Clustering of sports fields as specific construction objects aided by Kohonen’s neural networks
Michał Juszczyk, Krzysztof Zima · AIP conference proceedings · 2018
Sports fields can be considered as a type of sport facilities that must be built in the course of a construction project. The paper presents some results of a broader research on the problem of cost predicting for such objects supported by various mathematical tools. The Kohonen’s neural networks (also known as self-organizing map or self-organizing feature maps) are explored for the purpose of clustering data including characteristic parameters of sports fields built in Poland. Kohonen’s neural networks were applied to perform the transformation of n-dimensional input data space into a two dimensional in a map. In the course of the research, the data including characteristic parameters of sports fields was presented to the number of networks that varied in the configuration of an output layer. The two dimensional topologically ordered feature map of data clusters that describe groups of similar sports fields is proposed as a result of the analysis.