Rule-Based Models with Information Granules: Enhancements and Applications

  • Author / Creator
    Cui, Ye
  • Rule-based models have enjoyed a great deal of interest in the previous decades as one of the development paradigms of intelligent systems. They are regarded as a fundamental vehicle of knowledge representation, serving as a computational platform supporting an array of design practices and analysis of knowledge-based systems and their applications. With the rapidly growing complexity of real-world systems and their ensuing models, and diversity and quality of distributed sources of data, the quest for designing advanced rule-based models becomes evident. In the design of rule-based models, there are two challenging issues. The first is about the scalability of rule-based models when we are faced with high-dimensional data. The equally important task when building rule-based models is to endow such models with a sound measure of quality with which one can efficiently assess the relevance of the results produced by the rules.

    In this study, some key design objectives are formulated and pursued. When faced with high-dimensional data, some fundamental limitations such as the concentration effect hamper the design of high-quality rules or even make the design of monolithic models infeasible. To alleviate this problem, an idea of distributed fuzzy rule-based models is formulated, instead of a single monolithic (multivariable) rule-based model. In its realization, a slew of low-dimensional models is built and aggregated. The aggregation is realized by some linear linkage transformation. Such ensembles of models help us avoid a negative effect of the concentration effect. Next, a novel concept of the granular rule-based model is investigated. We show that granular models quantify the relevance of the original rule-based architectures and deliver a granular format of results, namely prediction intervals. The granular results are optimized by engaging the criteria of coverage and specificity of information granules. Subsequently, in order to improve the quality of the models, a comprehensive and systematic way of ranking alternatives in the environment of multicriteria group decision making is proposed by introducing information granules. The underlying decision process is realized with the use of the analytical hierarchy process (AHP), resulting in information granules (fuzzy sets) quantifying degrees of preference and relevance of the weights. A series of experiments are carried out to examine the feasibility of the proposed methods.

  • Subjects / Keywords
  • Graduation date
    Fall 2022
  • Type of Item
  • Degree
    Doctor of Philosophy
  • DOI
  • License
    This thesis is made available by the University of Alberta Library with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.