📈 Algorithmic Trading System

Python · Streamlit · Event-Based ML · Telegram Integration

---

📌 Project Overview

This project is an event-driven algorithmic trading analysis system built using Python, Machine Learning, and Streamlit.

It analyzes NIFTY 50 equity data, transforms raw OHLCV market data into interpretable trading events, and uses a RandomForest classifier to estimate the probability of next-day price movement.

Unlike black-box trading systems, this project prioritizes:

• ✅ Explainability
• ✅ Feature transparency
• ✅ Modular architecture
• ✅ Production-safe data handling

Outputs are visualized through an interactive Streamlit dashboard, logged to Google Sheets, and optionally accessible via a Telegram bot interface.

---

🧠 What This System Does (In One Paragraph)

The system ingests daily stock data, engineers event-based features (trend shifts, breakouts, volume confirmation, momentum, volatility), and trains a RandomForest model to predict the probability that tomorrow’s closing price will be higher than today’s.

The model does not auto-trade. It provides probabilistic decision support to help evaluate favorable market conditions.

---

🚀 Core Capabilities

📊 1. Automated Data Pipeline

• Daily OHLCV data via yfinance
• Data sanitization & preprocessing
• Multi-ticker support

📈 2. Event-Based Strategy Engine

Signals are derived from structural market conditions:

• SMA (20/50) crossover detection
• 20-day breakout logic
• Volume spike & trend confirmation
• Momentum & volatility regime detection

No fragile RSI-based heuristics.

---

🤖 3. Machine Learning Layer

• Model: RandomForestClassifier
• Problem Type: Supervised classification
• Target Variable:

1 → Tomorrow’s Close > Today’s Close  
0 → Otherwise

Why RandomForest?

• Handles non-linear interactions
• No feature scaling required
• Stable on tabular financial data
• Produces reliable probability estimates

---

📊 4. Interactive Streamlit Dashboard

• Multi-ticker selection
• Backtesting over custom date ranges
• Visualized SMA crossovers
• Buy/Sell markers
• ML accuracy display
• Indicator overlays using Plotly

---

📄 5. Google Sheets Logging

Not Working!

• Signal snapshots
• Model accuracy tracking
• Optional trade journal logging

---

🔔 6. Telegram Bot (Optional)

• Query crossover events
• View stock summaries
• Get signal breakdowns
• Designed for explainable outputs (not spam alerts)

---

🧩 Feature Groups Used for Prediction

Each trading day is treated as a market snapshot.

1️⃣ Trend & Structure

• ma_crossover
• strong_trend
• ma_diff

2️⃣ Breakout & Price Action

• breakout_20d
• price_sma20_diff
• price_sma50_diff

3️⃣ Volume Confirmation

• volume_spike
• volume_ma_ratio
• volume_change
• Normalized OBV

4️⃣ Momentum

• momentum_5d
• MACD
• MACD_hist

5️⃣ Risk / Regime

• volatility_5d

The model learns which combinations historically led to upward movement.

---

🛠️ Tech Stack

Layer	Technology
Language	Python ≥ 3.10
Data	yfinance
ML	scikit-learn (RandomForest)
UI	Streamlit + Plotly
Logging	Google Sheets (gspread)
Alerts	Telegram Bot API

Design Choices:

• ❌ No pandas-ta
• ❌ No RSI
• ✅ Stable dependency structure
• ✅ Fully modular architecture

---

📂 Project Structure

algo-trading-system/
│
├── main.py          # End-to-end pipeline
├── ui.py            # Streamlit dashboard
├── strategy.py      # Event-based signal logic
├── ml_model.py      # Feature engineering + ML model
├── sheets.py        # Google Sheets logging
├── bot.py           # Telegram bot integration
├── utils.py         # Logging utilities
├── requirements.txt
└── .gitignore

---

⚡ Quick Start

1️⃣ Clone the Repository

git clone https://github.com/Aryanonghub/algo-trading-system.git
cd algo-trading-system

2️⃣ Create Virtual Environment

python -m venv venv
source venv/bin/activate   # macOS/Linux
.\venv\Scripts\Activate.ps1  # Windows

3️⃣ Install Dependencies

pip install -r requirements.txt

---

💻 Running the System

▶ Streamlit Dashboard (Recommended)

python -m streamlit run ui.py

▶ CLI Pipeline

python main.py

▶ Telegram Bot

python bot.py

Make sure to set:

export BOT_TOKEN=your_telegram_token

---

🧭 Conceptual Architecture

[Streamlit UI]
        |
        v
[Data Fetcher - yfinance]
        |
        v
[Feature Engineering Engine]
        |
        +--> Google Sheets Logging
        +--> Telegram Queries
        |
        v
[RandomForest Model]
        |
        v
[Probability Output + Visual Dashboard]

---

📊 Expected ML Accuracy

Financial prediction is inherently noisy.

Typical realistic accuracy:

55% – 65%

Anything consistently above this in live conditions is strong.

---

📅 Roadmap

• Walk-forward validation
• Feature importance export
• Full backtesting engine (PnL, drawdown)
• Position sizing logic
• Risk management module
• Dockerization
• Unit testing
• Broker API integration (Zerodha / Upstox)

---

⚠️ Disclaimer

This project is for educational and research purposes only.

It does not constitute financial advice. Trading involves risk. Past performance does not guarantee future results.

---