针对现行方法在电商企业不平衡财务数据风险预测中存在真负类率和召回率较低的问题,提出基于模糊聚类的电商企业不平衡财务数据风险预测方法。先采用基于少数类样本的同类样本线性插值的过采样法对其进行处理,生成平衡财务数据样本集,...针对现行方法在电商企业不平衡财务数据风险预测中存在真负类率和召回率较低的问题,提出基于模糊聚类的电商企业不平衡财务数据风险预测方法。先采用基于少数类样本的同类样本线性插值的过采样法对其进行处理,生成平衡财务数据样本集,然后从电商企业投资与收益、偿债能力、盈利能力、经营能力四个方面选取预测变量,构建预测变量体系,最后利用模糊聚类算法对预测变量数据集分类,预测电商企业财务风险,实现基于模糊聚类的电商企业不平衡财务数据风险预测。经实验证明,设计方法真负类率和召回率均在95%以上,可以实现对电商企业不平衡财务数据风险的精准预测。In response to the issues of low true negative rate and recall rate in the risk prediction of imbalanced financial data of e-commerce enterprises using current methods, this paper proposes a risk prediction method for imbalanced financial data of e-commerce enterprises based on fuzzy clustering. First, an oversampling method based on linear interpolation of similar samples of the minority class is used to process the data, generating a balanced financial data sample set. Then, prediction variables are selected from four aspects of e-commerce enterprises: investment and return, debt repayment ability, profitability, and operational capability, to construct a prediction variable system. Finally, the fuzzy clustering algorithm is used to classify the prediction variable dataset, predict the financial risk of e-commerce enterprises, and achieve risk prediction of imbalanced financial data of e-commerce enterprises based on fuzzy clustering. Experimental results prove that the designed method has a true negative rate and recall rate of over 95%, enabling precise prediction of the risk of imbalanced financial data in e-commerce enterprises.展开更多
文摘针对现行方法在电商企业不平衡财务数据风险预测中存在真负类率和召回率较低的问题,提出基于模糊聚类的电商企业不平衡财务数据风险预测方法。先采用基于少数类样本的同类样本线性插值的过采样法对其进行处理,生成平衡财务数据样本集,然后从电商企业投资与收益、偿债能力、盈利能力、经营能力四个方面选取预测变量,构建预测变量体系,最后利用模糊聚类算法对预测变量数据集分类,预测电商企业财务风险,实现基于模糊聚类的电商企业不平衡财务数据风险预测。经实验证明,设计方法真负类率和召回率均在95%以上,可以实现对电商企业不平衡财务数据风险的精准预测。In response to the issues of low true negative rate and recall rate in the risk prediction of imbalanced financial data of e-commerce enterprises using current methods, this paper proposes a risk prediction method for imbalanced financial data of e-commerce enterprises based on fuzzy clustering. First, an oversampling method based on linear interpolation of similar samples of the minority class is used to process the data, generating a balanced financial data sample set. Then, prediction variables are selected from four aspects of e-commerce enterprises: investment and return, debt repayment ability, profitability, and operational capability, to construct a prediction variable system. Finally, the fuzzy clustering algorithm is used to classify the prediction variable dataset, predict the financial risk of e-commerce enterprises, and achieve risk prediction of imbalanced financial data of e-commerce enterprises based on fuzzy clustering. Experimental results prove that the designed method has a true negative rate and recall rate of over 95%, enabling precise prediction of the risk of imbalanced financial data in e-commerce enterprises.
