【摘要】 主体构建和学习环境(ABLE)是IBM开发的用于支持主体建模、构建多主体系统的框架.本文改进了ABLE的决策树算法,在其中计算属性的信息增益时加入一个模糊因子,使算法能更简单、有效地学习.在UCI数据集上的测试结果表明改进后的算法在不降低精度的前提下,可以生成更简洁的决策树,特别是处理复杂的数据时,可显著降低规则数目.更多还原

【Abstract】 Agent Building and Learning Environment(ABLE) is a Java-based framework for development and deployment of intelligent agents with learning abilities.The enhancement of the decision tree algorithm used in ABLE is studied.A fuzzy factor is introduced to calculate the information gain of an attribute during the course of the algorithm.Experimental results on UCI benchmark show that the enhanced algorithm can classify data in an easier and more effective manner,without loss of accuracy.The number of rules generated can be significantly reduced when a large 1data set is encountered.更多还原