摘要
目的建立氟性骨损伤人工神经网络模型,预测氟性骨损伤发生的危险性。方法选择氟接触工人的年龄、工龄、车间、饮酒史、降钙素受体(CTR)基因作为输入层,以是否患氟性骨损伤为输出层,采用Levenberg-Marquardt优化算法训练网络并建立人工神经网络模型,验证网络模型的实用性和可靠性。结果模型的拟合度为99.2%,预测的灵敏度、特异度、准确度、阳性预测值和阴性预测值分别为95.8%、77.3%、87.0%、82.1%和94.4%。结论人工神经网络模型在氟性骨损伤发病风险的预测中取得了较好的效果。
Objective To establish an artificial neural network model to predict the risk of occurrence of fluoride bone injury.Methods The variables including age,working-age,workshop,drinking history,and the gene type of calcitonin receptor（CTR）were taken as the input layer,while the fluoride bone injury（yes=1,no=0）was taken as the output layer.The Levenberg-Marquardt optimized calculation method was used to train the network and to establish the model,and finally to testify its practicability and reliability.Results The fitness of the model was 99.2%.The sensitivity,specificity,accuracy,positive and negative predictive values for fluoride bone injury were 95.8%,77.3%,87.0%,82.1% and 94.4% respectively.Conclusions The artificial neural network model achieved good efficiency in predicting the risk of occurrence of fluoride bone injury.
出处
《工业卫生与职业病》
CAS
CSCD
北大核心
2011年第4期218-220,共3页
Industrial Health and Occupational Diseases
基金
国家自然科学基金资助项目(30671742
81072255)
关键词
氟性骨损伤
预测
人工神经网络模型
Fluoride bone injury
Prediction
Artificial neural network model
