engleski

Price Prediction and Classification of Used-Vehicles Using Supervised Machine Learning

Due to the large growth in the number of cars being bought and sold, used-car price prediction creates a lot of interest in analysis and research. The availability of used cars in developing countries results in an increased choice of used vehicles, and people increasingly choose used vehicles over new ones, which causes shortages. There is an important need to explore the enormous amount of valuable data generated by vehicle sellers. All sellers usually have the imminent need of finding a better way to predict the future behavior of prices, which helps in determining the best time to buy or sell, in order to achieve the best profit. This paper provides an overview of data-driven models for estimating the price of used vehicles in the Croatian market using correlated attributes, in terms of production year and kilometers traveled. In order to achieve this, the technique of data mining from the online seller “Njuškalo” was used. Redundant and missing values were removed from the data set during data processing. Using the method of supervised machine learning, with the use of a linear regression algorithm for predicting the prices of used cars and comparing the accuracy with the classification algorithm, the purpose of this paper is to describe the state of the vehicle market and predict price trends based on available attributes. Prediction accuracy increases with training the model with the second data set, where price growth is predicted by linear regression with a prediction accuracy of 95%. The experimental analysis shows that the proposed model predicts increases in vehicle prices and decreases in the value of vehicles regarding kilometers traveled, regardless of the year of production. The average value of the first data set is a personal vehicle with 130, 000 km traveled and a price of EUR 10, 000. The second set of data was extracted 3 months after the previously analyzed set, and the average price of used vehicles increased by EUR 1391 per vehicle. On the other hand, average kilometers traveled decreased by 8060 km, which justifies the increase in prices and validates the training models. The price and vehicle type are features that play an important role in predicting the price in a second-hand market, which seems to be given less importance in the current literature of prediction models.

prediction ; machine learning ; used vehicles ; regression ; classification

Transportation Research Procedia Volume 64, Issue C