Support vector machines in ordinal classification

Risk assessment of credit portfolios is of pivotal importance in the banking industry. The bank that has the most accurate view of its credit risk will be the most profitable. One of the main pillars in assessing credit risk is the estimated probability of default of each counterparty, i.e., the probability that a counterparty cannot meet its payment obligations in the horizon of one year. A credit rating system takes several characteristics of a counterparty as inputs, and assigns this counterparty to a rating class. In essence, this system is a classifier whose classes lie on an ordinal scale. This thesis provides an extensive assessment of the abn amro credit rating system. The current rating tool, an expert system, is carefully reviewed. We show that this system has several drawbacks in both its mathematical fundamentals and its implementation. We propose a new credit rating framework, which incorporates an improved version of the current model. Aside from this expert system, we applied linear regression, ordinal logistic regression, and support vector machine techniques to the credit rating problem. The latter technique is a relatively new machine learning technique that was originally designed for the two-class problem. We propose two new techniques that incorporate the ordinal character of the credit rating problem into support vector machines. We show that the current rating model used at abn amro performs in line with statistical and support vector machine techniques. The results of our newly introduced techniques are promising.