Unsupervised spectral clustering for hierarchical modelling and criticality analysis of complex networks

Infrastructure networks are essential to the socioeconomic development of any country. This article applies clustering analysis to extract the inherent structural properties of realistic-size infrastructure networks. Network components with high criticality are identified and a general hierarchical modelling framework is developed for representing the networked system into a scalable hierarchical structure of corresponding fictitious networks. This representation makes a multi-scale criticality analysis possible, beyond the widely used component-level criticality analysis, whose results obtained from zoom-in analysis can support confident decision making.