Embedded Files
[BF-089] Pairwise Time Series Transformation: Copula Models, Mean-Reversion Metrics, and Future Directions in Paris Trade (SAS MWSUG Conference 2025)
[BF-089] Pairwise Time Series Transformation: Copula Models, Mean-Reversion Metrics, and Future Directions in Paris Trade (SAS MWSUG Conference 2025)
BF089MWSUG.pptxBF089PAPERMWSUG.docxReport
Report
I developed a command-line tool (available on Unix/Linux, MacOS, Windows) for quantitative finance that analyzes stock pairs using both mean-reversion models (Kalman Filter, z-scores, hedge ratios) and copula-based methods for capturing nonlinear dependencies. A report is on the right.
Paris Trade.docxKey features include:
Data pipeline: Automated retrieval of historical stock data via Yahoo Finance API, processed with Pandas and Dask for scalable analysis.
Strategy design: Implemented Paris (positive correlation) and All-Weather (negative correlation) strategies with dynamic spread estimation.
Statistical modeling:
Mean-reversion approach: Kalman Filter to estimate hedge ratios and generate trading signals.
Copula approach: Used Gaussian KDE to fit marginal distributions and applied Frank copulas to capture joint tail dependencies.
Visualization: Generated interpretable plots of spreads, z-scores, and cumulative returns with Matplotlib.
Configurable CLI: JSON-based configuration for tickers, thresholds, partitions, and strategy modes.
Impact: This project demonstrates how mathematical modeling, probability theory, and modern Python libraries can be combined to design and test financial trading strategies. It deepened my expertise in applied statistics, time-series analysis, and algorithmic trading frameworks.
Page updated
Google Sites
Report abuse