Evaluation Model Research Of Old Motor Vehicle Based On Hedonic And CatBoost

With the rapid development of China’s social economy,the volume of old motor vehicles has grown rapidly.However,due to the rapid development of the old motor vehicle market,a series of problems have arisen in the transaction process.For example,the information asymmetry in the transaction process can not guarantee the consumer rights,and a large number of offline evaluation work increases the human operation cost.And other issues.Therefore,in this fast-developing old motor vehicle market,how to make consumers understand the value of goods simply and b hdirectly,how to quickly evaluate the value of old motor vehicles,and how to reduce the labor cost of operation is urgently needed to be solved.problem.The main research contents of this paper are as follows:1)This paper uses the characteristic price theory that is more commonly used in the process of commodity value evaluation,and applies it to the field of old motor vehicle value assessment.By using this method,the value of the old motor vehicle is split into the respective functions and utility values of the 21 features.According to the market feature vector,the vehicle’s own vehicle condition feature vector,and the vehicle loss feature vector,the vehicle value is divided into three major aspects,and quantified according to certain rules.The characteristic price theory has higher objectivity and stronger credibility,which makes up for the neglect of the traditional way to consumer psychological perception.2)After using the feature price theory to filter the variables,the data is preprocessed to meet the application of the model.In the data preprocessing stage,the first step is to fill in the missing values in the data by means of manual interpolation and variable culling.The second step uses descriptive analysis and difference comparison to find out the outliers in the data and process them.The third step encodes the variables according to the business perspective and market perspective,mainly using mapping sequence encoding and one-hot encoding.Finally,the data is standardized to remove the influence of the data dimension on the model’s effect.3)After pre-processing,this paper uses XGBoost,Extratree,random forest and CatBoost to train the training set and obtain the evaluation model.Then compare the generalization performance of each model,use RMSE and coverage as evaluation indicators,select the best effect model,and establish the final old motor vehicle value evaluation model.During the training process,CatBoost has a generalization capability of more than 80% in the coverage(±5%)range,which is superior to other models’ performance ability and achieves the industry application level.4)Using the CatBoost algorithm to build an old vehicle evaluation model,the generalization performance and performance of the model is better than several other models.Using CatBoost to rank the importance of features,judging the importance of features to value,using this method to find that the use time has the highest impact,followed by the owner’s expected price,bare car price,production and brand market share accounted for the second place.In the fifth place,we can understand that the above five characteristics play a decisive role in the evaluation process of the old motor vehicle.Based on the research and analysis of this paper,an old motor vehicle value evaluation model based on the combination of characteristic price theory and CatBoost is established.By establishing a value assessment model,the input of labor cost is reduced,the operating cost is reduced,the limitations of the vehicle model in the manual evaluation process are reduced,and the accuracy of the evaluation is improved.