GA改进的统计组合模型在个人信用评估中的应用

Application of Combining Forecast Model Based on Statistical Methods Optimized by GA in Personal Credit Scoring

将分类精度作为遗传算法的适应度函数,利用遗传算法改进了基于线性回归和Logistic回归的统计组合预测模型,并将其应用于个人信用评估。结果表明,遗传算法改进后的组合预测模型在总分类精度和两类误判率方面都低于单一模型以及基于误差平方和最小的组合预测模型。利用遗传算法改进组合预测模型并用于个人信用评估是具有优势的。

Using predictive accuracy as the fitness function of genetic algorithm, this paper established a combining forecast model based on linear regression and logistic regression and used this combining forecast model in personal credit scoring. The resuits indicate that the combining model has a higher predictive accuracy and lower misclassification rates than each single model and the combining model based on the least sum of square error. The combining forecast model based on genetic algorithm has advantages in personal credit scoring.