Study On The Valuation Method Of Second-Hand Cars Based On Random Forest

Since 2000,with the rapid development of our economy,people’s concept of second-hand car consumption has changed greatly,and the state has provided a series of policy supports to second-hand cars.Meanwhile,the total number of automobiles in China exceeds 140 million,which promotes the development of high sales of second-hand cars in China.Second-hand cars are in disorder,the value evaluation system of second-hand cars is not unified,the evaluation process lacks objectivity,and it is difficult to gain the trust of consumers.The evaluation standards of the online trading market are vague and there is no reference basis and standard.Therefore,it is urgent to establish a scientific used car value evaluation system.Random forest model is a model formed by the combination of multiple decision trees,which is a kind of machine learning model.It has strong advantages in processing data noise reduction,precision prediction,missing value processing and not easy to overfit.Therefore,this paper chooses random forest as the used car evaluation model to study.First,the characteristic variables are divided into 7 entity variables,6 functional variables and 3 market variables,and the values are assigned.Secondly,the assigned data is imported into the model.According to the decreasing property of mean square error,the importance of the characteristic variables is sorted and selected.Finally,9characteristic variables are selected to build the used car value evaluation model based on random forest.Then,the sample number was divided into training data set and test data set according to the ratio of 7:3.The optimal value of random variable number(mtry)extracted during each growth of a single decision tree and the optimal value of decision tree number(ntree)extracted during each growth of a single decision tree were determined.Then,based on the two optimal values,the model was modeled based on the test set,and the goodness-of-fit R2,mean relative error MRE,mean absolute error MAE and root mean square error RMSE of the model were selected to verify the validity of the model operation.The accuracy of the model evaluation was verified by the error analysis of the estimated and actual values of the model of second-hand cars.The results of calculation and analysis show that the model runs well and the goodness of fit reaches92.21%.The error between evaluation value and actual value is mainly within 5%.Finally,it is concluded that random forest model is very suitable for a large number of second-hand car value evaluation.Finally,Random Forest algorithm and current market price method are respectively used to compare the case of a used car of Mazda series that has been traded on Guazi second-hand car.It is found that the average error rate of random forest algorithm is less than 5%,which proves the applicability of this model in the evaluation of second-hand car value.Therefore,the application of the random forest model established in this paper to the second-hand car market is the choice of The Times,providing a new direction for the value assessment of a large number of second-hand cars.This model effectively saves human and material resources,improves accuracy and efficiency,and is suitable for the study of vehicles with different configurations and brands.During the evaluation,only the quantified automobile data can be input into the random forest model,and the evaluation results can be obtained immediately.The process is simple and has certain applicability.