Installation of Intel OneAPI Toolkit

To optimize our project using Intel hardware, we installed the Intel oneAPI AI Tools as follows:

Downloaded Intel AI Tools: We downloaded and installed Intel AI Tools for full hardware optimization.

wget https://registrationcenter-download.intel.com/akdlm/IRC_NAS/f27e9e0e-ec27-4024-a4bf-b30c48c99564/l_AITools.2024.2.0.156.sh
sh l_AITools.2024.2.0.156.sh

Set Up Conda Environments: We created isolated environments for different phases of the project:

conda create -n modin intelpython3_core
conda activate modin

conda create -n pytorch-gpu intelpython3_core
conda activate pytorch-gpu

Installed Dependencies: Using a requirements.txt file, we ensured that all necessary dependencies were installed:

pip install -r requirements.txt

Running the Project

Follow the steps below in the specified order to run the project:

1. Preprocess the Data This script cleans and preprocesses the dataset. Ensure your data files are correctly placed and adjust file paths as needed. Command:

   python preprocess_data.py

2. Feature Extraction This script extracts relevant features from the cleaned data for model training. Command:

   python feature_extraction.py

3. Train the Model This script trains the AI model using the preprocessed features and saves the trained model for future predictions. Command:

   python orion_butterfly.py

4. Make Predictions This script loads the trained model and contains the function to make predictions based on new input data. Command: This file is indirectly run via the Streamlit application in the next step.

5. Run the Streamlit Application This is the main application where users can input data for predictions and visualize results through interactive graphs. Command:

   streamlit run app.py

6. Exploratory Data Analysis This script generates various visualizations using Plotly based on the dataset. The visualizations will be displayed in the Streamlit application when running app.py.

Prototype Details

Our AI-driven system leverages large-scale transportation datasets integrated with advanced deep learning models to forecast and optimize transit schedules and routes. By employing real-time demand forecasting, the system dynamically adapts to fluctuating user needs, effectively minimizing fuel consumption and carbon emissions while enhancing user convenience.

For seamless integration, we will utilize API pulls alongside Streamlit to create an intuitive and interactive user interface, enabling real-time insights and efficient data interaction.

---

Utility Summary

Through this project, we achieved:

• Optimized transportation routes and schedules
• Reduced fuel consumption and minimized carbon emissions
• Real-time predictions using an AI model deployed via Flask
• Efficient use of Intel-optimized tools to accelerate model training and deployment

The project demonstrates how AI can be applied to public transportation to enhance environmental sustainability and operational efficiency.