小額信貸自動核貸之研究-以 S 銀行為例

銀行走向 Bank3.0 的過程中，科技的特性其實為主要的關鍵，Bank2.0 時代都是被動式的讓顧客搜尋自己的交易狀況，以及被動的進行交易行為，是屬於非同步的行為。但是 Bank3.0 可以及時讓顧客掌控自己交易情況，甚至透過大數據(Big Data)的運算，主動提供適合顧客的交易商品。S 銀行推出小額信貸的線上申請服務，民眾只需要在家裡輸入資料，透過大數據運算來模擬放款金額跟利息，直接線上貸款，將可以省去麻煩的貸款流程，不過現行銀行只能做到線上申請，無法做到線上自動核貸，仍須都以人工核貸後放款，因此本文的主要研究為建立自核核貸系統，串接起線上貸款整體作業，並同時可再進一步探討自動核貸與人工核貸之差異;希望透過此研究結果，有助於應用於實務上，讓銀行能因應數位金融的變化。本研究蒐集了 2017 年 1 月 3 號至 2017 年 6 月 30 號之資料，合計筆數共有 15,480 筆，並針對探討的變數抽樣配對後，再進行 Logit 迴歸模型等計量方法建立實證模型，透過大數據與模型分析等方式驗證整合，利用演算法計算，找到最貼近人工核貸的方式，建立起自動核貸機制，串連起線上貸款整體作業之流程，自動審核機制效率優於傳統人工審核，改善 S 銀行之作業模式，精簡人工作業時間，省去作業成本、人力成本，並加以探討逾期貸款關鍵的因素，將性別、教育程度、收入、服務機構列為影響貸款逾期之重要因素。

關鍵字

小額信貸；逾期貸款；線上申請；自動核貸； Logit 迴歸

並列摘要

In the process of banks moving to Bank 3.0, the characteristics of science and technology are actually the key. In the era of Bank2.0, the passive behavior of allowing customers to search for their trading status and passively conduct trading activities is a non-synchronized behavior. However, Bank3.0 can prompt customers to control their own trading situation, and even through Big Data's calculations, they can actively provide customer-specific trading products. S Bank launched an online application service for microfinance. People only need to input data at home and simulate lending sums and interest through big data calculations. Direct online loans will save the troublesome loan process, but current banks can only do online. Application, can not be automated online loan, still have to use artificial loans after lending, so the main study of this article is to establish a self-nuclear loan system, contiguous starting the overall operation of online loans, and at the same time can further explore the automatic loan Differences from artificial nuclear loans; it is hoped that through the results of this study, it will be useful for practical applications and allow banks to respond to changes in digital finance. This study collected data from January 3, 2017 to June 30, 2017, a total of 15,480 data were used, and after pairing the variables discussed, a logit regression model and other measurement methods were used to establish an empirical model. Data and model analysis and other methods are used to verify and integrate, use algorithms to calculate, find the closest approach to artificial nuclear loans, establish an automatic credit mechanism, connect the whole process of online loans, and theefficiency of the automatic audit mechanism is better than the traditional manual audit. Improve the operating mode of S Bank, streamline the manual operation time, eliminate the operating costs, labor costs, and explore the key factors of overdue loans, and include gender, education, income, and service institutions as important factors affecting the overdue loans.