A Novel similarity measure based on eigenvalue distribution

Due to the rapidly increasing interests of effective and efficient data processing, the developments of similarity measure have been significantly expanded. This paper defines the eigenvalue distribution as a criterion of measuring similarity in a multivariate system. The primary evaluations are conducted by simulations with the assistances and comparisons of several empirical statistical tests. Furthermore, the proposed measure is conducted in simultaneous real case scenario by adopting the bootstrap re-sampling technique. It also overcomes the difficulty of different series lengths in the multivariate system. Moreover, it does not have pre-assumptions on distributions, and it can be easily employed and efficiently computed.