📈 Quantitative Trading: Nvidia Stock Prediction 🚀

The project includes a predictive model that generates trading signals ( buy / sell / hold ) for Nvidia’s stock based on historical data using an LSTM-based neural network.

You can run the model directly, and the predicted signal for the next trading day will be displayed in the command line:

python -m models.model_loder.lstm_loader

Screenshot on 2025-07-08:

Results (2024-12-05 to 2025-07-08):

• Initial Capital: $10,000.00
• Final Portfolio Value: $15,385.39
• Strategy Return: +53.85%
• Buy & Hold Return: +10.30%
• Max Drawdown: 17.80%
• Sharpe Ratio: 1.64
• Win Rate: 55.00%
• Total Trades: 120

Hope this helps you get rich 💰 — or at least lose money more elegantly. 💸

If this project makes you rich, feel free to buy me a coffee ☕ or a Tesla 🏎️.

---

🧠 LSTM Stock Prediction Model

The LSTM model is designed to predict the closing prices of Nvidia stock based on historical price data. It uses a sequence-based model where past prices are used to predict future prices.

Key Components:

• LSTM Model:
  • Input: Historical closing prices (last n_steps days).
  • Output: Predicted closing price for the next day.
  • The model is built using PyTorch and consists of an LSTM layer followed by a fully connected layer.
• Data Preprocessing:
  • The data is normalized using MinMaxScaler.
  • The dataset is split into training and testing sets, with a rolling window approach for time-series prediction.

Usage:

1. Train the Model:
   Use the StockPredictor class to preprocess the data and train the model.

python main.py

Use Optuna to get the best hyperparameters:

python -m util.optuna_tuner.py

2. Predict:

Install the required dependencies. Then, run the following command to generate the predictive trade signal for the next day:

python -m models.model_loder.lstm_loader

---

📌 Current Features

• Data Collection

  • Historical Stock Price Analysis:
    Utilizes the Yahoo Finance API to fetch Nvidia (NVDA) stock data, while also computing key technical indicators:
    • Simple Moving Averages (SMA)
    • Exponential Moving Averages (EMA)
    • Relative Strength Index (RSI)
    • MACD (Moving Average Convergence Divergence)
    • Bollinger Bands
  • Macro-Economic Data:
    Leverages the Federal Reserve Economic Data (FRED) API via the fredapi package to obtain:
    • Interest Rates
    • Inflation Rates
    • Unemployment Rates
    • GDP Data
  • News & Sentiment Analysis:
    Scrapes Nvidia-related news from the Google News RSS feed and computes sentiment scores using VADER.

• Stock Prediction Model (LSTM)

  • LSTM-based Stock Price Prediction:
    Implements a Long Short-Term Memory (LSTM) model to predict Nvidia's stock closing prices based on historical data.
  • Features:
    • Trains on historical stock data, incorporating sequence-based time series analysis.
    • Predicts closing prices for the next day.

---

🛠 Installation

Ensure you have Python 3.8+ installed. Then clone the repository and install the required dependencies:

git clone https://github.com/kongchenglc/QuantitativeTrading.git
cd QuantitativeTrading
pip install -r requirements.txt

⚠️ A FRED API key is required

• Head over to https://fred.stlouisfed.org/docs/api/api_key.html to generate an API key
• Once you generate a key, create an .env file in the root directory of the repo
• Add your API key here as FRED_API_KEY=<your key>

Conda

1. If you are using Conda, you can create a new environment by running conda env create -f environment.yml.
2. This command will create a new Conda environment for you with the dependencies declared in environment.yml.
3. Once Step 2 finishes, activate the environment: conda activate nvidia-quantitative-stock
4. You can now run the scripts 🎉