Analysis On Preservation Rate Of Used Cars

Due to the rapid development of China's automobile industry,as well as the introduction of a series of favorable policies such as the number limit,cancellation of used car moving,China's used car market is also expanding.However,there are still many problems in China's used car market compared with the new car market,like lack of scientific used car evaluation standards,information asymmetry and so on.The evaluation of used cars requires higher professional knowledge of consumers,who are often unable to accurately judge the value of a used car.At present,it is mainly priced by professional appraisers according to their experience.It hasn't been a long time since a series of data mining techniques were used to establish the evaluation or prediction model of used car preservation rate.Compared with Japan,Europe,the United States and other developed countries,China's used car industry chain is still immature,and there is room for further optimization.Firstly,the general situation of both domestic and foreign used car market and the estimation method of the preservation rate are studied in this paper.At present,China's used car market is still on its initial stage,and there is a large space for market progress,but it still lacks more effective supervision and evaluation mechanism.The used car market in many foreign developed countries has been more mature,with the trading volume of used cars higher than that of new cars,a sound market system and a complete used car certification system.Hence,there are a lot of experience is worth learning for china.On this basis,taking Tianjin's used car market as an example,the used car sales data of an e-commerce company in this area is crawled using web crawler,descriptive statistics and single factor impact analysis on the preprocessed data is carried out,the overall situation of the preservation rate in the sample is grasped,and how each factor specifically affects the preservation rate of used cars is explored in this paper.And then using the method of ordered logistics model,the factors are converted into ordered factors including three types of value(high,middle,low),the 14 key characteristics of the used vehicle high resale value are determined,including the registration time,table show mileage,vehicle wheelbase,automobile manufacturers,transmission type,structure type,emissions standards,emissions,maximum horsepower,whether there is equipped with GPS navigation or leather seats,whether there is easy to loss components,and whether the car has replaced the paint repair and appearance.The results of the model are explained in terms of usage,basic attribute,dynamic condition,internal and external configuration and fault troubleshooting.Then,based on these key characteristics,a multivariate linear regression model that can clearly reflect and explain the used-car preservation rate is established.The R squared after model adjustment is 0.813,and the goodness of fit is good.All the tests have passed,indicating that the model is reasonable and effective.According to the obtained regression coefficient,the model is analyzed,and it is concluded that the length of licensing time,wheelbase size,different automobile manufacturers and different body structure have the most significant influence on the used car preservation rate.Finally,in order to predict the used car insurance rate more accurately,a gradient boosting regression tree prediction model is established for the same data.The model is evaluated by the method of 10-fold cross validation,and the mean square error of the gradient boosting regression tree model is 31.58%,lower than that of the multivariable linear regression model.This indicate that compared with the multivariable linear regression model,the prediction accuracy of the gradient boosting regression model is significantly improved.By calculation,the average accuracy of the gradient boosting regression tree model is 97.83%,which can accurately predict the used car preservation rate.