基于时间序列支持向量机的信用额度预测

Prediction of automobile dealers’credit limit based on multidimensional time series SVR

汽车整车厂商通过经销商渠道销售整车时,会根据经销商自身的信用状况来决定信用额度,在该额度范围内,经销商可以先收车再按照规定的时间回款。以往的相关研究中,整车厂商常根据经销商某一时刻的信用状况评价其所属的信用等级,从而决定经销商的信用额度,但忽视了信用数据可能会产生“突变”,导致评价结果失真,产生信用风险。文章从受评经销商的历史业务数据出发,使用多维时间序列数据进行相空间重构,将得到的相点数据用于训练支持向量回归(support vector regression,SVR)模型,然后使用该模型给定经销商的信用额度。运用经销商信用数据的验证结果表明,该方法具有较好的预测效果。

When the automobile manufacturers sell the car through the dealers,they will decide the credit limit according to the dealers’credit condition.Within the limit,the dealers can get the car first,and then pay back within the stipulated time.In the previous research,automobile manufacturers evaluate the credit rating of the dealers based on their credit condition at a certain time,and decide the credit limit of the dealers,but ignore that the credit data may produce“mutation”,which will lead to the distortion of the evaluation results and cause the credit risk.In view of the historical business data of the evaluated dealer,this paper uses the multidimensional time series data for the phase space reconstruction,trains the support vector regression(SVR)model by the obtained phase points,and uses this model to give the credit limit of the dealer.The method is validated by the dealer’s credit data,which shows that the method has better prediction effect.