ExplainableAI - ML Model Analysis Dashboard

A comprehensive web-based platform for analyzing, visualizing, and understanding machine learning models through advanced explainability techniques powered by SHAP, AI explanations, and interactive visualizations.

🎯 Overview

ExplainableAI provides an intuitive interface for data scientists and machine learning practitioners to:

• Upload and analyze sklearn models (.joblib, .pkl) and ONNX models
• Generate comprehensive insights using SHAP (SHapley Additive exPlanations)
• Visualize feature importance and model behavior through interactive charts
• Perform what-if analysis to understand prediction changes
• Get AI-powered explanations using AWS Bedrock (Claude 3 Sonnet)
• Analyze model performance with ROC curves, confusion matrices, and threshold analysis
• Explore feature dependencies with partial dependence plots and SHAP dependence plots

🏗️ Architecture

Backend (FastAPI)

• Framework: FastAPI with async support
• ML Support: scikit-learn models and ONNX runtime
• Explainability: SHAP library for model interpretability
• AI Explanations: AWS Bedrock integration with Claude 3 Sonnet
• Authentication: Token-based authentication system
• Data Storage: Local file storage for models and datasets

Frontend (React + TypeScript)

• Framework: React 19 with TypeScript
• Build Tool: Vite for fast development and building
• Styling: Tailwind CSS for responsive design
• Visualizations:
  • Plotly.js for interactive charts
  • Recharts for data visualization
  • React Force Graph for network visualizations
• Navigation: React Router for SPA routing

🚀 Quick Start

Prerequisites

• Python 3.8+
• Node.js 18+
• npm or yarn

Backend Setup

1. Navigate to backend directory

   cd backend

2. Create virtual environment

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

3. Install dependencies

   pip install -r requirements.txt

4. Set environment variables (Optional - for AI explanations)

   export AWS_ACCESS_KEY_ID_LLM="your_aws_access_key"
   export AWS_SECRET_ACCESS_KEY_LLM="your_aws_secret_key"
   export AWS_SESSION_TOKEN_LLM="your_aws_session_token"  # Optional
   export REGION_LLM="us-east-1"  # Optional

5. Start the server

   uvicorn main:app --reload

   The API will be available at http://localhost:8000

Frontend Setup

1. Navigate to frontend directory

   cd frontend

2. Install dependencies

   npm install

3. Start development server

   npm run dev

   The frontend will be available at http://localhost:5173

📊 Features

Model Analysis

• Model Overview: Comprehensive statistics, performance metrics, and metadata
• Classification Stats: Accuracy, precision, recall, F1-score, AUC, confusion matrix
• Feature Importance: SHAP-based and built-in importance rankings
• ROC Analysis: ROC curves with optimal threshold detection
• Threshold Analysis: Performance metrics across different decision thresholds

Explainability & Interpretability

• Instance Explanations: SHAP values for individual predictions
• Feature Dependence: Partial dependence plots and SHAP dependence plots
• What-If Analysis: Real-time prediction changes with feature modifications
• Feature Interactions: Pairwise feature interaction analysis
• Decision Tree Visualization: Explore ensemble tree structures

Data Analysis

• Dataset Comparison: Training vs test dataset statistics and drift detection
• Feature Correlations: Correlation analysis between selected features
• Data Quality: Missing values, duplicates, and health scores
• Interactive Visualizations: Scatter plots, heatmaps, and network graphs

AI-Powered Insights

• Natural Language Explanations: AI-generated interpretations of analysis results
• Context-Aware Descriptions: Explanations tailored to different analysis types
• Business Impact: Translation of technical metrics into business insights

🛠️ API Endpoints

Authentication

All endpoints require a token parameter. For development, use token=dev_token.

