Using symbolic and connectionist algorithms to knowledge acquisition for credit evaluation

There are several techniques of artificial intelligence being applied on the financial market, including credit evaluation. This work investigates the performance achieved by different artificial intelligence techniques when applied to credit evaluation. The techniques used were MLP neural networks and two symbolic learning algorithms, CN2 and C4.5. In order to analyze the performance obtained by these techniques, two distinct data sets for credit evaluation were used. The knowledge used by these techniques was also compared to the knowledge extracted from trained neural networks using a knowledge extraction tool.