基于组合预测模型对门诊量的预测

Forecasting outpatient visits based on combined forecasting model

目的充分利用单个预测模型的有用信息，建立精度更高的组合预测模型。方法用1993--2002年某院门诊量实际数据建立3个单项预测模型，按“误差平方和最小”原则，将自回归预测模型、折扣最小二乘法模型和GM（1，1）模型优化组合成一新模型；然后分别用这4种模型对2003-2008年的门诊量进行预测，并对预测结果进行分析比较。结果折扣最小二乘法模型、GM（1，1）模型、自回归预测模型和组合预测模型的误差平方和分别为97．1885、57．4453、57．9136和49．5918。结论组合预测模型优于单个预测模型。

Objective To establish a better combined forecasting model so as to make full use of useful information of some individual forecasting models. Methods Three individual forecasting models were established with the actual data of the outpatient visits from 1993 to 2002. According to the principle of ＂the minimum value of the sum of error square＂, auto-regressive forecasting model, discount least square model and GM （ 1,1 ） model were combined into a new model, and the outpatient visits from 2003 to 2008 were predicted with the four models, and forecasting effects of them were compared and analyzed. Results The sum of error square of the discount least square model, GM （ 1,1 ） model, auto-regressive forecasting model and combined forecasting model was 97. 188 5, 57. 445 3, 57. 913 6, 49. 591 8 respectively. Conclusion The combined forecasting model is superior to any one of the individual forecasting models.