CAR PRICE PREDICTION USING ML METHODS

Abstract

The development of car manufacturing industry is consequence of fast-growing economics in developed countries. It is hard to deny the fact of strong integration of cars in all spheres of human life nowadays. Logistics, public transport, personal usage, delivery services, taxi depots, cars for rent, etc. Demand increases due to its flexibility of usage, manufacturers strive to increase amount of customers, which creates competition on the market. As a result, the amount of time spent on investigation of market and key features of each car is dramatically huge. That is why the aim of this research is to simplify process of observation of cars, their features and prices, using machine learning techniques, which will allow to choose desirable cars in shorter time.

Article offers researches of applying machine learning methods for car price prediction, train data preprocessing, suggesting own approach based on combination of several machine learning models. Decision trees and random forest models were chosen as basic methods for this research, general concepts and construction algorithms are discussed, highlighted pros and cons each of them. The main purpose of article is the comparison of proposed methods. Distinct models are capable to predict results quite preciously – 90% accuracy for decision trees and 95% for random forest. But R² (accuracy) is not the only metrics used in model’s effectiveness evaluation. RMSE and MAE have crucial influence on optimal work of the model. Relying on the results of this research, random forest gives better RMSE and MAE values comparing to decision trees, which proves the effectiveness of such approach. However, results might improve with usage of combined approach. The question about effectiveness of such approach will be answered during this research.