Levenberg-Marquardt (shorted as L-M) algorithm is improved and adopted to train the neural network. The improved training algorithm leads to better convergence, faster convergent speed and higher precision. The proposed L-M neural network is used for geological data mining of oil productive index basing on the geological database. The process of geological spatial data mining and the geological knowledge discovering with L-M neural network are discussed. As an engineering case, data mining and knowledge discovering of the oil productive index basing on the reservoir property stored in the geol...
