Third International Conference on Artificial Intelligence Applications on Wall Street (AIAW'95), June 7-9, 1995, New York, New York, USA
credit scoring; reinforcement learning; apprentissage par renforcement
There is an increasing need for credit decision making systems that can dynamically analyze historical data and learn complex relations among the most important attributes for loan evaluation. In this paper we propose the application of a new machine learning algorithm, QLC, to the credit analysis of consumer loans. The algorithm learns how to classify a loan by minimizing the expected cost due to both credit investigation expenses and possible misclassification. QLC is built upon reinforcement learning. A dataset of actual consumer loans issued for evaluating the algorithm. The experiments reported show that QLC performs better than other cost-sensitive algorithms on this dataset.
Proceedings of the Third International Conference on Artificial Intelligence Applications on Wall Street (AIAW'95).