Abstract: Abstract: To investigate the feasibility of machine learning in predicting the minimum ignition temperature of pulverized coal clouds, based on the minimum ignition temperature (MITc) of coal dust clouds and the influence factors obtained from previous tests using Godbert-Greenwal furnace, analyzed the correlation of the influence factors. The prediction effectiveness of the three machine learning models in terms of both the minimum ignition temperature of coal dust clouds and the incidence of ignition was evaluated and analysed using the AUC/ROC, Kappa coefficient, sensitivity, specificity, MAE and RMSE metrics. The results showed that the RSM model has the worst prediction effect; the RF model has the best accuracy and stability in predicting the MITc and ignition probability of pulverized coal clouds, and with the Bagging model, the AUC values are greater than 0.85 in predicting the ignition incidence, but the effect of predicting MITc is poor. The results provide a new research idea on prediction of coal dust cloud ignition sensitivity.

Key words: Key words: coal dust cloud; ignition sensitivity; random forest; Bagging model