Please wait a minute...
Data Analysis and Knowledge Discovery  2020, Vol. 4 Issue (6): 80-90    DOI: 10.11925/infotech.2096-3467.2019.1285
Current Issue | Archive | Adv Search |
Evaluation Model for Customer Credits Based on Convolutional Neural Network
Liu Weijiang1,2,Wei Hai2(),Yun Tianhe2
1Center for Quantitative Economics, Jilin University, Changchun 130012, China
2Businesses School, Jilin University, Changchun 130012, China
Download: PDF (912 KB)   HTML ( 18 ) 
Export: BibTeX | EndNote (RIS)      
Abstract  

[Objective] This paper analyzes customer loan information, and extracts their characteristics, aiming to more effectively predict customer defaults of online loans. [Methods] First, we collected customer credit data from Lending Club. Then, we integrated the characteristic variables from four aspects of customer information and created a grayscale map. Finally, we established a customer credit evaluation model based on convolutional neural networks. [Results] The proposed model had specificity of 99.4%, sensitivity of 68.7%, G-mean value of 82.7%, F1 value of 81.4% and AUC value of 99.5%. The performance of our new model was much better than those credit models based on feature processing. [Limitations] We only investigated the performance of a few models. More research is needed to study the impacts of unbalanced data. [Conclusions] The proposed model effectively predicts probability of customer defaults.
Key words： Convolutional Neural Networks      Indicator Imaging      Credit Evaluation      Information Value      PCA     
Received: 29 November 2019      Published: 07 July 2020
ZTFLH:  TP393 G250  
Corresponding Authors: Wei Hai     E-mail: weihai94@163.com
Service
E-mail this article
Add to my bookshelf
Add to citation manager
E-mail Alert
RSS
Articles by authors
Weijiang Liu
Hai Wei
Tianhe Yun

Cite this article:

Liu Weijiang,Wei Hai,Yun Tianhe. Evaluation Model for Customer Credits Based on Convolutional Neural Network. Data Analysis and Knowledge Discovery, 2020, 4(6): 80-90.

URL:

https://manu44.magtech.com.cn/Jwk_infotech_wk3/EN/10.11925/infotech.2096-3467.2019.1285     OR     https://manu44.magtech.com.cn/Jwk_infotech_wk3/EN/Y2020/V4/I6/80

