Auto MPG Prediction Project

Description

A multiple regression analysis model that predicts vehicle fuel efficiency (MPG) based on automobile technical specifications. This project implements a linear regression model to predict miles per gallon using features such as weight, model year, acceleration, displacement, cylinders, and horsepower.

The model achieves an R-squared score of 0.8244, meaning it explains 82.44% of the variance in MPG values with an average prediction error of approximately 3.07 MPG.

Installation

Install the required dependencies using pip:

pip install pandas numpy matplotlib seaborn scikit-learn

Run the prediction script:

python mpg_prediction.py

Tech Stack

• Python 3.x
• pandas for data processing and manipulation
• numpy for numerical computations
• matplotlib for data visualization
• seaborn for statistical data visualization
• scikit-learn for machine learning algorithms

Features

Data Processing

• Handles missing values in horsepower column using median imputation
• Excludes non-numeric car name column from analysis
• Applies StandardScaler for feature normalization

Model Training

• Implements multiple linear regression using scikit-learn
• Uses 80-20 train-test split for model validation
• Analyzes standardized coefficients to determine feature importance

Model Performance

• R-squared Test Set: 0.8244 (82.44%)
• RMSE Test Set: 3.0725 mpg

Key Findings

Most influential features based on standardized coefficients:

1. Weight: -5.9117 Strongest negative impact, heavier vehicles have lower MPG
2. Model Year: 2.8829 Second strongest positive impact, newer vehicles have better MPG
3. Acceleration: 0.2224 Weak positive impact
4. Displacement: 0.1649 Weak positive impact
5. Cylinders: 0.1162 Very weak positive impact
6. Horsepower: 0.1073 Very weak positive impact

Dataset Information

• Total Samples: 398
• Training Data: 318 samples, 80%
• Test Data: 80 samples, 20%
• Features Used: 6
• Target Variable: mpg

Generated Outputs

1. mpg_prediction.py - Main analysis script
2. mpg_prediction_results.png - Model evaluation visualizations including actual vs predicted MPG, residual plots, feature importance, and residual distribution
3. correlation_heatmap.png - Correlation matrix heatmap

Reference

Predictive Modeling of Auto MPG and Analysis of Engine Design Factors: Comparing the Impact of Weight and Displacement

License

MIT License

Author

Seonwoo Kang