基于混合BP神经网络算法的信用卡消费行为风险预测

Risk Prediction of Credit Card Consume Behavior Based on Mixed BP Neural Network Algorithm

随着中国信用卡市场的急速发展,信用卡消费行为的风险评估已成为业界研究的一个重要方向。目前风险预测的研究常采用单一的BP神经网络算法,但该算法存在一些固有缺点,如易陷入局部极小点、收敛速度较慢等,这些缺点会影响风险预测的效果。针对单一BP神经网络算法的不足,提出了一种将BP神经网络算法与遗传算法相结合的混合算法,它以BP神经网络作为基础,利用遗传算法对BP神经网络进行优化,并通过数据集的实验证明该混合算法要优于单一BP神经网络算法,可以有效提高信用卡消费行为风险评估中的检测率和准确率。

With the development of credit card market in China, consume behavior of risk prediction of credit card has been an important research area. BP algorithm is often used to solve the risk prediction problem, but it has some disadvantage, such as easily trapped into local minima value, slowly convergence, these limitations would effect risk prediction. In this paper, the author proposed a mixed BP neural network algorithm, based on BP neural network algorithm, it optimized by GA. Finally, this paper proved that the mixed algorithm is better than BP through data set test, and it could improve the test rate and correct rate of credit card consume behavior risk prediction.