目的：研究径向基函数（ RBF）神经网络与多重线性回归的组合模型在煤工尘肺发病工龄预测中的性能优劣。方法：采用RBF神经网络模型与多重线性回归模型对研究数据进行分析，对2模型进行加权拟合，采用均方根误差、均方误差、平均相对误差对模型的预测性能进行评价。结果：多重线性回归模型、RBF神经网络模型和组合模型真实值与预测值比较，差异均无统计学意义（t配对＝1．552、0．231、0．155，P均＞0．05）。多重线性回归模型、RBF神经网络模型和组合模型的均方根误差分别为（1．63±0．11）、（2．45±0．19）和（0．59±0．07）（F＝26．141， P＜0．001），均方误差分别为（2．6569±0．2412）、（5．9867±0．3804）和（0．3483±0．0653）（F ＝49．678， P＜0．001），平均相对误差分别为（7．15±0．82）％、（15．39±1．25）％和（3．68±0．59）％（F＝35．282，P＜0．001）。结论：在煤工尘肺发病工龄的预测中，组合模型预测性能优于单一模型。%Aim:To study the pros and cons of prediction performance of multiple linear regression model and radical basis function neural network combined model to forecast the work year of coal workers ′pneumoconiosis .Methods:Root of mean square error , mean square predict error , and mean percent error were applied to analyze the predicting outcomes of the three models in order to achieve the aim of comparing the prediction performance .Results:For multiple linear regres-sion model ,radical basis function neural network and the combination model , the difference between true and predicted val-ues were significant(tpaired =1.552,0.231, and 0.155, P>0.05).The root of mean square error of the multiple linear re-gression model,radical basis function neural network and the combination model was respectively (1.63 ±0.11),(2.45 ± 0.19),and (0.59 ±0.07)(F =26.141,P <0.001).The mean square predict error was respectively (2.656 9 ± 0.241 2),(5.986 7 ±0.380 4),and(0.348 3 ±0.065 3)(F=49.678,P<0.001).The mean percent error was respec-tively (7.15 ±0.82)%,(15.39 ±1.25)%,and (3.68 ±0.59)%(F=35.282,P<0.001).Conclusion:In the predic-tion of coal workers′pneumoconiosis incidence seniority , combined forecasting model is superior to a single model .
展开▼
