铁道科学与工程学报2025，Vol.22Issue(5)：2317-2332,16.DOI:10.19713/j.cnki.43-1423/u.T20241707
基于神经网络的心率变异性指标预测的研究
HRV index prediction based on neural network
摘要
Abstract
There is a clear correlation between working fatigue and unsafe behavior,and working fatigue is closely related to transportation production safety.The detection of working fatigue has been widely studied,but is mainly about detecting the state based on current indicators.The ability to predict these indicators would enable the anticipation of future fatigue states,facilitating proactive measures to manage occupational fatigue more effectively.The Gated Recursive Unit(GRU)Neural Network model had demonstrated superior performance in addressing time series prediction challenges.This paper endeavored to construct a predictive model based on the GRU framework to enhance the forecasting accuracy of railway transportation personnel fatigue,thereby providing a foundation for preemptive intervention.Fatigue detection based on heart rate variability(HRV),MEAN,SDNN,RMSSD,HR,and other time domain indexes were the most basic and commonly used.Overlapping window sampling was used to preprocess data,and the Bayesian-optimized GRU model was used to predict HRV indicators.Eight data sets were gathered from rail station duty officers,with MEAN,SDNN,RMSSD,and HR datasets extracted from each.Verify prediction accuracy by cross-validation using fitting degree(R2)and root mean square error(Erms).Compare the BO-GRU model with models based on long-term and short-term memory neural networks(LSTM).The results show that in MEAN,SDNN,RMSSD,and HR prediction tasks,the BO-GRU model can achieve high precision prediction,and the average fitting degree(R2)reaches 0.951 99,0.962 6,0.962 22,and 0.948 46,respectively.The mean values of root mean square error(Erms)are 11.217 ms,9.451 7 ms,12.657 ms,and 1.535 7 times/min,respectively,which are better than those of the comparison model.The research results can provide a reference for the railway transportation industry in dealing with working fatigue and ensuring production safety.关键词
铁路安全/运输岗位/疲劳预测/神经网络/心率变异性Key words
railway safety/transport post/fatigue prediction/neural network/heart rate variability分类
交通工程引用本文复制引用
王健,邢科家,王玺..基于神经网络的心率变异性指标预测的研究[J].铁道科学与工程学报,2025,22(5):2317-2332,16.基金项目
国家自然科学基金资助项目(52172323) （52172323）
