Membrane proteins are an important class of proteins that serve as channels, receptors, and energy transducers in a cell membrane. Knowledge of a given type of cell membrane protein is crucial for determining its function. This paper introduces an automated, in-silico method for identifying different types of membrane proteins based on their amino acid composition. Our method applies a novel, composite protein sequence representation that includes seven feature sets. The performance of the proposed method was tested on two large datasets and was compared with eight competing prediction methods. The results indicate that our method outperforms existing methods, provides improved predictions for the transmembrane protein types, and obtains 87% and 98% accuracy for the jackknife test and test on an independent dataset, respectively.