Global Market Intelligence

Multi-sector equity analytics platform — data ingestion, cleaning, EDA, forecasting (ARIMA + XGBoost), anomaly detection, and an interactive Streamlit dashboard across 36 blue-chip stocks.

Data source: Yahoo Finance via RapidAPI · Universe: 36 tickers across 6 sectors (Technology, Finance, Healthcare, Energy, Consumer, Industrials)

---

Dashboard Preview

Executive Overview	Forecasting	Anomaly Radar
KPI cards, sector heatmap	ARIMA vs XGBoost 90-day	Timeline + event table

---

Features

• Chunked ingestion of 5-year daily OHLCV + fundamentals (P/E, beta, EPS) via RapidAPI
• Multi-method outlier detection: IQR · Z-score · Isolation Forest · domain rules
• Feature engineering: lag/rolling windows, CAGR, HHI concentration, RSI, MACD, Bollinger Bands, ATR
• Dual forecasting: ARIMA (per-ticker, AIC-selected) + XGBoost (global model, 18 features, 90-day horizon)
• Anomaly detection: Isolation Forest + DBSCAN + residual-based, with SHAP explainability
• 7-page Streamlit dashboard with Plotly charts and full interactivity

Model Performance

Model	RMSE	MAE	MAPE	R²
ARIMA (per-ticker)	~$18	~$12	~3.2%	~0.89
XGBoost (global)	~$11	~$7	~1.8%	~0.96

Quick Start

git clone https://github.com/Jash-stack/global-market-intelligence.git
cd global-market-intelligence
pip install -r requirements.txt

# Optional: add your RapidAPI key (demo mode works without it)
cp .env.example .env

# Run full pipeline
python src/pipeline.py

# Launch dashboard
streamlit run app/dashboard.py

Project Structure

global-market-intelligence/
├── src/
│   ├── ingestion.py            # RapidAPI fetch + synthetic fallback
│   ├── cleaning.py             # Normalisation, interpolation, deduplication
│   ├── outlier_detection.py    # IQR, Z-score, IsolationForest, domain rules
│   ├── feature_engineering.py  # Lag, rolling, CAGR, HHI, RSI, MACD, BB, ATR
│   ├── forecasting.py          # ARIMA + XGBoost training & 90-day forecasts
│   ├── anomaly_detection.py    # IsolationForest + DBSCAN + residual-based
│   ├── evaluation.py           # RMSE/MAE/MAPE, SHAP plots, residual diagnostics
│   └── pipeline.py             # End-to-end CLI orchestrator
├── app/
│   └── dashboard.py            # 7-page Streamlit interactive dashboard
├── notebooks/
│   ├── 01_eda.ipynb
│   └── 02_modeling.ipynb
├── tests/
│   ├── test_cleaning.py
│   └── test_feature_engineering.py
├── Dockerfile
├── pyproject.toml
├── requirements.txt
├── .env.example
├── .gitignore
└── README.md

Tech Stack

Python 3.11 · pandas · numpy · scikit-learn · xgboost · statsmodels · shap · plotly · streamlit · pyarrow

Running Tests

pytest tests/ -v --cov=src

Docker

docker build -t global-market-intelligence .
docker run -p 8501:8501 global-market-intelligence

Author

Jash Shah — GitHub · LinkedIn

---

Data: Yahoo Finance via RapidAPI · License: MIT