Predikce cen aut pomocí metod strojového učení

Abstract

This bachelor thesis focuses on predicting the market prices of used cars using machine learning methods. The aim of the thesis is to design and compare selected regression models and subsequently identify the factors that have the greatest influence on the final vehicle price. The theoretical part of the thesis covers the principles of machine learning with a focus on regression tasks, data preprocessing methods, and approaches to model evaluation and optimization. In the practical part, a dataset containing information about used vehicles was processed. This included data cleaning, exploratory data analysis, and feature transformation. Selected regression models were implemented and evaluated on the processed data, namely linear regression, Ridge, Lasso, decision tree, random forest, and XGBoost. The models were compared using evaluation metrics such as mean absolute error, root mean squared error, and coefficient of determination. The best results were achieved by the XGBoost model. The thesis also highlights limitations in predictive accuracy resulting from the scope and structure of the available data and the absence of some difficult-to-quantify factors. The results indicate the potential of machine learning methods for estimating used car prices and provide space for further extension and improvement of the model.