Research & Working Projects

This paper develops the Rolling Fractal Dimension (RFD) as a real-time measure of price-path geometric complexity. The central hypothesis is that rising local irregularity can serve as an early-warning signal for shifts in market structure, including volatility transitions, drawdowns, and structural breaks.

The framework is benchmarked against conventional indicators such as realized volatility, EWMA-type measures, GARCH-family models, Hurst-based indicators, entropy statistics, and change-point detection approaches.

Research objective: to propose a simple, interpretable, and operational framework for regime-shift detection that can complement standard risk-monitoring tools.

This ongoing project aims to construct a unified, data-driven Antifragility Index for financial markets. The project investigates whether antifragility can be measured in a coherent way using complexity-based diagnostics rather than relying only on volatility or drawdown-based risk views.

The proposed index combines fractal and multifractal indicators, long-memory measures, and recurrence-based statistics. Its empirical usefulness is evaluated through predictive frameworks such as local projections and logistic hazard models across multiple asset classes, including equities, foreign exchange, and crypto-assets.

Research objective: to build a measurable and interpretable antifragility metric that improves risk forecasting and regime-shift diagnostics across heterogeneous markets.

This dissertation examined how antifragile behaviour and regime instability can be detected and modelled using data-driven, complexity-based methods. It established the theoretical and empirical foundation for the ongoing research projects presented above.

The methodological framework combined fractal and multifractal analysis, recurrence quantification, Hurst-based measures, complexity feature extraction, and machine learning models including LSTM, Random Forest, and SVM.

Main findings: financial markets exhibit strong fractal and multifractal structure, especially during crisis periods; rolling fractal measures tend to rise ahead of major volatility transitions; and complexity-enhanced machine learning models show stronger predictive performance in turbulent regimes than baseline approaches.