Application of support vector machine in lithological prediction of deep volcanic rocks in Songliao Basin

To evaluate the hydrocarbon potential and optimize the exploration deployment for the deep volcanic rocks in Songliao Basin, the study predicts the lithology of deep volcanic rocks by integrating regional gravity-magnetic data with seismic and well data. The boundary element gravity forward stripping method is applied to obtain the gravity anomaly effect reflecting deep geological bodies. Based on the integral iterative downward continuation, the flattening-curve method is developed to enhance and balance gravity-magnetic anomaly information related to volcanic rocks in deep fault depressions, which is then used to extract gravity-magnetic anomaly effects indicative of volcanic rocks in deep fault depressions, effectively obtaining the density and magnetic susceptibility reflecting the volcanic rocks in the basin through 2D physical property inversion. Constrained by lithology samples of the deep volcanic rocks from drilling, the artificial intelligence support vector machine (SVM) method is employed for effective lithology identification, achieving a cross-validation accuracy of 81.6%. This method is proved valuable application in geological mapping of covered areas.