Core Endpoints

• POST /upload/model-and-data - Upload model and dataset
• POST /upload/model-and-separate-datasets - Upload model with separate train/test data
• GET /analysis/overview - Get model overview and performance metrics
• GET /analysis/classification-stats - Get detailed classification statistics
• GET /analysis/feature-importance - Get feature importance rankings

Explainability Endpoints

• GET /analysis/explain-instance/{instance_idx} - Explain individual prediction
• POST /analysis/what-if - Perform what-if analysis
• GET /analysis/feature-dependence/{feature_name} - Get feature dependence
• POST /analysis/explain-with-ai - Get AI-powered explanations

Advanced Analysis

• POST /api/correlation - Feature correlation analysis
• POST /api/roc-analysis - ROC curve analysis
• POST /api/threshold-analysis - Threshold optimization
• POST /api/partial-dependence - Partial dependence plots
• POST /api/interaction-network - Feature interaction network

🧪 Supported Models

Model Formats

• scikit-learn: .joblib, .pkl, .pickle files

Model Types

• Binary Classification: Logistic Regression, Random Forest, SVM, XGBoost, etc.
• Multiclass Classification: Support for multi-class problems
• Tree-based Models: Enhanced support for decision trees and ensembles

Data Formats

• CSV Files: Training and test datasets
• Features: Numeric and categorical features
• Target: Binary and multiclass labels

🔧 Configuration

Environment Variables

# AWS Bedrock Configuration (Optional)
AWS_ACCESS_KEY_ID_LLM=your_access_key
AWS_SECRET_ACCESS_KEY_LLM=your_secret_key
AWS_SESSION_TOKEN_LLM=your_session_token
REGION_LLM=us-east-1

# Storage Configuration
STORAGE_DIR=./storage  # Default: backend/storage

Model Requirements

• Models must be trained and saved using supported formats
• Feature names should be consistent between training and inference
• Binary classification models should output probabilities for both classes

📁 Project Structure

ExplainableAI/
├── backend/                    # FastAPI Backend
│   ├── main.py                # Main application entry point
│   ├── requirements.txt       # Python dependencies
│   ├── app/
│   │   ├── core/             # Core configuration and auth
│   │   ├── services/         # Business logic services
│   │   │   ├── model_service.py
│   │   │   └── ai_explanation_service.py
│   │   └── storage/          # File storage
│   └── storage/              # Uploaded models and datasets
├── frontend/                 # React Frontend
│   ├── package.json         # Node.js dependencies
│   ├── src/
│   │   ├── components/      # React components
│   │   ├── services/        # API services
│   │   └── App.tsx         # Main application component
│   └── public/             # Static assets
├── *.csv                    # Sample datasets
├── *.joblib                 # Sample models
└── test_*.py               # Test scripts

🧪 Testing

Sample Data

The repository includes sample datasets and models for testing:

• breast_cancer_dataset.csv - Binary classification dataset
• wine_classification_dataset.csv - Multiclass classification dataset
• loan_approval_dataset.csv - Credit approval dataset
• Pre-trained models in .joblib format

Test Scripts

• test_upload.py - Test file upload functionality
• test_multiclass.py - Test multiclass model support
• test_decision_tree.py - Test decision tree analysis
• validate_ui_data_display.py - Validate frontend integration

Running Tests

# Test backend functionality
python test_upload.py

# Test multiclass support
python test_multiclass.py

# Validate API endpoints
python validate_ui_data_display.py

🤝 Contributing

1. Fork the repository
2. Create a feature branch (git checkout -b feature/amazing-feature)
3. Commit your changes (git commit -m 'Add amazing feature')
4. Push to the branch (git push origin feature/amazing-feature)
5. Open a Pull Request

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.

🆘 Support

Common Issues

• SHAP Errors: Ensure your model is compatible with SHAP explainers
• Memory Issues: Use smaller datasets or sample sizes for large models
• AWS Credentials: Set up proper AWS credentials for AI explanations

Documentation

• FastAPI Docs: Available at http://localhost:8000/docs when running
• SHAP Documentation: https://shap.readthedocs.io/
• AWS Bedrock: https://docs.aws.amazon.com/bedrock/

🔄 Recent Updates

• ✅ AI-powered explanations with AWS Bedrock integration
• ✅ Enhanced multiclass classification support
• ✅ Interactive decision tree visualization
• ✅ Feature interaction analysis
• ✅ Data drift detection
• ✅ Comprehensive test coverage
• ✅ ONNX model support

🚧 Roadmap

• Model comparison dashboard
• Automated report generation
• Model monitoring and alerting
• Integration with MLflow/Weights & Biases
• Support for regression models
• Advanced feature engineering insights
• Model fairness and bias detection