Enterprise Credit Confidence Assessment Based On TCM Framework

Enterprise credit assessment evaluates the trusted degree of an enterprise which indicates whether the enterprise has the ability of debt service. This assessment is characterized as a high-risk financial decision-making analysis. However, there is a somewhat neglected aspect of evaluation that estimates the confidence of the assessment as well as the trusted degree. We propose a confidence machine, named transductive confidence machine (TCM) for enterprise credit assessment for the first time.TCM, which is found rooted in the Kolmogorov algorithm randomness theory, is a universal framework of plugging machine learning algorithms to output prediction with confidence. Given the precise definition of nonconformity description function, the universal randomness test can be implemented approximately. In the paper, some classifiers (KNN&SVM) are embedded into the framework of TCM and then the corresponding TCMs are applied to the enterprise credit confidence assessment, which obtains the confidence of the assessment and enriches the output information of learning machines. The results of experiments show that enterprise credit confidence assessment based on TCM framework is more reliable and flexible.The outline of the paper is as follows: Firstly, the TCM-KNN algorithm based on Euclidean distance was applied to enterprise credit confidence assessment, then the improved algorithm that TCM-KNN based on kernel metric was proposed and then applied to enterprise credit confidence assessment to get an improved performance; Secondly, owing to the kernel technique and superiority model of SVM, TCM-SVM (applying Lagrange multiplier to be nonconformity function) was introduced into enterprise credit confidence assessment and got a satisfying result; Finally, except for the standard TCM-SVM, an improved TCM-SVM algorithm was proposed and then applied to enterprise credit confidence assessment to get a better result.In short, TCMs are applied to enterprise credit assessment successfully, on the other hand, it is facilitated that enterprise credit confidence assessment outputs with confidence value. We confirm that enterprise credit confidence assessment based on TCM framework is crucial in real-world application.