Upvestment

An AI-powered web application that predicts short-term S&P 500 (SPY) market trends using machine learning and live market data.

Disclaimer: For educational purposes only. Not financial advice.

---

Features

• One-click prediction — fetches live SPY data from Yahoo Finance and runs the model instantly
• 9 normalised technical indicators — SMA ratios, RSI, MACD %, Bollinger Band width, volatility, momentum
• 55%+ model accuracy — RandomForest trained on 10 years of SPY data
• 30-day backtest — see how the model performed on the last 30 real trading days
• Signal breakdown — RSI level, SMA cross, daily return, and momentum interpretation
• Feature importance chart — see which signals drive the model's decisions
• 15-minute data cache — avoids hammering Yahoo Finance on repeated requests
• Advanced Mode — manually enter all 9 indicators to test hypothetical conditions
• Production-ready — Docker support, env-based CORS, frontend served from FastAPI

---

Project Structure

Upvestment-Project/
├── app.py               # FastAPI backend — all API endpoints
├── features.py          # Feature engineering (indicators + normalisation)
├── train.py             # Model training script
├── requirements.txt     # Python dependencies
├── Dockerfile           # Multi-stage Docker build
├── docker-compose.yml   # Single-command deployment
│
├── frontend/            # React + Vite frontend
│   └── src/
│       ├── App.jsx
│       ├── components/
│       │   ├── InputForm.jsx        # Manual indicator form (Advanced Mode)
│       │   ├── PredictionResult.jsx # Result + signal breakdown
│       │   ├── Charts.jsx           # Feature importance bar chart
│       │   └── HistoryChart.jsx     # 30-day backtest table
│       └── services/api.js          # API client
│
├── models/              # Trained model artifacts
│   └── features.json    # Feature column schema
├── data/                # Processed datasets (gitignored)
└── reports/             # Metrics
    └── metrics.json

---

Setup

Requirements

• Python 3.10+
• Node.js 18+

Install Python dependencies

python -m venv .venv
source .venv/bin/activate        # macOS/Linux
# .venv\Scripts\Activate.ps1    # Windows PowerShell
pip install -r requirements.txt

Train the model

Downloads 10 years of SPY data from Yahoo Finance and trains the model:

python train.py

Artifacts saved to models/ and reports/.

---

Running Locally

Backend (Terminal 1)

source .venv/bin/activate
uvicorn app:app --host 0.0.0.0 --port 8000 --reload

API available at http://localhost:8000
Interactive docs at http://localhost:8000/docs

Frontend (Terminal 2)

cd frontend
npm install       # first time only
npm run dev

Frontend available at http://localhost:5173

The backend must be running before you use the app — it handles live data fetching and predictions.

---

Running with Docker

Build and run the full stack in a single container:

docker compose up --build

App available at http://localhost:8000

The React frontend is built and served directly by FastAPI — no separate server needed.

Note: Train the model before building the Docker image (python train.py) so models/ is populated.

---

API Endpoints

Method	Endpoint	Description
GET	/	Welcome message and feature schema
GET	/health	Model status, cache age, last data date
GET	/live-indicators	Fetch live SPY indicators from Yahoo Finance
POST	/predict	Run model prediction on provided features
GET	/feature-importance	Feature importances from the trained model
GET	/backtest?days=30	Model predictions vs actuals for last N trading days

POST /predict example

// Request
{
  "sma_10_ratio": 0.003,
  "sma_50_ratio": 0.012,
  "daily_return": 0.0031,
  "rsi": 57.4,
  "macd_pct": -0.0001,
  "macd_signal_pct": -0.00008,
  "bb_width": 0.045,
  "volatility_20": 0.009,
  "momentum_5": 0.018
}

// Response
{
  "prediction": "Up",
  "prob_up": 0.63
}

---

Model Details

Property	Value
Algorithm	RandomForestClassifier
Training data	10 years of SPY (via yfinance)
Features	9 normalised technical indicators
Test accuracy	~55%
Train/test split	80/20 chronological

Features

Feature	Description
sma_10_ratio	Price deviation from 10-day SMA: close/sma_10 − 1
sma_50_ratio	Price deviation from 50-day SMA: close/sma_50 − 1
daily_return	Today's percentage price change
rsi	14-day Relative Strength Index (0–100)
macd_pct	MACD line as fraction of price
macd_signal_pct	MACD signal line as fraction of price
bb_width	Bollinger Band width normalised by mid band
volatility_20	20-day rolling standard deviation of daily returns
momentum_5	5-day price momentum (percentage change)

---

Environment Variables

Variable	Default	Description
CORS_ORIGINS	http://localhost:5173,...	Comma-separated allowed origins

Set in production:

export CORS_ORIGINS=https://yourdomain.com

---

Deployment (AWS EC2)

1. SSH to your Ubuntu instance and install dependencies:

sudo apt update && sudo apt install -y docker.io docker-compose-v2

2. Clone the repo, train the model, then:

docker compose up --build -d

3. Set up Nginx to reverse proxy port 80 → 8000 if needed.