个人信用评分模型的发展及优化算法分析

为有效识别与计量个人信用风险,规避金融危机对商业银行的不利影响,保持我国信贷和金融市场的正常运转,对个人信用评分的主要模型及其发展进行了归纳,阐明个人信用评分研究中仍存在信用样本的有效性及完整性问题、信用指标体系的合理性问题以及模型的选择及适用性问题.鉴于此,基于相关性分析实现异常样本的预警,基于蒙特卡洛算法对样本进行补足;结合统计学模型及人工智能模型,采用步长遍历算法对指标体系进行优化及显著性加权;以精确度、稳健性、第1误判率、第2误判率及差异性作为选择指标,实现评分模型的选择与输出.分析表明:通过上述优化算法,将解决个人信用评分中存在的问题,提高商业银行的风险控制能力.

Research on algorithms development and optimization for personal credit scoring

In order to effectively recognize and measure individual credit risk and avoid the negative impact which financial crisis has on the commercial banks, and thus keep the normal operation of the credit and financial markets in China, this paper summarizes the main model of personal credit scoring and its development, and also clarifies the validity and integrity of the credit samples, the rationality of the index system and the selection, applicability of the model in the study of personal credit rating. In view of this, based on correlation analysis, early warning of abnormal samples can be achieved, and the samples are complemented through the Monte Carlo Algorithm. Combined with the statistical model and artificial intelligence model, the index system is optimized and given respective significant weight through traversal algorithm. With accuracy, robustness, first misjudgment rate, second misjudgment rate and difference as the selection index, the selection and output of the scoring model can be realized. Through the above optimization algorithm, the problem in personal credit scoring is solved, and the risk control ability of commercial banks is improved as well.