4. Conclusion
In conclusion, we have used the temporal network framework to analyze the temporal evolution of three major stock markets. The topology evolution of the correlation-based networks for three markets give some signals of corresponding financial turbulences in each market. With the help of temporal centrality measure, we can construct some risk diversified portfolios with high return and low risk. Under both the mean–variance and expected shortfall frameworks, the portfolios constructed with those peripheral stocks in both temporal and static centrality measures outperform those portfolios constructed with central stocks. Moreover, those peripheral portfolios selected with the guidance of temporal centrality measure performed way better than other portfolios(temporal central portfolios and aggregated peripheral portfolios) under both mean–variance and expected shortfall evaluation criterion. The in sample and out of sample tests have verified the robustness of the temporal peripheral portfolios. This is the first study to analyze the time evolving correlation-based networks with temporal network theory. The application of temporal centrality measure on portfolio selection has revealed the importance of the temporal attributes of the correlation-based networks of stock markets. Thus it should be quite interesting to investigate the temporal structure of the correlation-based networks with other tools developed for temporal network [16]. At the same instant, the investigation about the correlation-based networks of financial systems by using other tools from complex network theory, such as community detection [23,40–43] and network percolation theory [44], are also very promising directions. Those should subject to future researches.
