Grey Prediction Optimization Model And Its Application In The Enterprise Credit Risk Evaluation

By analyzing the characteristics of credit risk, this study reveals that enterprise credit risk consists of intricate relationships and it is a dynamic system which is comprehensive, covert as well as uncertain. Moreover, credit risk rating system of enterprise starts late in China, which leads to the scarcity of historical materials. Through comparing the current prediction models, it indicates that recently the grey prediction models are widely applied in economic areas since it suits for those have small number of samples and inadequate information. However, due to its limitation and applicability in credit risk evaluation index system of enterprise, the predict results of the grey prediction model (GM (1.1) model in this thesis) is slightly deviate from the actual situations. In order to increase the accuracy of prediction, this study analyzes and improves two factors that influences the accuracy of traditional grey prediction model GM(1.1):1. Previous studies found that time series of enterprise credit risk fluctuate randomly, while GM(1.1) model is more suitable for nonnegative monotone sequence. Based on this finding, this study proposes that smooth sequence transform should be employed to revise the initial sequence, therefore it can be transformed to smooth monotone increasing sequence, which matches the situations of grey prediction model. This study applies Themis credit risk data to exemplify, which reveals that the accuracy of prediction increases significantly after optimization.2. During the prediction of enterprise credit risk, the initial step-size exerts great influence on the predict results. Based on metabolism prediction model and regard relative error as selection criteria, this study tests the optimal step-size by employing sequence of average relative error and the optimal initial step length. The result indicates that it is possible to apply this model to select the optimal sequence step-size. Additionally, the combination of smooth sequence and optimal step-size enable the prediction model to achieve the optimum.