To achieve the early prediction of the latent fire source of coal spontaneous combustion in the long-distance spontaneous combustion coal seam double coal roadways, it is proposed to jointly utilize the Data Analysis Method of the coal mine monitoring system which can early predict and forecast the latent spontaneous combustion fire and improve the reliability, while it still keeps and ameliorates the Gas Analysis Method. In order to single out the fluctuations of CO generated by the exhaust gas of diesel vehicles in the coal mine, it is introduced the background of CO concentration, and the basic assumption is made: in one day, one or more diesel vehicles can finally be found intermittent moments in the busy work (monitoring time) ), then given the background CO concentration in a day which is determined by the minimum monitored value of CO. Taking the shortest natural combustion period as the examination period, and the continuous increase of CO concentration as the criterion of judgment method, we give two prediction and prediction models of regression trend method and integral method, furthermore,the forecasting and forecasting is supposed to divide into three levels, namely, key inspection, enhanced inspection and the reason query. In combination with Hong Qingliang Coal Mines, the lowest average rate of increase of CO, k1^*= 0.607 ppm/d is obtained. Based on the above-mentioned reliable predicting and forecasting methods, the long-distance roadway will be changed the comprehensive inspection every day to the targeted key inspection.