本文从优化初始簇入手,提出了改进的聚类算法,提高了信贷风险识别效率及准确率.主要工作包括:(1)实现基于信贷特色的申贷数据集标准化算法;(2)提出δ-相似度度量概念;(3)提出基于δ-K means的信贷风险识别算法δ-KCLR(δ-K-means-risk analysis of the bank credit)算法;(4)实验表明在银行信贷业务分析中,采用δ-KCLR算法可以有效识别隐含在信贷业务中的信贷风险.用这一模型可指导或预测新增贷款人中是否存在贷款风险.
The paper focus on the optimization initial clusters, proposed a novel modified clustering method,improved the identification efficiency and accuracy. The main contributions include. (1) implementing loan data standardization algorithm based on commercial loan features. (2) proposing a novel concept δ-similarity. (3)proposing loan risk identification algorithm based on δ-K means. (4) By experiments showing that δ-KCLR (δ-K-means-risk analysis of the bank credit) algorithm can effectively identify the risks in commercial loans.