🚗 Fuel Efficiency Prediction System ⛽

An interpretable, robust regression system for predicting vehicle fuel efficiency (MPG) using the Auto MPG dataset. This project goes beyond typical ML demos by emphasizing understanding over accuracy and treating regression as a decision-support system.

---

📋 Table of Contents

• What Makes This Different
• Project Structure
• Quick Start
• Dataset
• Model Architecture
• Key Insights
• Generated Outputs
• Detailed Guide
• Future Extensions
• Contributing
• License

---

🎯 What Makes This Different

Typical Auto MPG Projects	This Project
Train one neural network	Compare multiple models for honest evaluation
Print a loss value	Explain why predictions behave as they do
Declare success	Demonstrate when NNs help—and when they don't
Black-box predictions	Prioritize interpretability over marginal gains

💡 Philosophy: Numbers are only impressive when they explain something real.

---

📁 Project Structure

Fuel-efficiency-prediction/
├── 📂 data/
│   └── loader.py              # Data ingestion from UCI repository
├── 📂 preprocessing/
│   ├── cleaner.py             # Missing value handling, encoding
│   └── features.py            # Normalization, train-test split
├── 📂 eda/
│   └── visualizations.py      # Pairplots, correlation analysis
├── 📂 models/
│   ├── neural_network.py      # Keras Sequential model (64→32→1)
│   └── baselines.py           # Linear Regression, Ridge, Random Forest
├── 📂 training/
│   └── trainer.py             # Training loop with early stopping
├── 📂 evaluation/
│   ├── metrics.py             # MAE, RMSE, R² calculations
│   ├── plots.py               # Diagnostic visualizations
│   └── interpretability.py    # Feature importance analysis
├── 📂 utils/
│   └── config.py              # Hyperparameters and constants
├── 📂 outputs/                # Generated plots and artifacts
├── 🐍 main.py                 # End-to-end pipeline
├── 📄 requirements.txt
├── 📄 GUIDE.md                # Detailed usage guide
├── 📄 LICENSE
└── 📄 README.md

---

🚀 Quick Start

Prerequisites

• Python 3.8 or higher
• pip package manager

1. Clone the Repository

git clone https://github.com/pronzzz/fuel-effeciency-prediction.git
cd fuel-effeciency-prediction

2. Install Dependencies

pip install -r requirements.txt

3. Run the Full Pipeline

python main.py

Command Line Options

Option	Description
--eda-only	Only run exploratory data analysis
--skip-eda	Skip EDA for faster execution
--verbose	Enable detailed output

---

📊 Dataset

The Auto MPG dataset from the UCI Machine Learning Repository contains fuel efficiency data for automobiles from 1970-1982.

Feature	Description	Type
mpg	Miles per gallon	Target
cylinders	Number of cylinders	Discrete (3-8)
displacement	Engine displacement (cubic inches)	Continuous
horsepower	Engine horsepower	Continuous
weight	Vehicle weight (lbs)	Continuous
acceleration	0-60 mph time (seconds)	Continuous
model_year	Year (70-82)	Discrete
origin	Country (USA/Europe/Japan)	Categorical

Dataset Statistics:

• 📦 398 samples
• 🔢 8 features (7 predictors + 1 target)
• ❓ 6 missing values (horsepower)

---

🧠 Model Architecture

Neural Network

┌─────────────────────────────────────────────┐
│  INPUT (9 features after one-hot encoding) │
└─────────────────────┬───────────────────────┘
                      ▼
┌─────────────────────────────────────────────┐
│  Dense(64, ReLU)                            │
│  Purpose: Learn feature combinations        │
└─────────────────────┬───────────────────────┘
                      ▼
┌─────────────────────────────────────────────┐
│  Dense(32, ReLU)                            │
│  Purpose: Compress to efficiency-relevant   │
└─────────────────────┬───────────────────────┘
                      ▼
┌─────────────────────────────────────────────┐
│  Dense(1, Linear)                           │
│  Purpose: Predict MPG                       │
└─────────────────────────────────────────────┘

Total Parameters: ~2,753

Baselines for Comparison

Model	Purpose
Linear Regression	Simple interpretable baseline
Ridge Regression	L2 regularization for multicollinearity
Random Forest	Nonlinear tree-based ensemble

---

🔍 Key Insights

1️⃣ Weight Dominates MPG Prediction

Vehicle weight is the single strongest predictor of fuel efficiency:

• Physics: Heavier cars need more energy to accelerate (F = ma)
• Correlation: r ≈ -0.83 with MPG
• Engineering insight: 500 lb reduction → 3-5 MPG improvement

2️⃣ When Neural Networks Help (and When They Don't)

For tabular data with ~400 samples:

Observation	Implication
NN provides modest improvement over linear regression	Complexity may not be worth it
Random Forest often matches NN performance	Simpler is often better
Baseline models are highly competitive	Start simple, add complexity only when needed

3️⃣ Interpretability > Marginal Accuracy

In real-world engineering decisions:

Knowing "reduce weight by 500 lbs → +3 MPG" is more valuable than knowing "predicted MPG is 32.7"

---

📈 Generated Outputs

After running the pipeline, check outputs/ for:

File	Description
pairplot.png	Feature relationships and correlations
correlation_heatmap.png	Multicollinearity detection
target_distribution.png	MPG distribution analysis
features_by_origin.png	Feature differences by car origin
predicted_vs_actual_*.png	Prediction accuracy per model
residuals_*.png	Error pattern analysis
training_history.png	Neural network convergence
feature_importance_*.png	Which features matter most

---

📖 Detailed Guide

For a comprehensive walkthrough including:

• Step-by-step pipeline explanation
• Understanding each visualization
• Interpreting model results
• Customizing hyperparameters
• Extending the system

See GUIDE.md

---

🛠️ Future Extensions

• SHAP values for game-theoretic feature importance
• Uncertainty estimation for predictions
• Scenario simulation (e.g., "What if we reduce weight by 500 lbs?")
• Gradient Boosting (XGBoost, LightGBM) baselines
• Cross-validation for more robust evaluation
• Web interface for interactive predictions

---

🤝 Contributing

Contributions are welcome! Here's how to get started:

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

---

📄 License

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

---

🙏 Acknowledgments

• UCI Machine Learning Repository for the Auto MPG dataset
• Original dataset from StatLib library, Carnegie Mellon University

---

Made with ❤️ for interpretable machine learning