计算机科学
模糊性
推理规则
前因(行为心理学)
知识库
模糊规则
基于规则的系统
推论
知识表示与推理
人工智能
基于知识的系统
信念结构
代表(政治)
基础(拓扑)
机器学习
模糊逻辑
数据挖掘
模糊集
数学
法学
发展心理学
数学分析
政治
心理学
政治学
作者
Jun Liu,Luis Martı́nez,Alberto Calzada,Hui Wang
标识
DOI:10.1016/j.knosys.2013.08.019
摘要
Advancement and application of rule-based systems have always been a key research area in computer-aided support for human decision making due to the fact that rule base is one of the most common frameworks for expressing various types of human knowledge in an intelligent system. In this paper, a novel rule-based representation scheme with a belief structure is proposed firstly along with its inference methodology. Such a rule base is designed with belief degrees embedded in the consequent terms as well as in the all antecedent terms of each rule, which is shown to be capable of capturing vagueness, incompleteness, uncertainty, and nonlinear causal relationships in an integrated way. The overall representation and inference framework offers a further improvement and great extension of the recently developed belief Rule base Inference Methodology (refer to as RIMER), although they still share a common scheme at the final step of inference, i.e., the evidential reasoning (ER) approach is applied to the rule combination. It is worth noting that this new extended belief rule base representation is a great extension of traditional rule base as well as fuzzy rule base by encompassing the uncertainty description in the rule antecedent and consequent. Subsequently, a simple but efficient and powerful method for automatically generating such extended belief rule base from numerical data is proposed involving neither time-consuming iterative learning procedure nor complicated rule generation mechanisms but keeping the relatively good performance, which thanks to the new features of the extended rule base with belief structures. Then some case studies in oil pipeline leak detection and software defect detection are provided to illustrate the proposed new rule base representation, generation, and inference procedure as well as demonstrate its high performance and efficiency by comparing with some existing approaches.
科研通智能强力驱动
Strongly Powered by AbleSci AI