Generative modeling for financial time series.
Skanalytix Pty Ltd is a Melbourne-based quantitative research company developing generative models for financial time series.
Our work focuses on generating plausible alternative market histories that capture important characteristics of observed market behaviour, with applications in scenario analysis, stress testing, portfolio risk analysis and model validation.
Data-driven, non-parametric generative modeling.
Skanalytix takes a data-driven, non-parametric approach to generative modeling. Rather than assuming a particular distribution for market returns or training a neural generative model, the approach draws directly on information contained in historical market observations.
As a simulated history evolves, the distribution of what happens next depends on the market situation that has developed up to that point. This allows generated histories to evolve conditionally while capturing important distributional, temporal and cross-asset characteristics of financial markets.
The objective is not to predict a single future path or replay the historical record. It is to generate a range of plausible alternative histories for exploring outcomes that extend beyond the single market history we have observed.
Andrew Skabar, PhD
Founder, Skanalytix Pty Ltd
Andrew Skabar founded Skanalytix in 2023 following more than two decades working in artificial intelligence, machine learning and decision systems, initially in academia and subsequently through independent research and development. He holds a PhD in artificial intelligence and has a background in physics and mathematics.
His research has centred on methods for learning from complex data and estimating relationships under uncertainty. This work eventually led to the development of the modelling framework underlying Skanalytix and its application to financial time series.
Skanalytix brings this background together with a particular focus on generative modeling: using observed financial data to construct plausible alternative histories without relying on rigid distributional assumptions.
Equity time series and portfolios.
Skanalytix is currently focused on generative modeling of equity time series and multi-asset portfolios, with particular attention to whether synthetic market histories capture important distributional, temporal and cross-asset characteristics of observed markets.
Current work examines properties including fat tails, volatility clustering, mean reversion and cross-asset dependencies, as well as how these characteristics interact as market conditions evolve.
The broader objective is to provide a flexible basis for generating and analysing alternative market scenarios for portfolio risk analysis, stress testing and model validation.