Improved Debt Rating Model Using Choquet Integral
Toshihiro Kaino*, Ken Urata**, Shinichi Yoshida***, and Kaoru Hirota****
*School of Business Administration, Aoyama Gakuin University, 4-4-25 Shibuya, Shibuya-ku, Tokyo 150-8366, Japan
**DENTSU INC. 1-8-1 Higashi-shimbashi, Minato-ku, Tokyo 105-7001, Japan
***Information Science Research Center, Aoyama Gakuin University, 5-10-1 Fuchinobe, Sagamihara-shi, Kanagawa 229-8558, Japan
****Department of Computational Intelligence and Systems Science, Interdisciplinary Graduate School of Science and Engineering, Tokyo Institute of Technology, Midori-ku, Yokohama 226-8502, Japan
Improved long-term debt rating model using Choquet integral is proposed, where the input is qualitative and quantitative data of the corporations, and the output is the Moody’s long-term debt ratings. The fuzzy measure, which is given as the importance of each qualitative and quantitative data, is derived from a neural net method. Moreover, differentiation of the Choquet integral is applied to the long-term debt ratings, where this differentiation indicates how much evaluation of each specification influence to the rating of the corporation. A long-term debt rating model using Choquet integral was proposed by Kaino and Hirota . Under the ambiguous information which couldn’t be expressed by the statistics model, this Kaino and Hirota model  enabled analysis of the amount of influences of a specific variable, and showed the new possibility in the field of credit risk measurement. However, in order to develop a practical system for small and medium-sized corporations with many needs, this model must be improved so that it may correspond to the changing market or many types of industry. Moreover, this model is modified by the implementation of actual rating provider’s similar process to raise the relevance ratio. The advanced model proposed herein corporations than the model is more precise than conventional model using 2-layer type neural network model.
This article is published under a Creative Commons Attribution-NoDerivatives 4.0 Internationa License.