Deep Learning Based Prediction For The Price Of Second-hand Cars

The second-hand car trading market provides consumers with a new way to purchase vehicles,whether it is a physical second-hand car trading market or an online second-hand car trading platform,the second-hand car market is developing day by day.Potential markets are often accompanied by risks,and in the process of second-hand car trading,due to consumers’ own lack of experience and lack of transparency in the second-hand car trading market,the value assessment results of second-hand cars will be inaccurate.In view of the above problems,in this paper,we use machine learning and deep learning technology to evaluate and predict the price of secondhand cars,so as to provide reasonable and fair second-hand car value to protect consumers’ rights and interests as much as possible.The data used in this article is Ali Tianchi second-hand car transaction data,the data has been desensitized,with a total of 150,000 pieces of data.The machine learning and deep learning models are constructed by using this data set,and MAE is used to evaluate the predictive ability of the models.Firstly,for machine learning models,the data is preprocessed with missing values and outliers,and then the features are constructed and filtered,and the time variables that can be used by the machine learning models are constructed,and some features are filtered out according to the correlation coefficient.The processed data and features are input into the models for parameter tuning,and the five-fold cross is used for verification and construction.Secondly,for deep learning,data preprocessing and feature engineering are not the same as above,feature construction adopts addition and multiplication and a variety of codings,and after PCA dimension reduction,the features are normalized before entering the model.Finally,the MAE values of the above models are compared,and the results show the model based on the deep learning framework Keras is obviously better,and the MAE value is the smallest,which is far lower than the MAE values of each model of machine learning model.According to the deep learning model established in this thesis,it can better guide the pricing of the second-hand car trading market,drive the second-hand car evaluation system to be further improved,and provide more reliable technical support for consumers and merchant platforms.