SignalAI — Stock & ETF Signal Generation Platform

A full-stack platform that uses a global XGBoost model to generate BUY / HOLD / SELL signals for Indian stocks and ETFs. Built with Flask, React (Vite + TypeScript), and MongoDB.

---

Overview

SignalAI trains a single unified ML model across hundreds of tickers (Nifty 500 + core stocks) and exposes predictions through a REST API. A React SPA provides dashboards for signals, backtesting, portfolio analytics, alerts, and live market data.\

https://developmentof-python-based-stock-et.vercel.app/

Tech stack:

Layer	Technology
ML / Backend	Python 3.10+, XGBoost, scikit-learn, yfinance, Flask
Database	MongoDB
Frontend	React 18, Vite, TypeScript, Tailwind CSS
Auth	JWT

---

Project Structure

├── backend/
│   ├── app.py                    # Flask entry point
│   ├── train_global_model.py     # Model training script
│   ├── requirements.txt
│   ├── models/                   # Saved model artifacts
│   │   ├── global_model.pkl
│   │   ├── global_scaler.pkl
│   │   └── global_features.pkl
│   ├── routes/                   # Flask route blueprints
│   ├── services/
│   │   ├── data_fetcher.py       # yfinance fetching with .NS/.BO fallbacks
│   │   └── ...
│   └── database.py               # MongoDB connection & collections
│
└── front_end/
    ├── src/
    │   ├── App.tsx
    │   └── pages/                # Dashboard, SignalEngine, Backtesting, etc.
    └── package.json

---

Prerequisites

• Python 3.10+
• Node.js 18+ and npm
• MongoDB (local instance or Atlas URI)

---

Setup & Run

1. Backend

# Create and activate a virtual environment
python -m venv .venv

# Windows
.venv\Scripts\activate
# macOS / Linux
source .venv/bin/activate

# Install dependencies
pip install -r backend/requirements.txt

Create a .env file in the repo root (optional — defaults are provided):

MONGO_URI=mongodb://localhost:27017/    # MongoDB connection string
JWT_SECRET=your_secret_here            # JWT signing secret

Start the Flask server:

python backend/app.py
# Runs at http://127.0.0.1:5000

2. Frontend

cd front_end
npm install
npm run dev
# Runs at http://localhost:5173 (or as printed by Vite)

---

Model Training

Pre-trained artifacts are included in backend/models/. To retrain from scratch:

python backend/train_global_model.py

This downloads historical data for Nifty 500 + core tickers, engineers features, trains a global XGBoost classifier, and saves three artifacts:

File	Description
global_model.pkl	Trained XGBoost classifier
global_scaler.pkl	Feature scaler
global_features.pkl	Feature column names

Note: Training downloads data for many tickers and can take significant time. Rate-limit protection for yfinance is built in.

Labels are generated from next-day return thresholds:

• BUY — return above upper threshold
• HOLD — return within neutral band
• SELL — return below lower threshold

---

API Reference

Base URL: http://127.0.0.1:5000

Method	Endpoint	Description
GET	/	Health check & available endpoints
GET	/stocks	List of available tickers (Nifty 500 + core)
GET	/predict/<ticker>?email=<email>	BUY/HOLD/SELL signal + confidence + latest price
GET	/api/market/overview	Price, change %, market cap for core Indian tickers
POST	/api/backtest	Run RSI strategy backtest (see body below)
POST	/api/alerts	Create a price alert
GET	/api/alerts?email=<email>	List user's active alerts
DELETE	/api/alerts/<id>	Remove an alert
POST	/api/auth/register	Register a new user
POST	/api/auth/login	Login and receive a JWT token
POST	/api/user	Update user profile (name)

Backtest request body

{
  "startDate": "2020-01-01",
  "endDate":   "2024-01-01",
  "initialCapital": 100000
}

The backend runs an RSI Mean Reversion strategy against the requested date range using yfinance data.

---

Features

• ML Signal Engine — enter any NSE ticker and receive a BUY / HOLD / SELL signal with confidence score, powered by the global XGBoost model.
• Backtesting — test the RSI strategy against historical data over any date range with performance metrics (Sharpe ratio, max drawdown, win rate, CAGR).
• Portfolio Analytics — overview of positions and P&L.
• Alerts — set price-level alerts; a background daemon thread checks conditions every 30 seconds and writes notifications to MongoDB.
• Market Data — live table of Nifty large-cap stocks with price, change %, volume, market cap, P/E, and sector.
• Authentication — register / login with JWT; user profile stored in MongoDB.

---

Data Source

All market data is fetched via yfinance (Yahoo Finance). No API key is required. The fetcher automatically appends .NS or .BO suffixes for Indian tickers and falls back to yf.download if the primary method fails.

In production, consider a paid data provider or a caching proxy to avoid Yahoo Finance rate limits.

---

License

This project is a demo / proof-of-concept. See license.txt.