A multiple kernel classification approach based on a Quadratic Successive Geometric Segmentation methodology with a fault diagnosis case. (March 2018)

• Record Type: Journal Article
• Authors: Honório, Leonardo M.; Barbosa, Daniele A.; Oliveira, Edimar J.; Garcia, Paulo A. Nepomuceno; Santos, Murillo F.
• Is Part Of: ISA transactions. Volume 74(2018)
• Page Start: 209
• Page End: 216
• Publication Date: 2018-03
• Subjects: Successive Geometric Segmentation -- Support Vector Machine -- Quadratic classification -- Multiple kernel classifier -- Transformer fault diagnosis; Engineering instruments -- Periodicals; Engineering instruments; Periodicals; Electronic journals; 629.805
• DOI: 10.1016/j.isatra.2018.01.013
• Languages: English
• ISSNs: 0019-0578

Abstract

Abstract: This work presents a new approach for solving classification and learning problems. The Successive Geometric Segmentation technique is applied to encapsulate large datasets by using a series of Oriented Bounding Hyper Box (OBHBs). Each OBHB is obtained through linear separation analysis and each one represents a specific region in a pattern's solution space. Also, each OBHB can be seen as a data abstraction layer and be considered as an individual Kernel. Thus, it is possible by applying a quadratic discriminant function, to assemble a set of nonlinear surfaces separating each desirable pattern. This approach allows working with large datasets using high speed linear analysis tools and yet providing a very accurate non-linear classifier as final result. The methodology was tested using the UCI Machine Learning repository and a Power Transformer Fault Diagnosis real scenario problem. The results were compared with different approaches provided by literature and, finally, the potential and further applications of the methodology were also discussed. Highlights: The Quadratic Successive Geometric Segmentation (QSGS) is a clustering and classification algorithm based on spatial density. Data is represented by oriented bounding hyper boxes (OBHB) which may present a linear separation surface. By using Kernel Density Estimation, it is possible to break and identify the main core of an OBHB. Each linear-based OBHB core will be used as kernel of a quadratic discriminant separation surface. The methodology can find its own topology by organizing and clustering the provided data.