The classification of well-log responses into separate flow units for generating local permeability models is often used to predict the spatial distribution of permeability in heterogeneous reservoirs. The present research can be divided into two parts; fi rst, the nuclear magnetic resonance (NMR) log parameters are employed for developing a relationship between relaxation time and reservoir porosity as well as introducing the concept of relaxation group. This concept is then used for the de fi nition of electrofacies in the studied reservoir. A graph-based clustering method, known as multi resolution graph-based clustering (MRGC), was employed to classify and obtain the optimum number of electrofacies. The results show that the samples with similar NMR relaxation characteristics were classi fi ed as similar groups. In the second part of the study, the capabilities of nonlinear support vector machine as an intelligent model is employed to predict the electrofacies and permeability distribution in the entire interval of the reservoir, where the NMR log parameters are unavailable. SVM prediction results were compared with laboratory core measurements, and permeability was calculated from stoneley wave analysis to verify the performance of the model. The predicted results are in good agreement with the measured parameters, which proves that SVM is a reliable tool for the identi fi cation of electrofacies through the conventional well log data.