🌍 EU Carbon Tax (CBAM) Compliance & Prediction Pipeline

Advanced machine learning system for predicting EU carbon taxes on material imports with explainability, model comparison, and anomaly detection.

🚀 Live Demo

---

🎯 Problem Statement

EU companies importing materials face carbon taxes under the Carbon Border Adjustment Mechanism (CBAM). They need to:

• Predict carbon tax costs BEFORE purchasing
• Understand what factors drive costs
• Compare suppliers across countries
• Detect anomalies in shipment data

This system solves all of that.

---

✨ Key Features

Core Predictions

• Multi-Model Ensemble: XGBoost, Random Forest, Linear Regression
• 94% Accuracy: R² score on carbon emission predictions
• Instant Predictions: Sub-millisecond inference time
• Batch Processing: Predict for 100+ shipments at once

Explainability & Analysis

• SHAP Explainability: Understand why each prediction
• Model Comparison: Side-by-side accuracy comparison
• Cross-Validation: 5-fold testing for reliability
• Feature Importance: See which factors matter most
• Learning Curves: Monitor model improvement

Data Quality & Safety

• Anomaly Detection: Flag unusual shipments for review
• Validation Pipeline: Catch bad data before training
• Historical Tracking: Complete audit trail

Advanced Features

• Time Series Forecasting: Predict future emissions trends
• Active Learning: Identify highest-uncertainty shipments
• Bayesian Optimization: Auto-tune hyperparameters

---

🛠 Tech Stack

Machine Learning:

• xgboost - Primary prediction model
• scikit-learn - ML utilities, Random Forest, validation
• shap - Model explainability
• scikit-optimize - Bayesian hyperparameter tuning

Data & Database:

• pandas - Data manipulation
• numpy - Numerical computing
• sqlite3 - Lightweight database

Frontend:

• streamlit - Interactive dashboard (10 tabs)

Deployment:

• Streamlit Cloud - Free hosting
• GitHub - Version control

---

📊 Results

Metric	Value
Model Accuracy (R²)	0.94
Prediction Speed	<1ms
Training Time	30 seconds
Data Required	15+ records
Deployment Status	✅ Live

---

🚀 Quick Start

Live Demo

Open: https://cbam-system.streamlit.app/

No installation needed. Try it now!

Local Setup

1. Clone repository

git clone https://github.com/rathi29/cbam-system.git
cd cbam-system

2. Install dependencies

pip install -r requirements.txt

3. Run dashboard

streamlit run dashboard.py

4. Open browser

---

📋 How to Use

1. Load Data

• Go to "📥 Load Test Data" tab
• Click to add 15 fake shipments

2. Train AI

• Go to "🧠 Train the AI" tab
• Click "Train AI Now"
• Wait 30 seconds

3. Make Predictions

• Go to "💡 Quick Tax Prediction"
• Enter: Material, Tonnage, Country
• Get: Best country to buy from + cost comparison

4. Understand Costs

• Go to "🔍 Understand Your Cost"
• See SHAP breakdown of each factor

5. Advanced Analysis

• Model Comparison: Compare 3 models
• Cross-Validation: 5-fold reliability test
• Feature Importance: Which factors matter
• Anomaly Detection: Find risky shipments
• Learning Curves: Model improvement over time

📁 Project Structure

cbam-system/ ├── src/ │ ├── database.py # SQLite operations │ ├── utils.py # Mock data, validation │ ├── ml_forecaster.py # XGBoost training & prediction │ ├── shap_explainer.py # SHAP explanations │ ├── model_comparison.py # Compare 3 models │ ├── cross_validation.py # 5-fold validation │ ├── feature_importance.py # Feature analysis │ ├── anomaly_detection.py # IsolationForest │ ├── learning_curves.py # Learning curve generation │ ├── time_series.py # Trend forecasting │ ├── active_learner.py # Uncertainty sampling │ └── bayesian_optimizer.py # Hyperparameter tuning │ ├── dashboard.py # Main Streamlit app ├── requirements.txt # Python dependencies ├── README.md # This file │ ├── data/ │ └── carbon_ledger.db # SQLite database (auto-created) │ └── models/ └── cbam_model.pkl # Serialized XGBoost (auto-created)