The seismic attribute can be subtracted from seismic data, and then it is specially used for the measurement of seismic data in its geometry, dynamics or statistics feature. The purpose of seismic multi-attributes inversion is to provide reliable foundation data for reservoir prediction. There are two methods in forecasting well logging characteristics from seismic data: linear regression and nonlinear method. In nonlinear mode, neural network or support vector machine is trained, using the selected attributes as inputs. Seismic multi-attributes nonlinear inversion results match the geologic distribution regulation much more than linear regression. Taking spontaneous potential (SP) curve as the inversion goal, firstly, we can find the optimal seismic attributes for SP inversion by using linear regression. Secondly, the multi-attributes inversion is carried out by using the multilayer feed forward network (MLFN) based on these seismic attributes, and acquiring 3-D SP data volume. Finally, the progradation process of delta-front sandbody in Dongying delta are displayed clearly through seismic slice along the layer. The practical application results show that the proposed method has a good performance and is beneficial for further application and application.