基于熵权的医院门诊量组合预测模型

Combination prediction model based on entropy weight for forecasting hospital outpatient amount

目的：探讨基于熵权理论的组合预测模型预测医院门诊量,以提高预测的精度。方法依据信息熵理论,选取平方和误差（ SSE）、平均绝对误差（ MAE）、均方误差（ MSE）、平均绝对百分比误差（ MAPE）、均方百分比误差（ MSPE）等5个误差指标作为评价指标,根据这些预测误差所反映出的信息熵,确定所选各单项预测方法的权重,得到基于熵权的组合预测模型,并将模型应用于文献的数据,与该文献模型的预测效果作比较。结果文献[1]的折扣最小二乘法模型、GM（1,1）模型、自回归预测模型和组合模型的5个误差指标分别为：115.27911、1.99707、0.71579、0.04408、0.01401；67.53414、1.64571、0.54786、0.03871、0.01229；70.47580、1.37669、0.55967、0.03417、0.01340和58.64677、1.44235、0.51054、0.03471、0.01180。熵权组合模型的5个误差指标为50.39541、1.39713、0.47326、0.03364、0.01116。基于熵权的组合预测方法有较高的拟合精度,预测效果优于单项预测模型和文献的组合模型。结论熵权组合预测模型计算简单,预测效果好,可以在医院门诊量预测中推广应用。

Objective To investigate the combination prediction model based on the entropy weight theory for forecasting hospital outpatient amount so as to improve the prediction accuracy. Methods On the basis of the information entropy theory, five error indexes, sum of squares for error （ SSE） , mean absolute error（ MAE） , mean square for error （ MSE） , mean absolute percentage for error （ MAPE） , and mean square percent for error （ MSPE） , were selected as evaluating indexes and according to the information entropy which derived from forecasting errors, the weights of each prediction method were confirmed to get a combination prediction model. The model was used to forecast outpatient amount of hospital with the data from literature, and then the prediction effect was compared with that using the model from the literature. Results Five error indicators for discount least square model, GM（1,1） model, auto-regressive prediction model and combination model from literature[1] were 115. 279 11, 1. 997 07, 0. 715 79, 0. 044 08, 0. 014 01;67. 534 14, 1. 645 71, 0. 547 86, 0. 038 71, 0. 012 29;70. 475 80, 1. 376 69, 0. 559 67, 0. 034 17, 0. 013 40 and 58. 646 77, 1. 442 35, 0. 510 54, 0. 034 71, 0. 011 80 respectively. Those of our entropy weight model were 50. 395 41, 1. 397 13, 0. 473 26, 0. 033 64, and 0. 011 16. The higher fitting precision was achieved by the combination forecasting model based on entropy weight and its prediction effect was better than those using the monomial forecasting model or the combination model from literature. Conclusion The combination model based on entropy weight is simple in calculation, has good effect for prediction, and can be applied in the prediction of hospital outpatient amount.