Multiclass data classification using fault detection-based techniques

Multiclass classification of big data is a subject of broad interest in machine learning research nowadays, where it is necessary to extract important features from a dataset's variables in order to accurately detect unique variations between different classes of data. In this paper, we will discuss the application of a novel combination of different fault detection-based techniques towards the problem of multiclass classification of different types of faults found in the benchmark Tennessee Eastman Process. Moreover, the fault detection performance and data classification accuracy of our proposed method is compared to the respective performances of multiple data-driven methods tabulated in literature, including deep neural networks. The results show that a combined application of multiple fault detection techniques, in tandem with the one-versus-all and all-versus-all binary decomposition methods, can provide a competitive multiclass classification accuracy, comparative to other more complex methods in literature.