Research On Used Vehicle Valuation Model Based On Random Forest Algorithm

In recent years,the volume of used car transactions in China has expanded rapidly.In 2000,the volume of used cars was only 250,000,and by 2018 it had reached 13.82 million.The large middle and bottom income earners are important consumer groups of used cars.At the same time as the scale of the used car market continues to expand,the price evaluation system of the used car market in China has exposed the problem of not adapting to the market demand.The randomness in the price assessment and more human factors have affected the reasonable evaluation of the used car.The orderly development of the used car market.In contrast,in developed countries,the sustainable and stable development of the used car market is inseparable from its reasonable price assessment system.At present,China’s second-hand car market still has huge room for development.The orderly development of the used car market is also conducive to the benign development of the new car market and the entire automobile industry.Therefore,finding a more scientific and standard price assessment method has important practical significance.This paper establishes a second-hand vehicle valuation model based on random forest algorithm,conducts empirical analysis and research,and compares the valuation effect of random forest model with decision tree,K-nearest algorithm,neural network,multiple linear regression and ridge regression.In order to evaluate the price of used cars more comprehensively,this paper sorts out the domestic and foreign literatures.It is found that most scholars choose less of the characteristic variables when constructing the mathematical model valuation and cannot fully evaluate the price of used cars.Therefore,in addition to considering the influence of vehicle age and mileage on price,this paper also incorporates various factors such as vehicle power and type and brand,and uses random forest to measure the importance of these factors.The results of this paper:the evaluation results of random forest model and other models are compared by three evaluation indicators.The R2 of the random forest model,the MAE value,the variance score,and the random forest model results are best in three evaluation indicators.Secondly,the decision tree,the model R2 is 0.81.The remaining model K proximity algorithm,multiple linear regression,ridge regression and artificial neural network R2 are lower than 0.80;Among the influencing factors of car price,the top four influencing factors according to importance are car age,power,mileage and model.Among them,the importance of power is greater than the mileage,which verifies the applicability of the feature price theory in the selection of indicators.The random forest model established in this paper can provide a more applicative method to evaluate the price of used cars,such as the brand,model and power of different cars,according to the transaction records of used cars.The used car valuation method adopted in this paper can also be used for the valuation of used cars on the e-commerce trading platform.