AwareML Dashboard

AwareML Dashboard is an interactive, research-oriented AutoML platform that integrates Explainable AI (XAI), Fairness-Aware Machine Learning, and LLM-based explanations into a unified dashboard.

Unlike traditional accuracy-driven AutoML systems, AwareML emphasizes transparency, fairness, and human interpretability, enabling users to better understand why a model is selected and how it behaves across different populations.

The project is designed as both:

• a research prototype for experimentation and evaluation, and
• an educational tool for studying responsible and trustworthy AutoML systems.

Key Features

• Multiple AutoML frameworks (AutoClass, AutoStreamML, EvoAutoML, OAML, ChaCha)
• Fairness-aware model evaluation and bias analysis
• Explainability and interpretability methods (XAI)
• LLM-based natural language explanations
• Interactive Streamlit-based dashboard
• Support for user studies and reproducible experiments

Project Structure (Simplified)

AwareML-Dashboard/
│
├── frameworks/                   # AutoClass, AutoStreamML, EvoAutoML, OAML, ChaCha (EvoAutoML and ChaCha are integrated inside the backend file)
├── meta & ml recommender/        # Meta-learning and ML-based recommender systems
├── fairness & explainability/    # Fairness metrics, bias analysis, and XAI methods
├── meta data/                    # LLM-based explanation generation
├── datasets/                     # Streaming and test datasets
└── README.md

Installation

1️⃣ Clone the repository

git clone https://github.com/vikashmaheshwari97/AwareML.git
cd AwareML-Dashboard

2️⃣ Create and activate a virtual environment

python -m venv venv
source venv/bin/activate      # Linux / macOS
venv\Scripts\activate         # Windows

3️⃣ Python version requirements

• Python 3.8 or higher
• All experiments were conducted using Python 3.8.10

4️⃣ Install required dependencies

• Each AutoML model has its own requirements.txt file located in requirements.txt files
• Install dependencies model by model by running their corresponding requirements.txt files

Important notes:

• All models support river==0.10.1

• Exception: the OAML model requires river==0.8.0

• Helper libraries such as tqdm, psutil, etc. are required

  • These are preinstalled with Anaconda, so no additional setup is usually needed

Example installation command:

pip install -r requirements.txt

▶️ Running the Dashboard

Start the Streamlit application using:

streamlit run app/forntend.py

Once running, the dashboard will open in your browser.

🔄 Usage Workflow

1. Upload a dataset (CSV format)

2. Select the target variable

3. Review automatically detected sensitive attributes

4. Run AutoML

5. Explore:

   • Model performance
   • Fairness metrics
   • Explainability insights

6. Generate LLM-based explanations

7. Export figures or results for reports and publications

License

This project is intended for academic and research use.
Licensing details can be added as required.

🙌 Acknowledgements

This project builds upon ideas from the AutoML, Fair AI, and Explainable AI research communities and is developed as part of ongoing academic research.