Golgi membrane proteins contribute protein glycosylation, which is one of post-translational modification. It plays an important role in the cellular processes such as cell-cell adhesion, signal transfer, and subcellular localization. In this regard, the development of a computational method to discrimination of Golgi membrane proteins from the mammal genomes is desired. In this study, we succeeded in the feature extraction of the characteristics of Golgi type II membrane proteins (GLs) by comparison with post-Golgi type II membrane proteins (PGs) except the Golgi retention type mainly localized in the plasma membrane. The nonredundant datasets of GLs (344 sequences) and PGs (356 sequences) were obtained from Swiss-Prot release 57.0. GLs were detected by combining hydropathy alignment and position-specific score matrix (PSSM) around the transmembrane region. Each sequences were aligned by superpositioning the highest average hydrophobicity position. The PSSM was estimated based on position-specific amino acid propensities of the alignment position in the region of -14 to +18. Our method can discriminated GLs from PGs with 96.2% sensitivity and 93.5% specificity in a self-consistency test. Furthermore, 89.8% sensitivity and 87.0% specificity were achieved in 5-fold cross-validation test.