Data-driven fault classification for power converter systems has been taking more into consideration in power electronics, machine drives, and electric vehicles. It is challenging to classify the different topologies of faults in the real-time monitoring control systems. In this paper, a data-driven and supervised machine learning-based fault classification technique is adopted by combining and consolidating with Expectation Maximisation Principal Component Analysis (EMPCA) and Support Vector Machine (SVM) to substantiate the availability of fault classification. The proposed methodology is applied to the non-inverting Buck–Boost DC–DC power converter systems subjected to the incipient fault and serious fault, respectively. Finally, the feasibility of the approach is validated by intensive simulations and comparison studies.