Objective Accurate, easily accessible and economically viable cancer diagnostic tools are pivotal in improving the abysmal 5% survival rate of pancreatic cancer. Methods A novel, affordable, non‐invasive diagnostic method has been developed by combining measurement precision of infrared spectroscopy with classification using machine learning tools. Results Diagnosis accuracy as high as 90% has been achieved. The study investigated urine and blood from pancreas cancer patients and healthy volunteers, and significantly improved accuracy by focusing on sweet‐spots within blood plasma fractions containing molecules within a narrow range of molecular weights. The graphical table of contents depicts the flow of cancer information, data acquisition and diagnosis.