BOPCD @ Northwestern
at Northwestern, I'm applying Bayesian Online Change Point Detection (BOCPD) to cryptocurrency price distributions — building a real-time system that identifies when market regimes shift, as distinct from when prices simply move.
what is change point detection
financial time series aren't stationary. prices, volatility, and correlations all shift over time, driven by macro events, regulatory changes, and structural shifts in market participation. identifying when these structural breaks occur — in real time, not retrospectively — is valuable for risk management and systematic trading. BOCPD maintains a running probability distribution over when the last change point occurred, updating continuously with each new data point.
application to crypto
cryptocurrency markets are especially prone to distributional shifts: regime transitions between bull and bear markets, liquidity crises, exchange failures, and DeFi contagion cascades. these shifts are fast and often irreversible. BOCPD applied to crypto price distributions enables real-time detection of these transitions as they happen — not after the damage is done.
p2p lending contagion
the broader research at Northwestern also investigates contagion default in peer-to-peer lending systems. the core question: how do individual borrower defaults propagate through lending networks, and when does systemic failure become likely? the goal is to understand how network topology — who lends to whom — determines the speed and extent of default cascades, and to identify early warning indicators before systemic failure.