Research Process of Customer Credit Evaluation Combined with Convolutional Neural Network
CNN Structure Designed in This Paper
IV值 预测能力
[0,0.02) 无预测能力
[0.02,0.10) 预测能力低
[0.10,0.30) 预测能力中
[0.30,+∞) 预测能力高
IV Value Corresponding to Prediction Ability Interval
Graphical Sample Data
指标变量 变量名称 IV值 含义
贷款信息 loan_amnt 0.561 借款人申请的贷款金额
int_rate 0.724 贷款利率
偿付能力 annual_inc 0.560 借款人在注册期间自行报告的年收入
dti 0.333 使用借款人的总债务偿还总额（不包括抵押贷款和要求的LC贷款）除以借款人自我报告的月收入计算的比值
tot_cur_bal 0.555 所有账户的当前总余额
il_util 0.685 所有固定账户的总流量余额/信用额度的比率
max_bal_bc 0.710 所有周转账户的最大当前余额
acc_open_past_24mths 0.488 过去24个月的消费额
bc_open_to_buy 0.355 银行卡上可用于购买的资金
mort_acc 0.314 抵押账户数量
num_actv_rev_tl 0.560 当前活跃的循环交易数量
num_bc_tl 0.333 银行卡账户数量
installment_feat 0.306 客户每月还款支出占月收入的比值
历史信贷 open_acc 0.489 借款人信用档案中的未结信用额度
all_util 0.534 所有交易均衡信贷限额
total_bal_il 0.394 所有分期付款账户的当前总余额
revol_bal 0.780 总信贷周转余额
revol_util 0.564 循环利用率,或借款人相对于所有可用循环信贷使用的信贷额度
pct_tl_nvr_dlq 0.489 从未拖欠交易百分比
历史申请 mo_sin_old_il_acct 0.577 自最早开立银行分期账户以来的月数
mo_sin_old_rev_tl_op 0.441 自最早的循环账户开始以来的月数
mo_sin_rcnt_rev_tl_op 0.349 自最近一次的循环账户开通以来的月数
mo_sin_rcnt_tl 0.461 自最近一次开户以来的月数
mths_since_recent_bc 0.561 自最近一次开立银行卡账户以来的月数
mths_since_recent_inq 0.724 自最近的调查以来的月数
mths_since_rcnt_il 0.604 自最近的分期付款账户开通以来的月数
Index Variable System
模型 参数或结构设置
LeNet-5 参考图1结构
BP神经网络 采用传统三层BP神经网络,参数设定参考文献[19]
决策树 max_depth:7
支持向量机 kernal:rbf,c:100,gamma:0.01
随机森林 max_depth:8,min_samples_leaf:4
Logistic回归 c:0.1,penalty:l1
Machine Learning Models’ Parameter Settings
真实情况 预测结果
正例 反例
正例
(Positive)		 真正例
(True Positive,FP)		 假反例
(False Negative,FN)
反例
(Negative)		 假正例
(False Positive,FP)		 真反例
(True Negative,FN)
Confusion Matrix
特征处理方法 模型方法 灵敏度 特异度 查准率
LeNet-5① 0.687(1) 0.994(1) 0.998(1)
基于信息价值特征处理 BP神经网络 0.616 0.723 0.695
决策树 0.625 0.706 0.679
支持向量机 0.657(3) 0.751 0.724
随机森林 0.663(2) 0.717 0.744
Logistic回归 0.617 0.649 0.635
基于PCA特征处理 BP神经网络 0.648 0.889(3) 0.946(3)
决策树 0.636 0.839 0.911
支持向量机 0.657(3) 0.911(2) 0.956(2)
随机森林 0.641 0.872 0.930
Logistic回归 0.648 0.889(3) 0.656
Models’ Recognition Performance of Positive and Negative Samples
特征处理方法 模型方法 G-mean F1 AUC
LeNet-5 0.827（1） 0.814（1） 0.995（1）
基于信息价值
特征处理		 BP神经网络 0.668 0.653 0.715
决策树 0.665 0.651 0.724
支持向量机 0.703 0.689 0.668
随机森林 0.705 0.701 0.778
Logistic回归 0.633 0.626 0.747
基于PCA
特征处理		 BP神经网络 0.759（3） 0.769（3） 0.816
决策树 0.730 0.749 0.837
支持向量机 0.773（2） 0.778（2） 0.904（2）
随机森林 0.748 0.759 0.865
Logistic回归 0.759（3） 0.652 0.880（3）
Models’ Performance Evaluation
[1] 蒋辉, 马超群, 许旭庆, 等. 仿EM的多变量缺失数据填补算法及其在信用评估中的应用[J]. 中国管理科学, 2019,27(3):11-19.
[1] ( Jiang Hui, Ma Chaoqun, Xu Xuqing, et al. An EM-similar Imputation Algorithm for Multivariable Data Missing and Its Application in Credit Scoring[J]. Chinese Journal of Management Science, 2019,27(3):11-19.)
[2] 肖进, 刘敦虎, 顾新, 等. 银行客户信用评估动态分类器集成选择模型[J]. 管理科学学报, 2015,18(3):114-126.
[2] ( Xiao Jin, Liu Dunhu, Gu Xin, et al. Dynamic Classifier Ensemble Selection Model for Bank Customer’s Credit Scoring[J]. Journal of Management Sciences in China, 2015,18(3):114-126.)
[3] Altman E I. Financial Ratios, Discriminant Analysis and the Prediction of Corporate Bankruptcy[J]. The Journal of Finance, 1968,23(4):589-609.
doi: 10.1111/j.1540-6261.1968.tb00843.x
[4] Wiginton J C. A Note on the Comparison of Logit and Discriminant Models of Consumer Credit Behavior[J]. Journal of Financial Quantitative Analysis, 1980(15):757-770.
[5] 吴冲, 夏晗. 基于支持向量机集成的电子商务环境下客户信用评估模型研究[J]. 中国管理科学, 2008,16(S1):362-367.
[5] ( Wu Chong, Xia Han. Study of Customer Credit Evaluation Under E-commerce Based on Support Vector Machine Ensemble[J]. Chinese Journal of Management Science, 2008,16(S1):362-367.)
[6] Blanco A, Pino-Mejías R, Lara J, et al. Credit Scoring Models for the Microfinance Industry Using Neural Networks: Evidence from Peru[J]. Expert Systems with Applications, 2013,40(1):356-364.
