利用数据挖掘探讨信用卡违约用户分析

金融改革以来，这是消费金融产品中最活跃的业务，而推动“信用卡”是最成功的开证行面临申请的案件，如果仅凭是否出具信用卡和信用证审查的人员，会造成人力资源和浪费时间，也容易因为个人判断标准不影响信用质量和审核。另一方面，對已持卡者之信用狀況的管理亦相當重要，由於信用風險涉及許多層面，因此近年來所被重視的資料採礦技術就成為其重要的一項工具。若能及早發現可能產生呆帳等違約情形之持卡者，對其行為進行監控，將可有效預防違約的發生。另一方面，它也是管理持卡人的信用状况很重要。由于信用风险的诸多方面，数据挖掘技术已成为近年来的重要工具.。如果能够发现信用卡坏账可能发生的违约行为，对其行为进行监督，可以有效防止违约的发生。因此，本研究主要目的在於利用商業智慧與資料採礦的技術整合，希冀建立一套相對穩定且有效的預測模型，提供相關部門與發卡機構一個準則，以降低違約比例，進而降低信用風險，藉此提升該銀行在市場上的評價。因此，本研究的主要目的是利用技术集成的商务智能和数据挖掘，希望建立一个相对稳定的、有效的预测模型，为相关部门和发行人的标准，以降低违约金比例，从而降低信用风险，提高市场对银行的评价。根據本研究所選定羅吉斯迴歸當作分析工具後，在經過不同的抽樣方式，其所建置出來的模型，對於整體預測度、非違約戶的預測度、與違約戶的預測度皆有七成以上的預測能力，與國內外相關研究團隊所建出 70%～80%預測度的模型相距不大，顯示出本研究的模型具有一定的預測水準，所以在分析工具的選擇及抽樣的過程，亦可提供給爾後從事相關研究者作為參考。根据本研究选取 logistic 回归分析工具，通过不同的抽样方法，建立的模型，预测总体程度和非默认的家庭预期违约，和家庭的能力的预测 70%以上，与国内外相关研究团队建立了从70%到 80%的预测模型没有显示在这项研究中，该模型具有一定的预测水平，所以在选择的过程中，抽样分析工具，可以提供对未来的相关研究作为参考。

關鍵字

数据挖掘；信用卡；違約；决策树；类神经网络； R

並列摘要

Since the financial reform, which is the most active consumer financial products business, and promote the credit card is the most successful. The issuing bank faces for the case, if only by whether issued credit cards and credit review personnel, will cause the human resources and waste of time, but also easily because of personal judgment standard does not affect the credit quality and audit. On the other hand, it is also very important to manage the credit status of the cardholder. Because of the many aspects of the credit risk, the data mining technology has been an important tool in recent years. If we can find possible breach of credit card bad debts, to monitor their behavior, can effectively prevent the occurrence of default. On the other hand, it is also important to manage the cardholder's credit standing. Due to many aspects of credit risk, data mining technology has become an important tool in recent years. If we can find the credit card bad debt may be the default behavior, the conduct of its supervision, can effectively prevent the occurrence of default. Therefore, the main purpose of this study was to use the technology integration of business intelligence and data mining, hoping to establish a prediction model of relatively stable and effective, provide a criterion for the relevant departments and issuers, to reduce the proportion of default, thereby reducing the credit risk, to improve the evaluation of the bank on the market. Therefore, the main purpose of this study is the use of technology of integrated business intelligence and data mining, to establish a stable and effective prediction model for the relevant departments and the issuer's standard, to reduce the proportion of default, so as to reduce credit risk, improve the evaluation on the banking market. According to this study selected logistic regression as analysis tools, through different sampling methods, the establishment of the model, to predict the overall degree and non default households forecast, and default households are forecasting ability of more than 70%, and the domestic and foreign related research team built from 70% to 80% the prediction of the model are not shown in this study, the model has certain forecast level, so in the process of selection and sampling analysis tools, can be provided for the future related research as a reference. According to this study selected logistic regression analysis tools, through different sampling methods, establish the model to predict the overall degree and non default family anticipatory breach of contract, more than 70% predictive ability and family, and the domestic and foreign related research team built from 70% to 80% of the forecast model is not shown in this study, the the model has a certain level of prediction, so in the process of selection, sampling and analysis tools, can provide as a reference for future research.