Fuel efficiency prediction

Demonstration of the development and analysis of different regression models to predict fuel consumption using various car attributes from the Auto MPG dataset. An answer to a question - what decides how far a car will go?

A full, detailed report of the process in Polish can be read in report.pdf

Features

The project includes data preprocessing, exploratory data analysis, and the implementation of a linear regression model using the following features:

Table 1. Feature coefficient values for LSM-based linear regression

Functionalities

Data Preprocessing

• Reads car data from a file.
• Converts MPG (Miles per Gallon) to liters per 100 km.
• Normalises selected numerical features.
• Feature engineers categorical origin variable into Europe, USA or Asia via one-hot encoding.

Visualisation

• Provides scatter plots and histograms to explore data relationships and distribution.
• Offers correlation heatmaps to understand feature interactions.

Model Training and Evaluation

• Splits the data into training and testing sets.
• Trains a linear regression model and evaluates its performance using mean squared error (MSE).
• Compares the accuracy of different regression methods: LSM, Lasso, Ridge, ElasticNet.
• Outputs the parameter values for optimal LSM-based linear regression method.

Table 2. Mean Squared Error, intercept and coefficients comparison for different regression types

Installation

To run this project, you need Python installed along with the following libraries:

• NumPy
• pandas
• matplotlib
• seaborn
• scikit-learn

You can install these packages using pip:

pip install numpy pandas matplotlib seaborn scikit-learn

Usage

To use this project:

1. Ensure you have the auto-mpg.data file in your working directory. This file contains the dataset used for modeling.
2. Run the script via the command line:

python main.py