Death toll forecast of coal mines based on"grey+nonlinear regression"
To explore the current level of safety management in China coal mines,this paper uses the number of deaths in coal mines in recent years as an entry point to analyze the predictive performance and error sources of the GM(1,1)model and regression model for coal mine accidents.It proposes improvement methods and finally establishes a predictive model based on"grey+regression".Firstly,a GM(1,1)model is established and its predictive level is verified as good by using a post-residual ratio and small probability error-index,which meets the improvement requirements.Secondly,a linear regression model is fitted and its error sources are analyzed.A"gray+regression"predictive model is established with the help of gray correlation degree.Lastly,the predictive errors of three models are compared and a"gray+nonlinear regression"model is used to predict the number of deaths in coal mines in 2023.The data show that:based on the actual number of deaths in previous years and the data obtained by the"gray+nonlinear regression"predictive model,a significance analysis is conducted and F=1 661.67≥F0.05(1,5)=6.608,indicating that there is a significant difference between two groups of data.Thus,the predictive model is validated and can be used for predicting coal mine deaths.Based on data from 2021 to 2022,relative errors for"gray+nonlinear regression"predictions are 4.5％and 1.3％respectively,which further improves predictive accuracy compared with the GM(1,1)model.The predictive results are closer to actual values.Compared with other models,the"gray+nonlinear regression"predictive model can perform continuous predictions and overcome short-term prediction limitations of the GM(1,1)model.According to predictive analysis,the number of deaths in coal mines in 2023 will be controlled between 123 and 136 people and show a downward trend.Safety management level in coal mines has been further improved.
safety engineeringnumber of fatalitiesgray model"gray+nonlinear regression"modelaccident prediction
万方数据
