基于PSO-LSTM模型对上海某院糖尿病患者住院量的预测研究
Prediction Study of Hospitalization Volume in a Hospital in Shanghai Based on PSO-LSTM Model

目的：当前医疗管理在资源分配和效率提升方面缺乏精确的住院量预测方法，制约了卫生系统的优化。为解决这一问题，本研究提出一种基于粒子群优化的长短期记忆网络(PSO-LSTM)模型，为医疗资源配置提供科学支持。方法：采用2013年至2023上半年上海崇明地区某医院的住院量数据，构建并优化PSO-LSTM模型。PSO算法通过100次迭代对LSTM模型的关键超参数进行全局优化，并与ARIMA、SARIMA和SVR等模型进行对比分析。结果：PSO-LSTM模型在MSE、RMSE、MAE和R²等指标上表现最佳，相较于SVR模型，MSE降低81.7%，R²提高26.2%。结论：研究结果表明，PSO-LSTM模型有效弥补了传统方法的局限性，为精准预测住院量提供了一种科学的解决方案，有效提高医疗资源配置效率，为卫生系统管理决策提供支持。
Objective: The lack of accurate inpatient volume prediction methods in current healthcare management hinders effective resource allocation and system optimization. To address this issue, this study proposes a Particle Swarm Optimization-based Long Short-Term Memory (PSO-LSTM) model to provide scientific support for the allocation of medical resources. Methods: Using inpatient data from a hospital in Chongming, Shanghai, spanning 2013 to the first half of 2023, the PSO-LSTM model was developed and optimized. The PSO algorithm conducted 100 iterations to globally optimize key hyperparameters of the LSTM model. The performance of the proposed model was compared against ARIMA, SARIMA, and SVR models. Results: The PSO-LSTM model demonstrated superior performance across all metrics, including MSE, RMSE, MAE, and R². Compared to the SVR model, the PSO-LSTM model achieved an 81.7% reduction in MSE and a 26.2% improvement in R². Conclusion: The results indicate that the PSO-LSTM model effectively addresses the limitations of traditional methods, providing a robust solution for accurate inpatient volume prediction. This model enhances the efficiency of medical resource allocation and supports healthcare system management decisions.