Structural Clustering of Volatility Regimes for Dynamic Trading Strategies

We develop a new method to find the number of volatility regimes in nonstationary financial time series by applying unsupervised learning its structure. use change point detection partition into locally stationary segments and then compute distance matrix between segment distributions. The are clustered learned discrete via an optimization routine. Using this framework, we determine clustering structure for indices, large-cap equities, exchange-traded funds currency pairs. Our overcomes rigid assumptions necessary implement many parametric regime-switching models, while effectively distilling several characteristic behaviours. results provide significant simplification these strong descriptive analysis prior behaviours volatility. Finally, create validate dynamic trading strategy that learns optimal match current distribution past regimes, thereby making online risk-avoidance decisions present.