近年來,銀行業者營運重心由企業金融業務轉移至高獲利、高利差的消費金融業務,但因台灣市場規模的限制,銀行家數密集的情況下,使得銀行業的經營環境呈現高度競爭的狀況。為提升銀行業的授信品質,降低授信風險,以增進銀行業獲利的前提下,本研究之目的,為建構消費金融之信用風險模型,期望透過信用風險模型協助銀行業者深入了解客戶、區隔客戶類型,有效降低銀行業的授信風險與作業成本,並進一步擬定正確的市場區隔策略,以利拓展市場佔有率創造獲利空間。論文乃針對徵信內容所提供之各項資料,如客戶基本資料、信用狀況等變數做分析,透過分類迴歸樹、類神經網路、邏輯斯迴歸、分群分析等演算法,評估建置信用風險模型,並探討模型建置的方法與成功關鍵因素。
Recently, the business focus of banking industry has been shifting from corporate banking to consumer banking because of the high profit and high interest spread. Due to the relatively small market scale, the consumer banking business is indeed a highly competitive business in Taiwan. From the risk management’s point of view, the purpose of this research is to build up a credit risks model for consumer banking. By building the model, the bank could evaluate the underlying risk exposures on its consumer banking business so as to come out strategies to decrease the probabilities of bad accounts and reduce the operating costs, eventually, to maximize the profits. In the thesis, we first collect customer information and their transaction records to form a research database. Several algorithms, known as data mining techniques, are applied to build up the credit risk models. The algorithms are CART, Back-propagation Neural Network, Logistic Regression and K-Mean Clustering. The critical success factors for building up a reliable and robust model are discussed in this thesis as well.