At1uri/stochastic-financial-benchmarks

📊 Stochastic Financial Time Series Benchmarks

A quantitative benchmark dataset curated and generated by Ramprasad Atluri for validating statistical and machine learning time-series models.

📌 Dataset Overview

This dataset contains simulated continuous-time diffusion processes sampled daily across 1,000 trading periods. It is designed to evaluate model robustness across different market regimes:

1. Geometric Brownian Motion (GBM): Baseline log-normal continuous diffusion process with calibrated drift ($\mu = 0.09$) and volatility ($\sigma = 0.20$).
2. Ornstein-Uhlenbeck (OU) Mean Reversion: Stationary mean-reverting process modeling spreads, commodity pairs, and interest rate corridors.
3. Merton Jump Diffusion: Mixed compound Poisson jump process testing model resilience to non-Gaussian tail events, volatility clustering, and market shocks.

🚀 Usage

You can load this dataset directly in Python using the datasets library or pandas:

import pandas as pd

url = "https://huggingface.co/datasets/At1uri/stochastic-financial-benchmarks/raw/main/stochastic_benchmarks.csv"
df = pd.read_csv(url, parse_dates=["date"])
print(df.head())

🛠️ Field Descriptions

Column	Type	Description
date	String (YYYY-MM-DD)	Calendar trading timestamp
geometric_brownian_motion	Float	Continuous log-normal asset price path
ornstein_uhlenbeck_spread	Float	Mean-reverting stochastic spread
merton_jump_diffusion	Float	Jump-diffusion process exhibiting fat-tailed behavior

⚖️ License

MIT License. Free for research, commercial, and educational use.