The high rate of expansion of the pre-owned automotive market has enhanced the demand of viable and precise means of estimating the prices of vehicles. It is not easy to determine the resale value of a used car because it has to rely on various parameters including the age of the car, car mileage, the type of fuel and the transmission system and ownership. Traditional methods of pricing are largely based on subjective information by use of manual pricing and recommendations; this may lead to inconsistent and unrealistic pricing. Machine learning methods present an alternative approach which is data-driven by defining the patterns of pricing based on the past historical vehicle data. The paper provides comparative analysis of regression based machine learning models used to predict the price of used cars, such as Linear Regression, Random Forest Regressor and Gradient Boosting Regressor. These data preprocessing methods include feature encoding, outlier handling, and normalization which are embedded in the system. The standard regression evaluation metrics are the models used to measure model per- formance. Experimental experiments have shown that ensemble learning paradigms are much better than the traditional linear algorithms in their ability to describe non-linear associations between participants. The suggested system should help online platforms, dealerships, and customers make prices of their deals informed.
Siva et al. (Thu,) studied this question.