Aprendizado de máquina na precificação de carros usados: desenvolvimento de uma base de dados para modelos de regressão.

Abstract

The buying and selling of used cars is a vital activity for the Brazilian economy and for access to urban mobility, especially in a scenario of rising new vehicle prices. However, determining the fair price of a used car can be challenging. This final year project explores the use of Machine Learning techniques to improve the accuracy of used car pricing. A comprehensive dataset was developed, extracted from the OLX Brazil platform through Web Scraping using Scrapy. Subsequently, exploratory data analysis was carried out, as well as cleaning and preparation for training various supervised learning models focused on regression. The results were promising, especially with the Random Forest model, which achieved a determination coefficient of 0.9434 and a mean absolute error of 4855.27. These results indicate that regression models can be effective in predicting vehicle prices based on their characteristics, and suggest the need for further investigations to further improve used car pricing techniques.