Legal metrology focuses on consumer protection from unreliable, inaccurate, and imprecise measurements. Measuring instruments subject to legal control must comply with the requirements of metrological normative acts. Reliable measurement results are used for economic transactions and in industry. Our objective in this paper was to predict instrument behaviour and failures through the analysis of measuring instruments database history. Data analysis using Machine learning tools for prediction of instrument behaviour (specific type of non-automatic weighing instrument) was made. The data set covers all measurement ranges in relation to their permissible errors. Used analysis provides insight into the reliability of the instruments’ measurement results for all users, but also for state authorities responsible for end consumers protection and supervision of the measurement system in the country.