针对当下大学生心理健康预警方法存在计算复杂和准确率低的缺点,提出一种基于逻辑回归+轻度梯度提升树(LR+LightGBM)模型的大学生心理健康预警方法.模型选择中国常规模式评价指南和症状自评量表SCL-90中的90个项目作为输入数据,将数据先通过LR模型进行处理,输出结果作为一个复合评价指标,并同其他数据一起输入到LightGBM模型进行预测.融合模型结果与LR、决策树、随机森林、LightGBM对比发现,基于LR+LightGBM模型的心理健康预测准确率、召回率、AUC(AREA UNDER CURVE)值分别为0.917、0.903、0.97,优于其他对比模型,能够有效提升大学生心理健康预警的准确率,为高校大学生心理辅导提供科学决策依据.
Psychological Health Warning of College Students Based on LR+LightGBM Model
In response to the shortcomings of computational complexity and low accuracy in current methods for predicting the mental health of college students,a LR+LightGBM model based on the method for predict-ing the mental health of college students is proposed.The model selects 90 items from the Chinese Standard Mode Evaluation Guidelines and Symptom Checklist-90(SCL-90)as input data.The data is first processed through the LR model to obtain a composite evaluation index,which is then input into the LightGBM model along with other data for prediction.The results of the fusion model are compared with other single models.It is found that the accuracy,recall and AUC values of mental health prediction based on the LR+LightGBM model are 0.917,0.903 and 0.97,respectively,which are superior to other comparison models.It can effec-tively improve the early warning accuracy of college students' mental health,and provide scientific decision-making basis for college students'psychological counseling.
college student mental healthearly warning systemlogistic regressionLightGBM model