Graph Clustering Methods in Distributed Socio-Technical Systems

Socio-technical systems are characterized by dynamic relations between human actors, organizations, and technical components, requiring sophisticated mathematical and engineering methods for their modelling, analysis, and optimization. Graphs have been recognized as an efficient and powerful abstraction for modelling and analysing these systems, while graph clustering plays an important role in analysing and optimizing their structure and characteristics. A method for graph clustering based on attribute similarity of neighbouring nodes is introduced in this paper. The method contains two algorithms, the first one for checking similarity between neighbouring nodes and the second one for merging similar neighbouring nodes into clusters. The implementation of the method for determining the minimal number of air pollution measurement stations dispersed in urban city environments is presented to illustrate a possible scenario of use. Implementation is based on simulation of the method in a simulation environment that creates initial scenarios with measurement stations presented as graph nodes, create data sets for measurement stations, and calculates the minimal number of measurement stations by using the presented graph clustering method.