Credit Index Weight Model and Application of Multiclassification Support Vector Machine Based on Particle Swarm Optimization

Credit Index Weight Model and Application of Multiclassification Support Vector Machine Based on Particle Swarm Optimization

In the current credit evaluation environment with small default samples and unbalanced default status, it is a meaningful research work to explore the weight system of credit indicators with high discriminative power and high precision, which is relatively lacking in the existing researches. First by loss function to enterprise’s credit status is divided into high and low default default and no default three categories, and then support vector machine was optimized by using particle swarm optimization (pso) algorithm of punish coefficient and the kernel function parameter, building has three classification discriminant ability of support vector machine (SVM) discriminant model, finally through the credit identification ability to calculate each index index weights. The characteristic of this paper is that by combining the loss function with the support vector machine of particle swarm optimization, a new idea of credit index weight of multi-class support vector machine under particle swarm optimization is proposed, and a credit index weight system that can reflect the credit status of three types of enterprises is able to be constructed. Compared with the classical entropy weight method and logistic regression weight calculation method, the three-classification SVM weight calculation model studied in this paper has higher credit discrimination ability.