採用層級分析法（AHP）提升信用評分模型品質之研究

Abstract

信用評分模型將統計方法與專家群組的意見相結合，包含將風險因子和權重作組合。在建模過程中，權重的分配是大多數金融機構所關注的核心議題和主要挑戰。傳統上，專家群組根據其對各種風險因子所判斷重要性的觀點來確定權重。然而，該群組的成員往往具有目標上的衝突（如：風險最小化與市場份額最大化的選擇上，孰重孰輕的問題），並且最後的結果可能由單一的決策著的主要意見所主導，這將導致偏差和低落的模型績效。層級分析法（AHP）是一個多目標決策工具，已成功應用於各個領域。這項研究是發展出將AHP方法結合應用到信用評分的過程中，提高了模型的預測能力。在實證的案例資料研究中測試AHP模型，並將結果與原有的評分技術進行比較。研究結果顯示，AHP評分模型顯著提高了信用評分模型的預測能力。

A credit scoring model integrates a statistical method with the opinions of an expert group and therefore contains a combination of risk factors and weights. The assignment of weights in the modeling processes is a central interest of and a prime challenge for most financial institutions. Traditionally, an expert group determines the weights in accordance with the experts’ views on the importance of the various risk factors. However, the group’s members often have conflicting objectives (i.e., risk minimization vs. market share maximization) and can be dominated by a single dominating opinion maker, which leads to bias and poor model performance. The analytic hierarchy process (AHP) is a multiple-criteria decision-making tool that has been successfully used in various fields. This study is the pioneer to apply the AHP method to the credit scoring process to create a model that increases the predictive power. The study then tests the AHP model in a case study and compares the results with existing scoring techniques. The findings suggest that the AHP scoring model significantly improves the credit scoring model’s predictive ability.