基于特征提取和集成学习的个人信用评分方法

Personal Credit Scoring Method Based on Feature Extraction and Ensemble Learning

在大数据蓬勃发展的今天,信息经济已经深入社会方方面面,个人信用体系建设的重要性越发突出。而传统的信用体系存在覆盖率不足、评价特征维度高、数据孤岛等问题,为了解决以上问题,提出一种基于特征提取和Stacking集成学习的个人信用评分方法(PSL-Stacking)。方法首先利用Pearson和Spearman系数对数据进行初始化分析剔除不相关数据,利用LightGBM算法进行特征选择,减少冗余特征对模型的影响;其次选取XGboost、LightGBM、Random Forest以及Huber回归等算法,利用Stacking集成学习技术构造个人信用评分模型。最后,以某电信数据为研究对象,对该上述模型的个人信用评分能力进行验证。实验结果得出上述模型具有很好的预测能力,能够准确的对用户信用进行评分,有效降低企业遭受金融欺诈、团伙套利等问题的风险。

With the vigorous development of big data today,the information economy has penetrated into all aspects of society,and the importance of the construction of personal credit system has become more and more prominent.However,the traditional credit system has problems such as insufficient coverage,high evaluation feature dimensions,and data islands.In order to solve the above problems,a personal credit scoring model(PSL-Stacking)based on feature extraction and stacking integrated learning is proposed.The model first uses the Pearson and Spearman co--efficients to initialize and analyze the data to eliminate irrelevant data,and uses the LightGBM algorithm for feature selection to reduce the impact of redundant features on the model.An ensemble learning technique constructs a personal credit scoring model.Finally,taking a certain telecom data as the research object,the personal credit scoring ability of the model is verified.The experimental results show that the model has good prediction ability,can accurately score users'credit,and effectively reduce the risk of enterprises suffering from financial fraud,gang arbitrage and other problems.