4th International Workshop on Rough Sets, Fuzzy Sets, and Machine Discovery (RSFD'1996)
開催地
和文:
英文:
Tokyo, Japan
アブストラクト
This paper addresses issues of inductive learning in a multiagent system. The objectives of the research are twofold: the first is a practical one to develop a novel method to facilitate questionnaire data analysis, and the second is a theoretical one to provide a framework for analyzing emergent behaviors of inductive learning agents. The multiagent system we consider consist of agent, each of which has C4.5 inductive learning programs, a common learning goal, a subset of attributes of training examples, and a Genetic-Algorithm (GA) based feature exchange mechanism. This paper proposes a GA-based feature selection algorithm with a new crossover operator: Organizational crossover, which generates offspring from more than two parents. Intermediate experiments show that the system has discovered more accurate and / or simpler results than the ones by a single inductive learning system.
各種検索
サポート
ヘルプ