The rapid development of bank card industry brought huge amounts of transactions,and fraud transactions also increased.This paper applied GBDT to the field of bank card fraud detection.It grouped data and filtered variables by the methods of WOE and Ⅳ,and then combined models by the strategy of Bagging and finally judged whether a transaction was a fraud transaction by a weighted voting algorithm.The empirical study shows that the model could achieve the expected evaluation index,and the effect was significant.%随着银行卡行业的迅猛发展,在带来海量银行卡交易的同时,欺诈交易也与之俱增.将GBDT算法应用于银行卡欺诈侦测领域,通过WOE和Ⅳ方法对变量进行分组处理及筛选,以Bagging的方式对模型进行了组合,通过加权投票方式判断欺诈交易.实证研究表明,该模型的达到预期评价指标,效果显著.
展开▼