Porosity prediction using the group method of data handling

摘要

The relationship between the seismic data and the reservoirproperties can be modeled by using statistical approaches,such as regression and artificial neural networks(ANN); however, another nonlinear regression method,known as the group method of data handling (GMDH), hasbeen proven to perform better than regular statistical methods.GMDH is a supervised machine learning tool that automaticallyself-organizes (synthesizes) the models. Althoughit is self-organized, like unsupervised ANN, it learns fromthe examples introduced similar to the supervised ANN. Weapply the GMDH algorithm to seismic attributes to predictreservoir porosity. GMDH can automatically determine thebest network structure, as well as the number of nodes, thusgauging sensitivity of the input without overtraining thedata. Moreover, GMDH predicted porosity has better resolutionthan that predicted using ANN.