In order to accurately investigate the key microstructures in the spontaneous combustion exothermic process of coal, an ultrasonic extraction method was employed to extract the coal, and the complex microscopic groups within it were stripped and studied. On this basis, Fourier transform infrared spectroscopy and differential scanning calorimetry were employed to assess the content of microscopic groups and the exothermic characteristics of the raw and extracted coal samples. The findings indicated that toluene and methanol demonstrated a notable capacity for extracting aromatic and aliphatic hydrocarbon compounds from coal, whereas N-methyl pyrrolidone (NMP) and ethylenediamine (EDA) exhibited a pronounced effect on oxygen-containing functional groups and hydroxyl groups. The heat flow curves and spontaneous combustion risk indices of the extracted coal samples were reduced to varying degrees, and the coal-oxygen reaction was suppressed. The order of the coal samples' spontaneous combustion risk indices was E-EDA, E-NMP, E-methanol, and E-toluene, with the latter having the lowest value. The effects of −OH-a and oxygen-containing functional groups on the spontaneous combustion exotherm of the coal samples were greater. The Pearson correlation coefficient method was employed to identify the key groups with the highest correlation with the risk indices of coal. These were found to be −CH 2 − and −OH-a, respectively. A multivariate linear regression model for predicting the spontaneous combustion risk of coal was subsequently established. The results demonstrated that −CH 2 − had a more significant effect on the spontaneous combustion risk index than that of −OH-a.