RNN-based Stock Price Prediction for American Airlines (AAL)

Introduction

This project implements a Recurrent Neural Network (RNN) to forecast the next-day closing price of American Airlines (AAL) stock. Leveraging time series data, the model captures temporal dependencies to provide accurate price predictions.

Dataset

• Source: Yahoo Finance

• Ticker: AAL

• Period: December 1, 2022 to November 30, 2023

• Features: Open, High, Low, Close, Adj Close, Volume

Features

• Utilizes raw OHLC data along with adjusted closing price and trading volume

• Constructs sequences of 5-day historical windows to predict the following day's closing price (input shape: (5, 6))

Preprocessing

1. Normalization:

   • Scales all features to the [0,1] range using MinMaxScaler

2. Sequence Generation:

   • Custom rnn_data_setup function creates training samples with lookback windows and corresponding targets

3. Data Splitting:

   • Training set: 144 samples

   • Validation set: 48 samples

   • Test set: 48 samples

Model Architecture

• Type: Sequential RNN

• Layers:

  • SimpleRNN with 16 units (linear activation)

  • Dense output layer (1 unit)

• Hyperparameters:

  • Learning rate: 0.0012

  • Batch size: 512

  • Loss: Mean Squared Error (MSE)

  • Metric: Mean Absolute Error (MAE)

  • Optimizer: Adam

Training

• Epochs: 100

• Validation: Monitored on a hold-out validation set during training (verbose=0)

Results

• Prediction Shape: (48, 1)

• Evaluation Metrics on Test Set:

  • Mean Squared Error (MSE): 0.08615

  • Root Mean Squared Error (RMSE): 0.29352

Usage

1. Clone the repository:

   git clone https://github.com/CSwebD/RNN-AAL-Prediction.git

   cd RNN-AAL-Prediction

2. Install dependencies:

   pip install -r requirements.txt

3. Launch Jupyter Notebook:

   RNN.pred.AAL_jupyter.ipynb

4. Execute all cells to reproduce preprocessing, training, and evaluation.

Dependencies

Python 3.x

numpy, pandas, matplotlib, seaborn, yfinance, tensorflow (keras), scikit-learn

Contributing

Contributions are welcome! Please fork the repo, create a branch for your feature or bugfix, and submit a pull request.

License

This project is licensed under the MIT License. Feel free to use and modify for educational and research purposes.