Stock Price Prediction with Machine Learning

This project uses machine learning to predict stock prices based on historical data. It employs a Long Short-Term Memory (LSTM) neural network model to forecast future stock prices for various companies.

Features

• Data collection from Yahoo Finance using the yfinance library
• LSTM model for time series prediction
• Training on multiple stock data to improve generalization
• Prediction visualization with matplotlib
• GPU support for faster training (if available)

Requirements

• Python 3.7+
• TensorFlow 2.x
• yfinance
• pandas
• numpy
• matplotlib
• scikit-learn

You can install the required packages using:

pip install -r requirements.txt

Usage

Training the Model

_{The weights are already trained and saved in the repo 'weights' folder for the current version. This is optional.}

To train the model on historical stock data:

python scripts/train.py

This script will:

1. Download historical stock data for predefined tickers
2. Prepare the data for training
3. Build and train the LSTM model
4. Save the trained model as model.h5 and the scaler as scaler.npy

Making Predictions

To make predictions using the trained model:

python scripts/inference.py --stock <stock symbol> --output <output filename> --epoch <use model from a specific epoch>

By default, this will predict stock prices for NVIDIA (NVDA) and save the plot as predictions.png. You can modify the ticker and output filename by passing the appropriate arguments as shown above.

How it Works

1. Data Collection: Historical stock data is fetched using the yfinance library.
2. Data Preprocessing: The data is scaled using MinMaxScaler to normalize the values.
3. Model Architecture: An LSTM-based neural network is used for sequence prediction.
4. Training: The model is trained on multiple stock data to capture general market trends.
5. Prediction: The trained model predicts future stock prices based on recent data.
6. Visualization: Predictions are plotted against actual prices, including a simple trading recommendation.

Customization

• To train on different stocks, update the list of tickers in train.py.
• Adjust the time_step and future_days parameters in both scripts to change the input sequence length and prediction horizon.

License

This project is released under the GPL-3.0. For more information, see the LICENSE file.

Disclaimer

This may provide inaccurate predictions. Dont rely on it too much until it is improved.