论文部分内容阅读
运用聚类方法把公司财务状况分为5个等级,分别为财务状况健康,良好,一般,预警和危机,与以往将研究样本分为ST和非ST两类的财务预警模型相比,5分类模型更加精确合理,贴近实际。同时基于指标相关性和指标重要度对33个财务指标进行了约简,得到9个能够反映企业财务状况的财务指标。以约简后的9个指标及5个等级的财务状况来建立决策树,指标体系和财务等级更加合理。树的生成过程运用粗糙集中的变精度加权平均粗糙度作为选择测试属性的方法,每次选择变精度加权平均粗糙度值最小的属性作为分支结点。变精度加权平均粗糙度的应用提高了决策树的防噪声能力,复杂性较低且能有效提高分类效果。实证研究表明将它应用到财务预警领域,提高了财务预警的分类精度。
Using the clustering method, the company’s financial status is divided into five levels, which are healthy, good, general, early warning and crisis of financial status. Compared with the financial early-warning model which divided the research sample into ST and non-ST in the past, The model is more accurate and reasonable and closer to reality. At the same time, 33 financial indicators were reduced based on the relevance of the indicators and the importance of the indicators, and nine financial indicators that can reflect the financial status of the enterprises were obtained. With the reduction of 9 indicators and 5 levels of financial status to establish a decision tree, the index system and financial grade more reasonable. The tree generation process uses the variable precision weighted average roughness of the rough set as a method of selecting test attributes. Each time the attribute with the smallest variable precision weighted average roughness value is selected as the branch node. The application of variable-precision weighted average roughness improves the anti-noise ability of the decision tree, has less complexity and can effectively improve the classification effect. Empirical studies show that it is applied to the field of financial early warning, improve the classification accuracy of financial early warning.