Clustering of Japanese Stock Returns by Recursive Modularity Maximization

This paper analyses high dimensional correlation structure of Japanese stocks to find a more data-oriented and flexible grouping than the Japan standard sector classification for better portfolio risk management. Modularity maximization and spectral clustering are employed to recursively divide GARCH filtered stock returns into subgroups. The standard sector classification is proved to be valid for group identification, though partially. Our method based on community detection can be applicable for clustering other fat-tailed financial asset returns.