions. The dataset used in this study includes student records from Chaoyang University of Technology spanning five academic years (2017–2021). Data from the second semester of the 2021 academic year were designated for model testing, while data from the other nine semesters were used for training and validation. The proposed model achieved an accuracy of 81.70%, demonstrating its potential to provide valuable insights. Future efforts will focus on further optimizing the model and deploying it in real-world applications to serve as a critical tool for counseling resource allocation and institutional decision-making.