王永斌, 李向文, 柴峰, 袁聚祥, 尹素凤, 武建辉. 采用灰色-广义回归神经网络组合模型预测我国尘肺病发病人数的方法探讨[J]. 环境与职业医学, 2016, 33(10): 984-987, 999. DOI: 10.13213/j.cnki.jeom.2016.15643
引用本文: 王永斌, 李向文, 柴峰, 袁聚祥, 尹素凤, 武建辉. 采用灰色-广义回归神经网络组合模型预测我国尘肺病发病人数的方法探讨[J]. 环境与职业医学, 2016, 33(10): 984-987, 999. DOI: 10.13213/j.cnki.jeom.2016.15643
WANG Yong-bin, LI Xiang-wen, CHAI Feng, YUAN Ju-xiang, YIN Su-feng, WU Jian-hui. Application of Grey Model-Generalized Regression Neural Network Combination Model to Prediction on Incidence of Pneumoconiosis in China[J]. Journal of Environmental and Occupational Medicine, 2016, 33(10): 984-987, 999. DOI: 10.13213/j.cnki.jeom.2016.15643
Citation: WANG Yong-bin, LI Xiang-wen, CHAI Feng, YUAN Ju-xiang, YIN Su-feng, WU Jian-hui. Application of Grey Model-Generalized Regression Neural Network Combination Model to Prediction on Incidence of Pneumoconiosis in China[J]. Journal of Environmental and Occupational Medicine, 2016, 33(10): 984-987, 999. DOI: 10.13213/j.cnki.jeom.2016.15643

采用灰色-广义回归神经网络组合模型预测我国尘肺病发病人数的方法探讨

Application of Grey Model-Generalized Regression Neural Network Combination Model to Prediction on Incidence of Pneumoconiosis in China

  • 摘要: 目的

    探讨灰色-广义回归神经网络组合模型GM(1,1)-GRNN在我国尘肺病发病人数预测中的应用,并比较其与灰色模型(GM)和反向传播网络(BPNN)模型的预测效果。

    方法

    收集2003—2012年我国尘肺病发病资料,用SAS9.3建立GM(1,1)模型,用Matlab 8.0建立BPNN模型和GM(1,1)-GRNN组合模型,并用2013年的数据评价模型的预测效果。

    结果

    GM(1,1)模型拟合及预测的平均相对误差(MRE),平均误差率(MER),均方误差(MSE)和平均绝对误差(MAE)分别为12.041%,0.122,4999 319.100,1781.100和20.033%,0.200,2151 104.000,4638.000;BPNN模型拟合及预测的MRE,MERMSEMAE分别为9.891%,0.124,3615 099.600,1802.000和6.932%,0.069,2576 025.000,1605.000;GM(1,1)-GRNN组合模型拟合及预测的MRD,MERMSEMAE分别为6.490%,0.069,1900 198.400,1001.200和3.939%,0.039,831744.000,912.000。GM(1,1)-GRNN组合模型预测的2014—2015年我国尘肺病的发病人数分别为23 768和23 434。

    结论

    GM(1,1)-GRNN组合模型的拟合及预测效果优于GM(1,1)模型和BPNN模型。

     

    Abstract: Objective

    To apply gray model plus generalized regression neural network GM(1, 1)-GRNN combination model to the prediction on incidence of pneumoconiosis in China and compare the predictive effects among GM(1, 1)-GRNN combination model, grey model (GM), and back-propagation network (BPNN).

    Methods

    The data of pneumoconiosis incidence from 2003 to 2012 in China were collected, SAS9.3 was used to construct GM(1, 1) model, and Matlab 8.0 was used to establish BPNN model and GM(1, 1)-GRNN combination model. Afterwards, the data in 2013 were used to evaluate the predictive effects.

    Results

    The mean relative error (MRE), mean error rate (MER), mean square error (MSE), and mean absolute error (NAE) fitted and forecasted by GM(1, 1) model were 12.041%, 0.122, 4 999 319.100, and 1 781.100 (fitted), and 20.033%, 0.200, 2 151 104.000, and 4 638.000 (forecasted); by BRNN, 9.891%, 0.124, 3 615 099.600, and 1 802.000 (fitted), and 6.932%, 0.069, 2 576 025.000, and 1 605.000 (forecasted); by GM(1, 1)-GRNN combination model, 6.490%, 0.069, 1 900 198.400, and 1 001.200 (fitted), and 3.939%, 0.039, 831 744.000, and 912.000 (forecasted), respectively. The incidences of pneumoconiosis from 2014 to 2015 forecasted by GM(1, 1)-GRNN combination model were 23 768 and 23 434, respectively.

    Conclusion

    GM(1, 1)-GRNN combination model is superior to GM(1, 1) model and BPNN model with better fitting and predictive effects.

     

/

返回文章
返回