A Machine Learning Based Approach for Automated Used-Car Price Evaluation
by Jong Pyeong Lee, Hiep Nguyen, Konrad Lykowski, and Thien Phan
Abstract – Artificial Intelligence and Machine Learning have been successfully applied in many fields recently for automated systems. Recent business trends raise a demand of automating the process of predicting the prices of used cars as it helps to reduce a lot of human effort. In this paper, we investigate the application of supervised machine learning techniques to evaluate the price of used cars. Our analysis is based on real datasets collected for over 8 years in a cars dealing business of a used cars seller in Japan. Possible data processing techniques and regression analysis methods, e.g., multiple linear regression and decision tree regression, have been tested to achieve a reliable prediction accuracy. Our evaluation results show an accuracy of over 90% for random forest, which is a promising achievement, especially when the datasets contain rather large amount of noises such as missing values and outliers. This paper summarizes our work including data processing methods, machine learning models building, models accuracy evaluation and achieved results at the current stage.