Prediction of the positive psychology of college students using LightGBM algorithm based on SMOTE and Bayesian optimization

In our work, the LightGBM algorithm based on Bayesian optimization is employed to classify and predict the characteristic factors affecting the college students' positive psychology; the unbalanced data is processed using the SMOTE algorithm, a high-precision prediction model of positive psychology is constructed on the basis of the data preprocessing, and the commonly used classification assessment indexes are used for evaluation of the model performance; finally, the characteristic factors affecting positive psychology are ranked in accordance with their importance, and analyzed. The results show that the LightGBM algorithm based on SMOTE and Bayesian optimization performs optimally on the assessment indexes of Accuracy, Recall, Precision, and f1, and outperforms the Xgboost, lr, and catboost algorithms.