Abstract: In order to predict mine dust concentration accurately and prevent mine dust disaster effectively, this paper puts forward a differential autoregressive moving average prediction model based on a time series of mine dust concentration.Based on the characteristics that the dust concentration is a non-stationary random sequence and the ARIMA prediction model can process non-stationary data, SPSS statistical analysis software is used to establish an ARIMA dust concentration prediction model.Firstly,the dust concentration data should be taken stable processing.Model parameters are determined according to the autocorrelation and partial autocorrelation coefficients and BIC criterion.ARIMA(1, 2, 1) model is preliminarily selected,and then the model is tested by residual autocorrelation and partial autocorrelation functions, which further verified the rationality of the model.Finally,the model is used to predict the dust concentration.The results show that the relative error is 8.34% at the maximum and 2.40% at the minimum, and all the relative errors are controlled within 10%.ARIMA model can be used to predict mine dust concentration and the prediction effect is good.