Analysis and Prediction of Abnormal Behavior of Higher Vocational Students Based on XGBoost Model

Abnormal behaviors such as dropouts or suspensions in higher vocational colleges should not be attributed solely to a single factor or based on anecdotal evidence. Instead, these behaviors must be understood as the result of multiple interrelated influences. This paper examines and predicts such behaviors by analyzing recent student data from a higher vocational college. It investigates the correlations among twelve key variables–including student gender, major, admission scores, average academic performance, admission education level, and the occurrence of abnormal behaviors. To address the challenge of imbalanced sample distribution, a balancing algorithm is applied before training an XGBoost model on 80% of the dataset, while the remaining 20% is reserved for validation. Feature extraction techniques are then employed to identify a set of indicators strongly associated with student behavioral anomalies, and further parameter tuning is conducted to optimize the model. The resulting predictive model offers an objective reference for early warning mechanisms within higher vocational institutions and serves as a valuable tool for implementing targeted intervention measures to mitigate potential abnormal behaviors.