论文部分内容阅读
借鉴物理学中动力学原理,提出基于动力学理论的聚类参数挖掘策略,并应用于银行贷款数据风险评估。定义了聚类动力学参数挖掘概念、g-平均、簇的-相似、风险相似度等概念,提出基于聚类动力学参数挖掘的聚类策略挖掘算法CSMA(clustering strategy mining algorithm),分析了该策略在不同参数下对实验结果的影响。实验结果表明,CSMA策略使得聚类分析的精度提高了9%～13%。
By referring to the principle of dynamics in physics, a clustering parameter mining strategy based on dynamics theory is proposed and applied to the risk assessment of bank loan data. The concepts of clustering dynamics parameter mining, such as g-average, cluster-similarity and risk similarity are defined. Clustering strategy mining algorithm CSMA (clustering strategy mining algorithm) is proposed based on clustering dynamics parameter mining. The impact of strategy on experimental results under different parameters. Experimental results show that the CSMA strategy improves the accuracy of clustering analysis by 9% ~ 13%.