Approaching a Knowledge Discovery Process for Corporate Bond Classifications via Neural Clustering Technique

摘要

We propose a knowledge discovery process for corporate bond classifications. We implement the process on a financial decision support system, which is able to convert data from various sources into data warehouse, to retrieve data cubes based on different power users' commands using OLAP tool to support data mining modules. One of the data mining solutions we implement is for corporate bond classifications. We first apply the SOFM technique to classify thousands of corporate financial data into clusters. We then employ the LVQ technique to refine selected features. With the use of MyInfoTeller, bond managers can freely select groups of factors to simulate, and to examine and fine-tune initial clustering results based on either their expertise or the. shared knowledge from preliminary results of qualitative analysis. The final learned cluster information will then be stored and managed in the system, and can further be applied to rating predictions of newly issuing bonds, or rating changes of existing bonds.