Traditional methods of multivariate statistical analyses, used to recognize anomalous hydrocarbon signatures in petroleum exploration, have at least four shortcomings: (1) it is difficult to isolate anomalies where the data are not normally distributed; (2) it is difficult to separate distinct anomaly populations corresponding to distinct formation mechanisms while separating anomalies from background; (3) it is not fitting to present illustrations of multivariate anomalies on contour maps; and (4) it is not suitable for preparation for multivariate pattern recognition. These present serious obstacles to the application of exploration geochemistry in hydrocarbon exploration. This study demonstrates that the Back Propagation Artificial Neural Network (BP-ANN) with logic multiplication cluster analysis (a new cluster analysis proposed in this paper) overcomes the difficulties exhibited by traditional multivariate methods. The logic multiplication cluster analysis was designed to produce a training set for the BP-ANN. This approach was established on the basis of geochemical characteristics and origin of the various populations in geochemical data, such as background, micro-seepage anomalies and seepage anomalies. The topology of the BP-ANN was optimized using the outputs of the BP-ANN and the correct rate. In order to illustrate the multivariate anomalies recognized using BP-ANN on contour maps, we designed the expression functions for BP-ANN application in this field. With traditional methods of anomaly recognition, acid-extractable hydrocarbons in soils have not proven to be efficient indicators for hydrocarbon potential in East Anan Sag, Inner Mongolia. However, the BP-ANN has indicated that these indicators are efficient and that areas of East Anan Sag have potential reserves of petroleum.