摘要
针对贷款市场中的客户信用评分建立需求,本文基于采集的某区域贷款信用数据,使用机器学习算法进行信用评分模型的构建从而预测出客户的放贷风险等级。首先进行数据清洗以及探索性分析,获取到完整备用的信用贷款数据;其次利用皮尔森相关性分析和热力图完成特征自变量以及因变量的选取和处理;最后采用LightGBM模型进行训练,并与多个主流预测算法进行对比分析。本文完整算法模型在预测精确度、召回率以及F-1 Score评价指标均可达到97%以上。
To meet the demand of customer credit scoring in the loan market,the paper uses machine learning algorithm to build a credit scoring model based on the collected credit data of a region to predict the customer’s lending risk level.Firstly,data cleaning and exploratory analysis were carried out to obtain complete standby credit loan data;Secondly,Pearson correlation analysis and thermodynamic diagram are used to select and process characteristic independent variables and dependent variables;Finally,LightGBM model is used for training,and compared with many mainstream prediction algorithms.The complete algorithm model in the paper can achieve more than 97% in prediction accuracy,recall rate and F-1 Score evaluation index.
作者
赵兴文
ZHAO Xingwen(Department of Information Technology,Zhejiang Financial College,Hangzhou,China,310018)
出处
《福建电脑》
2023年第2期31-34,共4页
Journal of Fujian Computer
基金
浙江金融职业学院青年科研项目课题基金(No.2022YB44)资助。