Technical Case Study · Quant & Data

Market Sentiment and Financial Volatility Analyzer

8th-semester major project

Empirical econometrics and machine learning platform analyzing whether investor sentiment drives volatility across Nifty 50, BSE Sensex, and Bank Nifty (2007–2024). Combines Nelson’s Exponential GARCH (EGARCH) with a leak-free chronological Random Forest directional classifier and a tri-level real-data sentiment cascade.

Trading Horizon4,236+ Days (17 Yrs)

EGARCH Leverage γ-0.082 (p < 0.001)

RF Directional AUC0.65 – 0.72 AUC

Python 3.12arch (EGARCH 1,1)Scikit-Learn (RF & PCA)Statsmodels (Granger)GDELT v2 + VADERYFinanceStreamlit Cloud

Problem Statement

Symmetric GARCH models ignore the financial leverage effect where negative news shocks amplify volatility far more than positive shocks, and random K-fold splits leak future regimes into training.

Deterministic Architecture Advantage

Fits Nelson's asymmetric log-variance EGARCH(1,1) model via Maximum Likelihood alongside a tri-level sentiment cascade (GDELT v2, Google Trends PCA, VADER) and a chronological 80/20 Random Forest regime classifier across 17 years of Nifty 50, Sensex, and Bank Nifty data.

Execution Pipeline & System Topology

STEP 01

OHLCV & Tri-Level Sentiment

Ingests 2007–2024 Yahoo Finance log returns alongside GDELT, Google Trends PCA, and VADER tone.

yfinance + GDELT + PCA4,236+ Daily Observations

STEP 02

Asymmetric EGARCH(1,1)

Fits log-variance formulation capturing leverage effect (γ) and sentiment elasticity (δ).

arch Library + AIC/BICβ = 0.987 Persistence

STEP 03

Leak-Free Chronological RF

Trains 200-tree Random Forest on 15 lagged features using a strict 80/20 temporal split.

Scikit-Learn RF Classifier+12% to +18% Lift

STEP 04

Streamlit Cloud Analytics

Interactive 6-tab dashboard with Granger Causality tables, ROC curves, and crisis comparisons.

Streamlit Community CloudLive Public Deployment

Verification & Engineering Highlights

Benchmarks Symmetric GARCH(1,1), GJR-GARCH(1,1), and EGARCH(1,1) via Log-Likelihood, AIC, and BIC across Nifty 50, Sensex, and Bank Nifty.
Tri-level real-data sentiment cascade: GDELT v2 global news tone → Google Trends retail anxiety compressed via Principal Component Analysis (PCA) → VADER financial headlines.
Includes automated crisis comparative analysis contrasting the 2008 Global Financial Crisis (~24-month prolonged shock) against the 2020 COVID-19 liquidity shock (~89% annualized peak volatility).

core.py

python

from arch import arch_model
import numpy as np

def fit_egarch_volatility_model(returns_series):
    # Scale log returns for numerical stability during MLE convergence
    scaled_returns = returns_series.dropna() * 100.0
    model = arch_model(scaled_returns, vol="EGARCH", p=1, o=1, q=1, dist="normal")
    res = model.fit(disp="off")
    cond_vol = res.conditional_volatility / 100.0
    metrics = {
        "omega": float(res.params.get("omega", 0.0)),
        "alpha": float(res.params.get("alpha[1]", 0.0)),
        "gamma_leverage": float(res.params.get("gamma[1]", 0.0)),
        "beta_persistence": float(res.params.get("beta[1]", 0.0)),
        "aic": float(res.aic),
        "bic": float(res.bic),
    }
    return cond_vol, metrics