Membrane proteins can be classified among the following five types: (i) type I membrane protein. (ii) type II membrane protein. (iii) multipass transmembrane proteins. (iv) lipid chain-anchored membrane proteins, and (v) GPI-anchored membrane proteins. T. Kohonen?s self-organization model which is a typical neural network is applied for predicting the type of a given membrane protein based on its amino acid composition. As a result, the high rates of self-consistency (94.80%) and cross-validation (77.76%), and stronger fault-tolerant ability were obtained.