Installation and Setup

Prerequisites

• Python 3.8+
• Node.js 14+
• Docker
• Google Cloud account

Backend (Flask)

1. Navigate to the Flask directory:

   cd src/Flask

2. Create a virtual environment and activate it:

   python -m venv venv
   source venv/bin/activate   # On Windows use `venv\Scripts\activate`

3. Install the required packages:

   pip install -r requirements.txt

4. Run the Flask application:

   flask run

Frontend (React)

1. Navigate to the React directory:

   cd src/React

2. Install the dependencies:

   npm install

3. Start the React application:

   npm start

Forecasting Models

Each forecasting model has its own subdirectory under src/Forecasting. Follow the instructions in the respective README.md files to set up and run each model.

TensorFlow LSTM

1. Navigate to the Tensorflow_LSTM directory:

   cd src/Forecasting/Tensorflow_LSTM

2. Install the required packages:

   pip install -r requirements.txt

3. Run the model script:

   python model.py

Prophet

1. Navigate to the Prophet directory:

   cd src/Forecasting/Prophet

2. Install the required packages:

   pip install -r requirements.txt

3. Run the model script:

   python model.py

StatsModel

1. Navigate to the StatsModel directory:

   cd src/Forecasting/StatsModel

2. Install the required packages:

   pip install -r requirements.txt

3. Run the model script:

   python model.py

Deployment

Docker and Google Cloud

1. Build the Docker images for the Flask and React applications:

   docker build -t flask-app ./src/Flask
   docker build -t react-app ./src/React

2. Push the Docker images to Google Container Registry:

   docker tag flask-app gcr.io/your-project-id/flask-app
   docker tag react-app gcr.io/your-project-id/react-app
   docker push gcr.io/your-project-id/flask-app
   docker push gcr.io/your-project-id/react-app

3. Deploy the applications on Google Cloud using Kubernetes or Cloud Run. Follow the respective Google Cloud documentation for detailed steps.

Design Diagram

Final Report

The final report provides a comparative analysis of the forecasting models (TF/Keras/LSTM, Prophet, StatsModel). The recommendation for the best time-series forecasting model is based on the experimental results obtained.

Contact

For any questions or feedback, please contact:

• Vishal Gawade: [email protected]
• Shlok Chaudhari: [email protected]
• Mihira Gudimetla: [email protected]