Zhou Xian-Feng, Feng Zi-Jian, Yang Wei-Zhong, Li Xiao-Song
Department of Health Statistics, West China School of Public Health, Sichuan University, Chengdu 610041, China.
Sichuan Da Xue Xue Bao Yi Xue Ban. 2011 Jul;42(4):544-7.
To apply Wavelet Neural Networks (WNN) model to forecast incidence of Syphilis.
Back Propagation Neural Network (BPNN) and WNN were developed based on the monthly incidence of Syphilis in Sichuan province from 2004 to 2008. The accuracy of forecast was compared between the two models.
In the training approximation, the mean absolute error (MAE), rooted mean square error (RMSE) and mean absolute percentage error (MAPE) were 0.0719, 0.0862 and 11.52% respectively for WNN, and 0.0892, 0.1183 and 14.87% respectively for BPNN. The three indexes for generalization of models were 0.0497, 0.0513 and 4.60% for WNN, and 0.0816, 0.1119 and 7.25% for BPNN.
WNN is a better model for short-term forecasting of Syphilis.
应用小波神经网络（WNN）模型预测梅毒发病率。
基于四川省2004年至2008年梅毒的月发病率建立反向传播神经网络（BPNN）和WNN。比较两种模型预测的准确性。
在训练近似中，WNN的平均绝对误差（MAE）、均方根误差（RMSE）和平均绝对百分比误差（MAPE）分别为0.0719、0.0862和11.52%，BPNN分别为0.0892、0.1183和14.87%。模型泛化的三个指标，WNN为0.0497、0.0513和4.60%，BPNN为0.0816、0.1119和7.25%。
WNN是梅毒短期预测的较好模型。
