Risk Analysis of Financial Time-Series Using Multi-Scale Entropy

Recently, there are growing concerns about the time-scale dependency of complexity in financial data. To evaluate the risk of financial data, we adopt a multi-scale entropy analysis, which can measure complexity with timescale dependency, to time-series of TOPIX during 1/7/1992-6/1/2016. The results confirm that sample entropy exhibits higher value, especially with large-scale factor, near main financial incidents. Furthermore, we classify a multi-scale entropy profile against the time-scale by K-means. Furthermore, we confirm that the multi-scale entropy profile during main financial incidents belongs to the class with larger 1 st component of principal component analysis. Therefore, we conclude that multi-scale entropy is a useful tool for evaluating the risk of financial data.