volatility-forecasting

Systematic Volatility Research and Backtesting for equity options

TensorFlow implementation of the HARNet model for realized volatility forecasting.

Official code - M2VN(Multi-Modal Learning Network for Volatility Forecasting)

IBOVESPA volatility forecasting

An autonomous risk-overlay system simulating a hedge fund Investment Committee. Uses Multi-Agent Architecture (LangGraph) to validate algorithmic signals by combining deep-learning volatility forecasts (VolSense) with fundamental semantic reasoning and CVaR constraints.

A comprehensive analysis and forecasting project for Samsung stock data, utilizing historical data to build predictive models and analyze volatility.

A financial forecasting research prototype containing multiple competing forecasting approaches, with an LSTM price model currently being used by the Streamlit application.

Comparing the performance of the GARCH(1,1) model and historical volatility, close-to-close volatility, Parkinson volatility, Garman-Klass volatility and Rogers-Satchell volatility in the rolling window method to forecast future volatility on the NASDAQ composite.

FRE6123 (Financial Risk Management) Group Project: Volatility Forecast Using GARCH and Temporal Convolutional Networks

Independent R&D bridging classical financial econometrics and modern continuous-time deep learning. Projects on PINNs for Value-at-Risk and Neural SDEs for density forecasting. "Complexity must earn its place."

VolFlux is a quantitative framework for analyzing and forecasting financial market volatility using time series and statistical models (e.g. GARCH). It studies volatility dynamics across multiple asset classes to help quantify market risk.

A modular Python toolkit for advanced options pricing, volatility modeling, Greeks computation, and risk analysis. Includes Monte Carlo and Black-Scholes models, machine learning volatility surfaces, and interactive visualizations via Streamlit.

Out-of-sample volatility forecasting and Value-at-Risk backtesting for 14 currencies (2000–2026): GARCH/EGARCH/GJR vs. RiskMetrics, with QLIKE and Diebold-Mariano model comparison, Kupiec/Christoffersen VaR coverage tests, and sparse PCA on FX returns. Python.

A leak-free evaluation of the ARFIMA–Wavelet–LSTM hybrid: where the signal is, where it isn't, and why information beats architecture.

Forecasting realized volatility for 5 US sector ETFs using statistical models (ARIMA, GARCH, EGARCH), machine learning (Ridge, XGBoost, SVR), and deep learning (LSTM) on 20 years of daily data.

Global Industry Challenge 26

Forecasting 21-day realised volatility on the URA uranium ETF using a stacked LSTM network, benchmarked against a GARCH(1,1) baseline.

C++17 realised-volatility forecasting pipeline: Bloomberg BLPAPI → SQLite → feature engineering → Chronos-2 inference via ONNX Runtime, benchmarked against HAR-RV and persistence with Diebold-Mariano tests.

Regime-conditional volatility forecasting framework using HAR-RV as a baseline and XGBoost on either residual vol or directly on log(RV), implemented for Germany and France electricity markets. Metric: Spearman ranking. Model validation and market-neutral cross-country trading strategy.

Transformer for FX realized-volatility forecasting. Each hourly block encodes the joint market state (10 forex pairs + 14 macros + events + HAR features) into a single context vector; 24-horizon output for one target symbol.