Support Vector Machine Model Construction Forecast of Patient Visits Hospital Operation
江苏省医院协会医院管理创新研究课题（JSYGY-2-2023-415）
目的：通过机械学习算法支持向量机（SVR）建立模型，评估其用于门诊量及出院人次预测的表现。方法：以2016年1月至2024年4月某三甲医院门诊人次及出院人次为研究对象，采用不同的核函数，构建三种不同的模型（线性核SVR，多线程SVR，高斯核SVR）。结果：线性核SVR预测门诊量RMSE、MAE、MAPE值分别为13 470、10 510和9.8%，出院人次RMSE、MAE、MAPE值分别为1 005、753和10.8%，均优于多线程SVR和高斯核SVR模型预测结果。结论：在构建的三种SVR模型测算中，线性核SVR模型预测表现最佳，更适合用于医院门诊和住院就诊人次预测，能够为医院在医疗资源调度与分配等管理决策方面提供更为坚实的数据支撑，有效确保医院高效、有序运营。
Objective：To develop a model utilizing the support vector machine（SVR）machine learning algorithm and assess its efficacy in predicting outpatient visits and discharges. Methods：The study used the number of outpatient visits and discharges from a three-class hospital from January 2016 to April 2024 as the research object，and different kernel functions were used to build three different models（linear kernel SVR，multi-thread SVR，Gaussian kernel SVR）. Results：The RMSE，MAE，and MAPE values for predicting outpatient visits using the linear kernel SVR were 13470，10510 and 9.8%. The RMSE，MAE，and MAPE values of the number of discharged patients were 1005，753，and 10.8%，respectively.，which were all better than the prediction results of the multi-thread SVR and Gaussian kernel SVR models.Conclusion：In the calculation of the three SVR models，the linear kernel SVR model had the best prediction performance and is more suitable for predicting the number of outpatient and inpatient visits in a hospital，which can provide more solid data support for hospital management decisions on medical resource allocation and scheduling，and effectively ensure the efficient and orderly operation of the hospital.