Model-independent analysis (MIA) methods are generally useful foranalysing complex systems in which relationships between theobservables are non-trivial and noise is present. Principal ComponentAnalysis (PCA) is one of MIA methods allowing to isolate components inthe input data graded to their contribution to the variability of thedata. In this publication we show how the PCA can be applied todigitised signals obtained from a cavity beam position monitor (CBPM)system on the example of a 3-cavity test system installed at theAccelerator Test Facility 2 (ATF2) at KEK in Japan. We demonstratethat the PCA based method can be used to extract beam positioninformation, and matches conventional techniques in terms ofperformance, while requiring considerably less settings and data forcalibration.