中低强度噪声致高频听力损伤的BP神经网络模型研究

目的]建立中低强度噪声对高频听力损伤关系的BP神经网络模型,探讨BP神经网络模型在职业医学方面的应用。方法]选择590名中低强度噪声接触者作为研究对象,应用SPSS10.0进行中低强度噪声和高频听力损伤关系的多元线性回归分析,同时应用Matlab7.0建立中低强度噪声和高频听阈关系的BP神经网络模型,比较人工神经网络模型和统计模型的预测效果。结果]得到由4个输入单元、3个输出单元、18个隐含单元组成的BP神经网络模型,应用BP神经网络预测的误差绝对值平均值和相对误差绝对值都低于应用线性方程的预测结果。结论]BP神经网络应用于中低强度噪声对高频听力损伤关系的分析效果较好,值得推广应用。

Study of the Model of BP Neural Network of High Frequency Hearing Loss by low-middle Dose Noise

Objective]To build and evaluate the back-propagation(BP) neural network mode on the relationship between low-middle dose noise and high frequency hearing loss and to apply the models in occupational medicine.Methods]590 workers exposed to low-middle dose noise were selected as exposed group.Multiple linear regression analysis of the relationship between low-middle dose noise and high frequency hearing loss was conducted with SPSS10.0.The BP neural network model was built by using Matlab7.0.Forecast effect of the model of artificial neural network and statistical model was compared.Results]The model of BP neural net-work was composed of 4 input units and 3 output unit and 18 implicit units.The average relative error and the average error of the forecasting values appeared to be less than that of the traditional forecasting model.Conclusion]The BP neural network model would be valuable to be popularized as an important tool to forecast hearing loss induced by low-middle dose noise.