유전자 다중 에이전트 규칙유도를 이용한 기업 부도 예측 방안

초록

The purpose of this paper is to develop a new corporate bankruptcy prediction method based on a mixture of multiagent rule induction system and genetic algorithm and to show its efficiency by applying it to real world data. To serve this purpose, we propose an algorithmic model, which derives rules from decision trees, calculates matching rates, and selects the optimal rules using genetic algorithm’s crossover methodology. Specifically, on rule selection stage, we use a voting mechanism. We next locate the optimal set of rules, which maximizes the correct classification ratios. Experimental result shows that the application of our method raises the prediction accuracy gradually as the generation increases, comparing to other existing methods while providing more diversity in genetic methods.

키워드