由於科技不斷進步、網際網路的快速發展加上各種儲存設備的製造成本大幅的降低，企業可以輕易地以較低的成本獲得顧客各種資料，大幅節省了資料調查、蒐集所需的人力、物力及財力。也因此在快速地累積了龐大的資料後，常無法有效地加以分析。而這些資料往往隱藏許多重要的資訊，若能有效地運用這些資訊，進行正確的處理，並經由適當的分析方法取得相關的重要資訊，在經營或行銷策略的制定上，定有相當的輔助效果。 本研究嘗試提出一整合類神經網路與分類迴歸樹（classification and regression tree, CART）及多元適應性雲形迴歸（multivariate adaptive regression splines, MARS）於信用評等模式問題之建構程序，而為了評估此二階段分類模式的判別能力，以某信用卡發卡銀行提供之信用卡持有者之信用資料建構分類模式，並將分析之結果與單純使用倒傳遞類神經網路之分析結果進行比較。研究結果顯示，二階段分類模式的判別結果明顯有優於倒傳遞類神經網路模式之結果。此外，在考慮銀行較關心的損失成本上，二階段分類模式亦有較低的型二錯誤率。 The artificial neural network is becoming a popular alternative in handling credit scoring tasks due to its associated memory characteristic and generalization capability. However, the decision of network's topology, importance of potential input variables and the long training process has often long been criticized and hence limited its application. The classification and regression tree (CART) and multivariate adaptive regression splines (MARS) are becoming very popular alternatives in prediction and classification task due to their capability in modeling nonlinear relationship among variables. Aiming at improving the above-mentioned drawbacks of neural network, the objective of the proposed study is to explore the performance of data classification by integrating neural network, CART, and MARS. To demonstrate the effectiveness of the proposed approach, classification tasks are performed on one credit card dataset. As the results reveal, the proposed integrated approach outperforms the traditional neural network approach in terms of classification accurancy and misclassification cost.
