Infinite Decision Agent Ensemble Learning System for Credit Risk Analysis

Abstract

Considering the special needs of credit risk analysis, the Infinite DEcision Agent ensemble Learning (IDEAL) system is proposed. In the first level of our model, we adopt soft margin boosting to overcome over fitting. In the second level, the RVM algorithm is revised for boosting so that different RVM agents can be generated from the updated instance space of the data. In the third level, the perceptron kernel is employed in RVM to generate infinite subagents. Our IDEAL system also shares some good properties, such as good generalization performance, immunity to over fitting and predicting the distance to default. According to the experimental results, our proposed system can achieve better performance in term of sensitivity, specificity and overall accuracy.