基于改进k-means聚类算法的上市公司信用风险评估

Credit risk assessment of listed companies based on improved k-means clustering algorithm

针对传统的k-means聚类算法初始聚类中心具有随机性,聚类结果会随着初始聚类中心的不同而波动的问题,改进了初始聚类中心选取的方法.利用UCI数据库中的Iris数据集进行实验并计算准确率,对比发现改进算法后的聚类准确率较传统算法有了明显提高.选取25家上市公司,运用因子分析法构建信用风险评估指标体系,利用改进后的评估模型对这25家上市公司的信用风险进行评估,按照信用风险等级将这些公司分为高风险、中风险、低风险三类,其中高风险1家,中风险21家,低风险3家.将改进后的k-means聚类评估模型和传统k-means聚类评估模型进行比较分析,结果表明,改进后的评估模型算法性能有所提高,评估结果更为合理.

In response to the problem that the initial clustering centers of traditional k-means clustering algorithm have randomness and the clustering results fluctuate with different initial clustering centers,the selection of initial clustering centers was improved.Experiments were conducted using the Iris dataset in the UCI database and accuracy was calculated,comparison showed that the improved algorithm significantly improved the clustering accuracy compared to traditional algorithms.Then,25 listed companies were selected,and a credit risk assessment index system was constructed using factor analysis.The improved evaluation model was used to evaluate the credit risk of these 25 listed companies,these companies were classified into three categories based on their credit risk levels:high-risk,medium risk,and low-risk,among them,there is 1 high-risk company,21 medium risk companies,and 3 low-risk companies.The improved k-means clustering evaluation model was compared and analyzed with the traditional k-means clustering evaluation model,the results showed that the algorithm performance of the improved evaluation model was improved,and the evaluation results were more reasonable.