Method of oil/gas prediction based on optimization of seismic attributes and support vector machine

Seismic attribute analysis technique is major studied content in oil/gas reservoir exploration and development. It should optimize the attribute group that is sensitive to oil and gas in work zone and no strong cross-correlation before carrying out the oil/gas prediction. The paper presented a new feature-selecting algorithm based on support vector machine. The class separability of kernel space is deduced by defining kernel feature similarity. The subset of attributes having most discriminating ability is selected iteratively based on the variation of class separability. In combination with support vector machine, the algorithm presented in the paper was applied to the issue of oil/gas prediction for Upper Yangxin Series carbonate reservoir in Sichuan Guanyinchang structure and G development block in Daqing Oilfield respectively. The predicted results proved the effectiveness of the method in the paper, which is able to become optional method in oil/gas prediction.