A Proposed Methodology for Online Implementation of a Support Vector Machine With Applications in Power Transformers Event Classification
Abstract
The constant evolution of resources in computational processing and machine learning algorithms, combined with the increasing complexity of embedded systems, made the hardware implementation of machine learning models more viable. This paper proposes a methodology for online implementation of a support vector machine classifier through the development of a simple, concise, and easily adapted algorithm for data classification. The system was validated through the development of an application that classifies disturbances in a power transformer, followed by a comparison with the results obtained with the Library for Support Vector Machines (LIBSVM). Besides the very similar results when compared with the LIBSVM, the proposed methodology achieved high overall accuracy and fast classification time.