摘要
目的探讨贝叶斯正规化BP神经网络在我国医院床位预测中的应用。方法利用1990-2006年我国医院床位历史数据,建立贝叶斯正规化BP神经网络进行拟合预测,并与指数平滑预测、自回归模型的预测结果进行比较。结果三种方法预测结果的相对误差分别为0.58%,3.62%,1.48%。贝叶斯正规化BP神经网络模型预测精度更高,效果更好,优于传统方法。结论贝叶斯正规化BP神经网络预测模型可以用于我国医院床位预测。
Objective To explore the appfication of Bayesian-regularization BP neural network for predicting bospital beds. Methods According to the data of hospital beds from 1990 to 2006 ,Bayesian-regularization BP neural network was established. The forecast results were compared with exponential smoothing and autoregressive model. Results The relative errors of results with the three methods were 0.58 %, 3.62 % and 1.48 % in forecasting hospital beds, respectively. The forecasting result with Bayesian-regularization BP neural network was more precise and effective than that with traditional prediction methods. Conclusion Bayesian-regularization BP neural network predictive model can be used to predict hospital beds.
出处
《山西医科大学学报》
CAS
2008年第9期833-835,共3页
Journal of Shanxi Medical University
