Superiorities of support vector machine in fracture prediction and gassiness evaluation

Multiple regression analysis (MRA), artificial neural network (ANN), and support vector machine (SVM) were applied to two case studies to contrast the application results. Case 1 is the fracture prediction based on studies of 34 samples from Wells An1 and An2 in the Anpeng Oilfield of the Biyang Sag, Nanxiang Basin. Case 2 is the gassiness evaluation of 40 samples in tight sandstones in the Tabamiao area, Ordos Basin. The results are as follows: (1) The nonlinear methods, SVM and ANN, are far superior to the linear method, MRA; (2) SVM presents absolute superiority due to zero error and fast speed, indicating that it is the best machine-learning method till date; (3) ANN is almost as accurate as SVM in Case 1, but ANN is less precise than SVM in Case 2; (4) MRA is fast and can establish the order of dependence between the study target and its related multi-geological-factors that cannot be estimated using SVM and ANN. Therefore, SVM is recommended when describing any complex relationship between a target and its related geological factors and MRA can be used as an auxiliary tool.