基于深度学习的二手车价格预测模型及影响分析  

Prediction Modelling of Second-Hand Car Price Based on Deep Learning and Influence Factors Analysis

作　　者：李富强 彭海丽 杨熙 张文静 LI Fuqiang;PENG Haili;YANG Xi;ZHANG Wenjing(Equipment Industry Development Center,Ministry of Industry and Information Technology,Beijing 100846,China)

机构地区：[1]工业和信息化部装备工业发展中心,北京100846

出　　处：《汽车工程学报》2021年第5期379-385,共7页Chinese Journal of Automotive Engineering

摘　　要：运用2015~2019年中国二手车行业相关数据,基于深度学习模型,采用对比分析法、控制变量法探究我国二手车成交价格影响因素的重要性程度。结果表明,我国二手车成交量一直处于增长状态且保持较高增速;DNN模型比年限估计法及重置成本法预测的二手车成交价格精度高;新车指导价、已使用时间、已行驶距离、省份及汽车品牌重要性占比依次为67%、13.06%、9.08%、6.22%和4.64%。In order to study the importance of the factors influencing the transaction price of second-hand cars in China,firstly the present situation of second-hand cars in China was studied and analyzed.Then a model for predicting the transaction price of second-hand cars was established based on deep learning.Additionally the prediction results were compared with those of applying the year estimation method and the replacement cost method.Finally based on the deep neural network(DNN)model,the control variable method was used to analyze the importance degree of influencing factors of transaction price.The results show that China’s second-hand car transactions have been growing at a rapid rate.The DNN model is more accurate than the year estimation method and the replacement cost method in predicting the second-hand car transaction price.And the importance of the guiding price,usage time,the driving distance,the province and car brands accounts for 67%,13.06%,9.08%,6.22%and 4.64%,respectively.

关 键 词：二手车 深度学习 DNN模型 成交价格 

分 类 号：TP392[自动化与计算机技术—计算机应用技术]