doi: 10.1016/j.eswa.2012.07.051
[7] Chen F L, Li F C. Combination of Feature Selection Approaches with SVM in Credit Scoring[J]. Expert Systems with Applications, 2010,37(7):4902-4909.
doi: 10.1016/j.eswa.2009.12.025
[8] 熊志斌, 吴维烨. 基于深度信念网络的信用评估研究[J]. 科研信息化技术与应用, 2019,10(3):28-36.
[8] ( Xiong Zhibing, Wu Weiye. Credit Evaluation Research Based on Deep Belief Networks[J]. E-science Technology & Application, 2019,10(3):28-36.)
[9] 吴星泽. 财务危机预警研究:存在问题与框架重构[J]. 会计研究, 2011(2):59-65,97.
[9] ( Wu Xingze. Problems on Research of Predicting Financial Distress and Framework Reconstructure[J]. Accounting Research, 2011(2):59-65, 97.)
[10] He K M, Zhang X Y, Ren S Q, et al. Delving Deep into Rectifiers: Surpassing Human-Level Performance on ImageNet Classification [C]// Proceedings of the 2015 IEEE International Conference on Computer Vision. 2015: 1026-1034.
[11] Sun S N, Zhang B B, Xie L, et al. An Unsupervised Deep Domain Adaptation Approach for Robust Speech Recognition[J]. Neurocomputing, 2017,257:79-87.
doi: 10.1016/j.neucom.2016.11.063
[12] Williamson D S, Wang D L. Time-frequency Masking in the Complex Domain for Speech Dereverberation and Denoising[J]. IEEE/ACM Transactions on Audio, Speech, and Language Processing, 2017,25(7):1492-1501.
doi: 10.1109/TASLP.2017.2696307 pmid: 30112422
[13] Zhang Y, Marshall I, Wallace B C. Rationale-augmented Convolutional Neural Networks for Text Classification [C]// Proceedings of the 2016 Conference on Empirical Methods in Natural Language Processing. 2016: 795-804.
[14] 李慧, 柴亚青. 基于卷积神经网络的细粒度情感分析方法[J]. 数据分析与知识发现, 2019,3(1):95-103.
[14] ( Li Hui, Chai Yaqing. Fine-Grained Sentiment Analysis Based on Convolutional Neural Network[J]. Data Analysis and Knowledge Discovery, 2019,3(1):95-103.)
[15] Hosaka T. Bankruptcy Prediction Using Imaged Financial Ratios and Convolutional Neural Networks[J]. Expert Systems with Applications, 2019,117:287-299.
doi: 10.1016/j.eswa.2018.09.039
[16] Lencun Y, Bottou L, Bengio Y, et al. Gradient-Based Learning Applied to Document Recognition[J]. Proceedings of the IEEE, 1998,86(11):2278-2324.
doi: 10.1109/5.726791
[17] 阚士行. 商业银行信用评级筛选财务指标方法效果对比与校验[D]. 济南:山东大学, 2010.
[17] ( Kan Shixing. Performance Comparison of Several Methods for Selecting Indices of Commercial Bank Credit Ranking[D]. Ji’nan: Shandong University, 2010.)
[18] 刘丹, 李战江, 郑喜喜. 基于WOE-Probit逐步回归的信用指标组合筛选模型及应用[J]. 数学的实践与认识, 2018,48(2):76-87.
[18] ( Liu Dan, Li Zhanjiang, Zheng Xixi. Selection Model of Credit Index Combination Based on WOE-Probit Stepwise Regression and Its Application[J]. Mathematics in Practice and Theory, 2018,48(2):76-87.)
[19] 杨淑娥, 黄礼. 基于BP 神经网络的上市公司财务预警模型[J]. 系统工程理论与实践, 2005,25(1):12-18,26.
[19] ( Yang Shu’e, Huang Li. Financial Crisis Warning Model Based on BP Neural Network[J]. Systems Engineering-Theory & Practice, 2005,25(1):12-18, 26.)
[1] Lin Kerou,Wang Hao,Gong Lijuan,Zhang Baolong. Disambiguation of Chinese Author Names with Multiple Features[J]. 数据分析与知识发现, 2021, 5(4): 90-102.
[2] Xu Yuemei,Lv Sining,Cai Lianqiao,Zhang Xiaoya. Analyzing News Topic Evolution with Convolutional Neural Networks and Topic2Vec[J]. 数据分析与知识发现, 2018, 2(9): 31-41.
[3] Huai Mengjiao,Pan Yuntao,Yuan Junpeng. Evaluating Academic Credits of Scientific Research Project Leaders[J]. 数据分析与知识发现, 2017, 1(11): 94-102.
[4] Yaming Zhang,Na Li,Peiqing Zhao. Study on Credit Evaluation Model of Online Group-buying by Using ACO and Similarity Weight Algorithm[J]. 现代图书情报技术, 2016, 32(1): 40-47.
[5] Yu Xianzi, Gao Yinglian, Ma Chunxia, Liu Jinxing. The Penalized Matrix Decomposition Method of Extracting Core Characteristic Words——Taking Co-word Analysis as an Example[J]. 现代图书情报技术, 2014, 30(3): 88-95.
  Copyright © 2016 Data Analysis and Knowledge Discovery   Tel/Fax:(010)82626611-6626，82624938   E-mail:jishu@mail.las.ac.cn