📈 Financial Market Analytics Pipeline

End-to-end quantitative analytics system for multi-asset financial markets — combining technical analysis, statistical feature engineering, and machine learning for market direction forecasting.

---

🗂️ Project Overview

This project implements a production-grade financial data analytics pipeline covering:

Stage	Description
Data Generation	Realistic synthetic OHLCV data via Geometric Brownian Motion with regime changes
Feature Engineering	40+ technical indicators (RSI, MACD, Bollinger Bands, ATR, OBV, momentum)
Machine Learning	Time-series-safe train/val/test splits, 3 classifiers, full evaluation suite
Visualization	Interactive Plotly dashboards: candlesticks, correlation heatmaps, volatility charts

Assets covered: AAPL · TSLA · MSFT · NVDA · GOOGL · BTC-USD · ETH-USD · SOL-USD

---

🚀 Quickstart

# Clone and install
git clone https://github.com/thed700/financial-market-analytics.git
cd financial-market-analytics
pip install -r requirements.txt

# Run the full pipeline
python main.py

# Run tests
pytest tests/ -v

After running, open any file in reports/ in your browser for interactive charts.

---

📁 Project Structure

financial-market-analytics/
│
├── main.py                    # Pipeline orchestrator
├── requirements.txt
│
├── src/
│   ├── data_generator.py      # GBM synthetic OHLCV generator
│   ├── feature_engineering.py # Technical indicators & features
│   ├── ml_models.py           # ML training, evaluation, feature importance
│   └── dashboard.py           # Plotly interactive visualizations
│
├── data/                      # Generated CSVs (gitignored)
│   ├── market_data.csv
│   ├── market_features.csv
│   ├── model_metrics.csv
│   └── feature_importance.csv
│
├── reports/                   # HTML dashboards (gitignored)
│   ├── candlestick_AAPL.html
│   ├── correlation_heatmap.html
│   ├── volatility_comparison.html
│   ├── model_comparison.html
│   └── ...
│
└── tests/
    └── test_pipeline.py       # 12 unit + integration tests

---

🔬 Technical Indicators Implemented

Trend

• SMA — Simple Moving Average (7, 14, 21, 50, 200-day)
• EMA — Exponential Moving Average (7, 14, 21, 50-day)

Momentum

• RSI — Relative Strength Index (14-day)
• MACD — Moving Average Convergence/Divergence (12-26-9)
• Momentum — Rolling price change (5, 10, 21-day)

Volatility

• Bollinger Bands — Width, %B position (20-day, 2σ)
• ATR — Average True Range, normalized (14-day)
• Rolling Volatility — Annualized (5, 10, 21-day)

Volume

• OBV — On-Balance Volume
• Volume Ratio — Current vs 20-day average

Price Structure

• Candlestick body size, upper/lower shadow, daily range
• Log returns, daily returns, bullish/bearish flag

---

🤖 Machine Learning

Methodology

• No data leakage: strict chronological train (70%) / val (15%) / test (15%) split
• Target: 5-day forward price direction (binary: up/down)
• All features computed using only past data

Models

Model	Accuracy	ROC-AUC
Logistic Regression	~51%	~0.52
Random Forest	~48%	~0.49
Gradient Boosting	~51%	~0.51

Note: Market direction forecasting is inherently close to random (~50%) for efficient markets. These results are realistic and expected for raw technical indicators on synthetic data. The project demonstrates the full ML workflow correctly.

---

📊 Visualizations

All charts are fully interactive (zoom, hover, export):

• Candlestick charts — OHLCV with overlaid SMA, EMA, Bollinger Bands, volume, RSI, MACD
• Correlation heatmap — Pairwise return correlations across all 8 assets
• Volatility comparison — Rolling annualized volatility by asset
• Normalized returns — Cumulative performance starting from base 100
• Model comparison — Grouped bar chart: accuracy, AUC, precision, recall
• Feature importance — Top-20 features ranked by tree-based importance

---

🧪 Tests

tests/test_pipeline.py  — 12 tests
  TestDataGenerator       (5 tests)  OHLCV validity, price positivity, asset coverage
  TestFeatureEngineering  (5 tests)  RSI bounds, MACD columns, ATR, target encoding
  TestIntegration         (2 tests)  Smoke test, no-infinity check

---

🛠️ Tech Stack

• Python 3.10+ — Core language
• Pandas / NumPy — Data manipulation and numerical computing
• Scikit-learn — ML models, pipelines, evaluation metrics
• Plotly — Interactive financial charts and dashboards
• Pytest — Unit and integration testing

---

📌 Key Skills Demonstrated

• Quantitative financial data modeling (GBM simulation)
• Time-series feature engineering without look-ahead bias
• Production-style modular Python code with type hints and docstrings
• Full ML workflow: training, validation, evaluation, interpretability
• Interactive data visualization for business stakeholders
• Automated testing with pytest

---

👤 Author

Akmal — @thed700

Economics & Data Analytics student | Aspiring Data Analyst

---

📄 License

MIT License — feel free to use, fork, and build on this.