一种改进的专家系统自学习算法的研究  

Research of Self-learning Algorithm of a improved Expert System

作　　者：王继东[1] 常瑞[1] 苏海滨[1] 王玲花[1] 

机构地区：[1]华北水利水电学院电力学院,河南郑州450008

出　　处：《微计算机信息》2009年第18期263-264,250,共3页Control & Automation

基　　金：河南省科技攻关资助项目(0524260049);华北水利水电学院青年科技基金项目(HSQJ200514)

摘　　要：基于证据推理的置信规则库推理方法(RIMER)已被提出,由此产生了一种新型专家系统—RIMER专家系统。该专家系统的学习训练模型是一个带有线性约束的复杂非线性优化模型,传统优化方法求解该类模型有一定困难和局限性。本文结合梯度法和二分法提出一种新的优化算法实现了RIMER专家系统的自学习。采用该算法对一个实例进行了训练,训练结果令人满意。训练实例表明新算法具有简单、速度快、收敛精度高等特点。A belief rule-base inference methodology using the evidential reasoning approach （RIMER） has been developed, the RIMER expert system is explored on the basis of RIMER. Parameters in the belief rule-base such as rule weights, weights of antecedent attributes and belief degrees are determined by experts. But generally, it may be difficult to determine these parameters objectively and precisely, which may hinder the RIMER expert system from imitating real systems. So the learning model of RIMER expert system is developed. The model is a complex nonlinear optimization problem with linear constraints and the conventional optimization methods have limitations in solving such a training model. Based on the gradient and dichotomy combined methods, a new simple optimization algorithm is proposed in this paper to train parameters in a belief rule-base. The algorithm has been applied to train parameters in an existing belief rule-base and the trained results are satisfactory. The results show that the new algorithm is simple, fast and effective.

关 键 词：专家系统 学习模型 置信规则库 梯度法 二分法 

分 类 号：TP182[自动化与计算机技术—控制理论与控制工程]