A Used Cars' Price Forecasting Model Based On Artificial Neural Network

Accompanying the development of auto industry,car parc of China is continuously increasing.With the changes in consumption concept,customers stimulate market's boosting either.However,without a proper used cars' pricing system,used cars' industry chain is dragged.It is eagerly to develop a new pricing model,which appears more robust and precise,to fit users' complicated demands such as risk assessment and subsidy policies.This thesis introduces background of used cars market in China and abroad.Through analyzing consulting existing research and pricing methods nowadays,this thesis determines ANN as foundation of model.With the transition from macro to micro,this thesis analyzes arguments which affect used cars' price.Policy,society,economy and technique are macro elements.Micro elements focus on detail of vehicle itself.Considering difficulty of acquiring data,this thesis locates twelve arguments.Original data distracted from B2 C car sales website CARKING and reinforce by CATARC data.Focusing on ANN and multivariable linear regression model,this thesis build,train and simulate both models.Compared with accuracy and stability,ANN method get a better model.Its tolerance,robustness and self-adaption provide excellent solution of none-linear problem such as pricing used cars.This model contributed an introduction of used car pricing to regulating market and protecting customers' benefits.