Thermodynamic motif analysis for directed stock market networks

• This paper presents a novel thermodynamic framework for directed time evolving networks, especially in financial domain, by building the connection between graphs and dilute gas system. The motif successfully combines structural characteristic of motifs with their statistical characteristics to represent entire complex networks. based on performs well revealing evolution series network detection crisis events network. In this paper, we present thermodynamically analysis method particular time-evolving networks finance domain. Based an analogy mechanics, develop partition function composed motifs. relies decomposition into frequently occurring graphlets, or According gas, have same topological structure as low-order interactions particles gas. means that can use so-called cluster expansion from mechanics decomposition. prior work, reported detailed case undirected graphs, showed how resulting entropy be used analyse [1]. extend work compute quantities including energy, temperature three constitute evolution. We apply our biological domains real world systems time-varying Experimental results demonstrate effectiveness representing anomalous event detection.