應用分類與迴歸樹模型於台灣地區無店面零售產業之信用風險評估

Abstract

以Live直播型態的電視購物頻道在台灣近年來快速蓬勃發展，加上郵購、網路購物等以虛擬通路做為銷售方式的無店面零售業市場產值不斷上升，市場已達到成熟階段。各無店面零售業者為了衝業績，紛紛的推出信用卡分期付款的方式，減少顧客每月負擔的金額，以降低商品購買門檻，刺激消費者的購買意願。 另一方面，在信用卡發卡銀行的宣傳下，近年來信用卡的發卡張數不斷創新高，消費者能更輕易的膨脹自身的信用。在銀行以及企業兩方面的推波助瀾下，消費者的還款能力被過度高估，以致過度濫用個人信用，終於在2005年年底發生了震驚全台的雙卡債風暴，除了造成消費者自身負債累累外，連帶的衝擊發卡銀行與許多的產業，造成企業極為嚴重的壞帳損失。 為了防範企業再次遭受如此巨大的損失，一套良好的信用風險評估模型是極其必要的。然而，目前企業使用的信用風險評估模型較為簡單，多數僅使用從業人員的經驗法則進行信用風險的控管，容易造成偏誤。 本研究利用較客觀精確的統計模型-分類與迴歸樹(CART)-進行信用風險評估模型的建置，並以東森購物的顧客資料庫做為驗證分析的資料來源，希望在考慮多元廣範圍的變項下，可以建立出精確度較高的信用風險評估模型，幫助企業降低壞帳損失。

TV shopping companies using live broadcast are growing rapidly in Taiwan recently. Plus the sales channels of mail order and web shopping, the market value of the virtual store in retail industry is continuously escalating, and the market has arrived the mature stage. In order to raise the sales, all virtual-store companies in the retail industry provide the service of installment payment by credit card to reduce customers’ burden each month. It also lowers down the threshold of the product purchase and inspires consumers’ purchase willing. Moreover, under the marketing by the credit card issue banks, the number of credit cards outstanding is soaring high year by year. The consumers now can expand their credit easily by using the credit cards. Therefore, under the promotion by both banks and firms, consumers’ abilities of installment payment are strongly overvalue. One can use his credit freely without any limitation. As a result, in the end of 2005, there was a card debt crisis which shocked the nation. The crisis not only caused consumers to have heavy debt burden, but also struck card issue banks and many industries, which leaded firms to the serious loss of the overdue payment. In order to prevent firms from having such huge loss again, a well-designed credit risk evaluation model is extremely necessary. However, the credit risk evaluation model which firms use now is simple. Most of them only use the experience from their employees to control the credit risk, which can cause bias easily.

In this research, a more objective and precise statistics model- classification and regression tree (CART) - is used for the construction of the credit risk evaluation model. It also uses consumers’ data from the database of the Eastern Home Shopping and Leisure Company as the source for the verification of the model we construct. It’s expected that under the consideration of various variables, the credit risk evaluation model with higher precision could be established to help firms lower the loss of the overdue payment.