A half of melanomas in non-white populations are from acral volar area and an appearance of these lesions is completely different from that of other parts of lesions. However, no research for classifying these lesions has been conducted. In this paper, we describe a diagnosis classifier of acral volar melanomas. We used our dermatologist-like tumor area extraction algorithm and extracted a total of 428 features from 199 acral dermoscopic images (169 nevi and 30 melanomas) and built a linear classifier. Our classifier selected 17 features and achieved a sensitivity of 90.0%, a specificity of 92.3% and an area under the ROC curve (AUC) value of 0.944 using a leave-one-out cross-validation strategy.