Mohammad Valizadehmoghadam

It analyzes unstructured news to surface emerging fraud and scam trends across jurisdictions, automates monitoring workflows, and supports proactive operational risk stakeholders with API-driven delivery.

The notebook covers instrument pricing and sensitivities, historical simulation for VaR and Expected Shortfall, VaR backtesting, a severe macro stress scenario, risk limits, DV01 by tenor, factor correlations, and FRTB liquidity horizon scaling.

It follows the same order of steps a desk would use so the story stays grounded in practice.

The index combines long-horizon cleaned data, GARCH-based resilience signals, and anomaly detection (e.g. Isolation Forests, K-Means) to produce stability scores and flag outliers compared with traditional econometric baselines.

I combined returns, fundamentals, and index data into a panel and replicated seven style factors. Each month names land in five equal weight quintiles; the factor return is the spread between top and bottom. I compare to Capital IQ with Newey-West inference, apply a turnover haircut, and include a book to price case study plus subsample checks.

The work evaluates alternative risk transfer (ART) structures, benchmarks peers, scenarios for setup and annual fees, and highlights capital optimization opportunities for senior stakeholders.

Each fund’s excess returns are regressed on MKTRF, SMB, HML, UMD, RMW, and CMA. Portfolios include long-only and long–short global minimum variance and max-Sharpe, a tracking-error-capped max-Sharpe versus equal weight, and a factor-neutral max-Sharpe (±200% per asset). Inputs are estimated in-sample (Jan 2010–Dec 2019) and evaluated out-of-sample (Jan 2020–Dec 2024) without re-estimation; reported statistics are annualized.

A consistent takeaway: unconstrained max-Sharpe and long-short designs overfit in-sample; GMV and tracking-error-bounded strategies tend to generalize better out-of-sample because they rely less on hard-to-estimate expected returns. Hedge fund indices show the most stable in-sample to out-of-sample transfer among the three universes.

Fund return data is bundled in the repo with permission; factor data draws on Kenneth R. French’s library. Data-quality work (e.g. correcting a bad monthly return for FSPHX) is documented in the notebook. Completed as part of the Master of Financial Risk Management (MFRM) program, Rotman School of Management, University of Toronto.