Performance Evaluation of Linear Regression Models Based on Specific Features in the Turkish Second-Hand Car Market
The 3rd International Conference on Applied Mathematics in Engineering (ICAME’24), Balıkesir, Türkiye, 26 - 28 Haziran 2024, cilt.1, ss.41, (Özet Bildiri)
- Yayın Türü: Bildiri / Özet Bildiri
- Cilt numarası: 1
- Basıldığı Şehir: Balıkesir
- Basıldığı Ülke: Türkiye
- Sayfa Sayıları: ss.41
- Dokuz Eylül Üniversitesi Adresli: Evet
Özet
The automotive industry experiences steady growth and intense competition annually. In the secondhand car market, accurate pricing is critical for both sellers and customers [1]. Customers, especially those unfamiliar with automotive details, often face challenges during car purchases, making reliable price predictions valuable [2]. Linear regression methods can assist in providing these accurate price estimates by analyzing various car attributes [3,4]. This study aims to evaluate the performance of linear regression models in predicting second-hand car prices in Turkey by considering various car attributes. Using data from Turkey's leading second-hand car platform, arabam.com, we developed and compared several linear regression-based models, including Linear Regression, Ridge, Lasso, and Elastic Net. The dataset, collected over three months in 2023, included over 17,000 records with 25 different attributes, such as brand, model, age, mileage, and fuel type. Our findings indicate that while traditional Linear Regression provides a solid baseline, models like Ridge, Lasso, and Elastic Net significantly improve prediction accuracy by addressing multicollinearity and enhancing feature selection. The performance of these models was thoroughly compared to determine their effectiveness in price prediction. This study demonstrates the potential of linear regression algorithms to enhance the transparency and efficiency of the second-hand car market in Turkey. By providing accurate price estimates, these models can help customers make informed buying decisions and assist sellers in setting competitive prices, ultimately contributing to a fairer and more efficient marketplace.