Time Series Forecasting: Multi-Framework Comparison

This project implements time series forecasting using four different frameworks (TensorFlow, PyTorch, Statsmodels, and Keras) to compare their performance and accuracy on different datasets. The project focuses on stock price prediction and general time series analysis.

• Implementation of different time series forecasting models
• Comparative analysis of model accuracies
• Interactive web interface for SARIMAX model
• Visualization of predictions vs actual values
• Performance metrics calculation (MAE, RMSE, MAPE)

# Clone the repository
git clone https://github.com/yourusername/time-series-forecasting.git

# Create virtual environment
python -m venv venv

# Activate virtual environment
# For Windows
venv\Scripts\activate
# For Unix/MacOS
source venv/bin/activate

# Install required packages
pip install -r requirements.txt

├── data/
│   ├── ACGL_DATA.csv
│   ├── MSFT_stock.csv
│   └── temperature.csv
├── models/
│   ├── tensorflow_model.py
│   ├── pytorch_model.py
│   ├── statsmodels_model.py
│   └── keras_model.py
├── web_app/
│   ├── app.py
│   ├── templates/
│   └── static/
├── requirements.txt
└── README.md

The project includes a Flask web application for the SARIMAX model that allows users to:

• Input prediction timeframe
• View forecasted values
• Analyze prediction accuracy
• Visualize results through interactive plots

cd web_app
python app.py

Access the application at http://localhost:5000

• Python 3.8+
• TensorFlow 2.x
• PyTorch 1.9+
• Statsmodels 0.12+
• Keras
• Flask
• Pandas
• NumPy
• Matplotlib
• Scikit-learn

• Add real-time data fetching
• Implement ensemble methods
• Add more visualization options
• Include hyperparameter tuning
• Add cross-validation