Biomarkers for the early diagnosis of pancreatic cancer (PC) are urgent needed. Plasma microRNAs (miRNAs) might be used as biomarkers for the diagnosis of cancer. We analyzed 361 plasma samples from 6 surgical centers in China and performed machine learning approach. We gain insight of the association between the aberrant plasma miRNA expression and pancreatic disease. 671 microRNAs were screened in the discovery phase and 33 microRNAs in the training phase and 13 microRNAs in the validation phase. After the discovery phase and training phase, 2 diagnostic panels were constructed comprising 3 microRNAs in panel I (miR-486-5p, miR-126-3p, miR-106b-3p) and 6 microRNAs in panel II (miR-486-5p, miR-126-3p, miR-106b-3p, miR-938, miR-26b-3p, miR-1285). Panel I and panel II had high accuracy for distinguishing pancreatic cancer from chronic pancreatitis (CP) with area under the curve (AUC) values of 0.891 (Standard Error (SE): 0.097) and 0.889 (SE: 0.097) respectively, in the validation phase. Additionally, we demonstrated that the diagnostic value of the panels in discriminating PC from CP were comparable to that of carbohydrate antigen 19-9 (CA 19-9) 0.775 (SE: 0.053) (P = 0.1 for both). This study identified 2 diagnostic panels based on microRNA expression in plasma with the potential to distinguish PC from CP. These patterns might be developed as biomarkers for pancreatic cancer.