Prediction of the length of service at the onset of coal workers’ pneumoconiosis based on neural network

Abstract

Three environmental parameters, i.e. dust concentrations, dust dispersion, and free silica content, were introduced into the traditional indices of the neural network model in order to construct a new prediction index and explore a new method for preventing the incidence of pneumoconiosis with intelligent accuracy and universality. Data of the pneumoconiosis patients from Huabei Mining Group (HBMG) of China from 1980 to 2017 were collected. SPSS22.0 was used to develop the combined models based on Back Propagation (BP) neural network model, Radial Basis Function (RBF) neural network model, and Multiple Linear Regression (MLR) model. The paired sample t-test was performed between the real and predicted values. According to this model, it was predicted that 382 coal workers in HBMG were likely to suffer from pneumoconiosis in 2022 and the incidence rate was 4.48%. It is necessary to take prevention measures and transfer these workers from their current positions. In four combined models, the BP-MLR combined model achieved the optimal error parameters and the most accurate prediction. This study provided a scientific basis for effective control and prevention of the incidence of the pneumoconiosis.