one of the most prevalent tumors that are considered incurable is a pancreatic tumor. It is one of the most common polyps that are likely to be lethal. It’s been predicted to become the second deadliest disease by 2030. Currently, the Food and Drug Agency (FDA) approved only CA19-9 as a biomarker as a part of the screening program. But, the sensitivity and Specificity of CA19-9 are below 90%. The extensive growth of artificial intelligence techniques enables solutions in medical health systems including the automatic diagnosis of disease to monitor the progression of the disease. In this work, we proposed a novel panel of biomarkers LYVE1, REG1B, TFF1, HbA1C, CALCIUM, MAGNESIUM, ZINC, and COPPER through a modified random forest feature extraction technique. The missing values of the dataset are handled by the hybrid KNN and iterative imputation method, which gives a better standard derivation of about 0.0600. We also propose a modified Random Forest Machine learning Classifier, differentiating the pancreatic ductal Adenocarcinoma patients from healthy controls and Chronic Pancreatitis in the early stages such as Stage I and Stage II. The proposed techniques achieved a Sensitivity of 96.15% and a Specificity of 91.1% for the SS Institute of Medical Sciences and Research Centre dataset of 560 urine samples.