📊 UCIAdultXAI — Interactive Model Interpretation Dashboard

UCIAdultXAI is a production-ready, interactive dashboard for training, evaluating, and interpreting classification models (Logistic Regression, Random Forest, XGBoost, CatBoost) using SHAP and feature importance plots. It is designed for data scientists and business stakeholders to visualize model behavior and performance with ease.

---

🖼️ Project Overview

📊 Class-wise Metrics

🔥 SHAP Summary Plot

🧠 Feature Importance

• Model Accuracy & Metrics: Visualize class-wise precision, recall, and F1 score.
• Top Feature Importances: Sorted bar chart of the top 20 impactful features.
• SHAP Summary Plot: Explains global feature contributions via SHAP.

---

🚀 Features

• Supports 4 classification algorithms:

  • Logistic Regression
  • Random Forest
  • XGBoost
  • CatBoost

• Dynamic dropdown to toggle between models.

• SHAP summary visualization with matplotlib.

• Classification report parsing and rendering.

• FastAPI backend with an interactive Dash frontend.

• Dockerized with reproducible builds.

---

🧪 Tech Stack

• Backend: FastAPI, Scikit-learn, XGBoost, CatBoost
• Visualization: Dash by Plotly, SHAP
• Deployment: Docker & Docker Compose
• Logging: Custom structured logger
• Dataset: UCI Adult Income Dataset (binary classification)

---

🐳 Docker Setup

Step 1: Clone the repo

git clone https://github.com/smasoudrezvani/UCIAdultXAI.git
cd ml-insighthub

Step 2: Add dataset

Place your adult.csv file inside the data/ directory.

Step 3: Build and run the app

docker-compose up --build

Step 4: Access the dashboard

Open your browser and navigate to: 👉 http://172.0.0.1:5050/dashboard/

---

📝 Project Structure

.
├── app/
│   ├── config.py
│   ├── data.py
│   ├── dashboard.py
│   ├── logger.py
│   ├── main.py
│   ├── model.py
│   ├── pipeline.py
│   └── data/adult.csv
├── Dockerfile
├── docker-compose.yml
├── requirements.txt
└── README.md

---

🛠️ Development

To run locally:

uvicorn app.main:app --reload --host 0.0.0.0 --port 5050

Then access the dashboard at http://localhost:5050/dashboard

---

📈 Example Output

• Accuracy: 0.8569 (Logistic), 0.8783 (XGBoost)
• SHAP: Highlights high-impact features like capital_gain, age, relationship_*
• Feature Importance: Consistently emphasizes capital_gain, education, native_country

---

✅ Requirements

• Python 3.8+
• See requirements.txt for all dependencies

---

Contributing

Contributions are welcome! Please open an issue or submit a pull request for any improvements or bug fixes.

License

This project is licensed under the Apache-2.0 license.