Forecasting resale value of the car: Evaluating the proficiency under the impact of machine learning model

Abstract

Car resale value prediction is the technique for estimating the prices of an old car. Second hand cars are in high demand but the process for evaluating their value is flawed which may lead them to be priced at unrealistic rates. Here, we offer our contribution for this issue. In this paper, an ensemble machine learning approach, XG Boost is proposed, which takes the important features of a car into account and assigns the cost based on it. Our proposed system uses multiple gradient boosted trees which minimizes the overfitting and eliminates irregularities from the predictions. Further, the accuracies for different machine learning algorithms such as Linear regression, Lasso Regression, Random Forest, KNN, and CART are calculated and compared with our model. From experimental results, major factors which contributed towards the car prices are found to be year of manufacture, distance driven, fuel type and transmission. The proposed approach contributes promising results in forecasting the resale value of the car, and achieves an accuracy of 93.73%.