一种决策树增量学习算法在故障诊断中的应用

在决策树增量学习算法ID5R上,结合粗糙集理论较强的简化知识、缩小知识库空间的能力,提出了一种增量学习方法,该算法在面对新事例时,能够先对新知识与知识库内容进行比较,将其中已为知识库所有的知识进行压缩,仅对其余部分学习,该过程能够有效减少新示例学习过程的消耗.同时针对知识获取的瓶颈问题,将其应用于故障诊断的知识积累过程中,并通过一个柴油机油样光谱分析故障诊断的算例,验证该算法的有效性.

Application of incremental learning algorithm of decision tree to fault diagnosis

With the ability of incremental learning, fault diagnosis expert system can deal with a new instance without relearning the entire example set. While treating with a new training instance, the existing algorithms almost needed to calculate the entropy of each attribute without considering its importance. Rough set was used in the incremental learning approach of decision tree to find out the most important attributes. Through a set of oil samples, the approach is demonstrated during the learning process of an oil fault diagnosis system.