Multiclass Support Vector Machines for Determination of Reinforcement Degree in Constructing Large-scale Structures Based on Data Characteristics

In recent years, there has been concern about the decrease in the labor force due to the declining birthrate and aging population, and the shortage of manpower has become a problem in the field of infrastructure development and maintenance as well. In this paper, we apply the machine learning methods to determination of the reinforcement degree when constructing large-scale structures, which requires the experience and skills of experts. Then, we formulate the tasks as a recognition problem based on data observed at a construction site, and propose two kinds of multiclass support vector machines by exploiting relevance between evaluation data observed in construction sites. In addition, we also show how to resolve the inconsistency between the data. Finally, we apply the proposed methods to real data of some large-scale structures in order to evaluate the formulation and the validity of the proposed methods.