3种模型在煤工尘肺病预测中的比较研究  

Comparative study of three models in prediction of coal workers’ pneumoconiosis

作　　者：张毓媛 张延松 刘博 ZHANG Yu-yuan;ZHANG Yan-song;LIU Bo(Shandong University of Science and Technology,College of Resources and Environmental Engineering,Qingdao 266590,China)

机构地区：[1]山东科技大学矿业与安全工程学院,山东青岛266590 [2]山东科技大学矿山灾害预防控制-省部共建国家重点实验室培育基地,山东青岛266590

出　　处：《工业卫生与职业病》2020年第2期94-97,共4页Industrial Health and Occupational Diseases

基　　金：国家重点研发计划(2017YFC0805200)。

摘　　要：目的基于当前煤工尘肺预测方法的现状,研究尘肺病的发病特征,对比分析BP神经网络模型、多元线性回归模型及两种模型的组合模型在尘肺病发病工龄预测中的优劣。方法采用SPSS 22.0中的BP神经网络模型、多元线性回归模型和组合模型,在传统模型的基础上加入粉尘浓度、粉尘分散度和游离SO2含量3个变量,对H矿业集团1963-2017年煤矿工人尘肺病患者的数据进行分析,并对3种模型的真实值与预测值之间进行配对t检验,对模型的预测结果进行对比分析。结果组合模型(t=0.363,P=0.816>0.05)在3种模型中误差参数最优,均方误差、平均相对误差和平均绝对误差都小于其他两种单一模型,对尘肺病发病工龄的预测精度最高。结论通过组合模型预测,H矿业集团到2020年有376名煤矿工人可能患有尘肺病,发病率为4.3%,须及时采取措施,调离岗位。Objective Based on the current situation of coal workers’pneumoconiosis prediction methods,the incidence characteristics of pneumoconiosis were studied,and the advantages and disadvantages of BP neural network model,multiple linear regression model and the combination model of the two models in the prediction of the duration of pneumoconiosis were compared and analyzed.Methods Using BP neural network model,multiple linear regression model and combination model in SPSS22.0,the data of coal miners suffering from pneumoconiosis in H Mining Group from 1963 to 2017 were analyzed by adding three variables of dust concentration,dust dispersion and free SiO2 content on the basis of the traditional model.The paired test between the real value and the predicted value of the three models was carried out,and the prediction conclusion of the model was made.The results were compared and analyzed.Results The combined model(t=0.363,P=0.816>0.05)had the best error parameters among the three models.The mean square error,average relative error and average absolute error were all smaller than the other two single models,and the prediction accuracy of the duration of pneumoconiosis was the highest.Conclusions The combined model predicts that 376 coal miners in H Mining Group may suffer from pneumoconiosis by 2020,with an incidence rate of 4.3%.It is necessary to take timely measures to remove them from their posts.

关 键 词：BP神经网络 组合模型 煤工尘肺 发病工龄 

分 类 号：R135.2[医药卫生—劳动卫生]