Multiclass piecewise-linear SVMs based on division of data space for determining the reinforcement degrees of large-scale structures

In this paper, we focus on automatical determination of the reinforcement degrees at construction sites of a kind of large-scale structure to address labor shortages. In the previous study, the reinforcement degree at each section of the structure can be appropriately determined by a linear discriminant function of the risk scores and natural condition in the simplest and monotonic form, which is obtained by training a multiclass linear support vector machine (SVM).Then, we propose multiclass piecewise linear SVMs (MPLSVMs) to obtain a piecewise linear discriminant function that uses different weights for each group of data, and we propose two kinds of methods of dividing data space: one based on the natural conditions, and the other based on the similarity of risk scores. In addition, we introduce a local search method tuning the boundaries of divided regions. Numerical experiments demonstrate the superiority of the proposed MPLSVMs.