1 AIT Asian Institute of Technology

Case-based reasoning using conceptual graphs

AuthorSookapat Niyomthai
Call NumberAIT Thesis no.CS-90-22
Subject(s)Problem solving--Data processing
NoteA thesis submitted in partial fulfillment of the requirements for the degree of Master of Engineering, School of Engineering and Technology
PublisherAsian Institute of Technology
AbstractCase-based reasoning is a problem solving method that involves recalling a previous similar situation and the application of its solution to a current problem. Using previous experiences in problem solving, a case-based problem solver can suggest a plan to use in resolving a new problem, can focus on relevant details of a new situation, and can also avoid errors made in the past. This study introduces the conceptual graph notation to be used as a knowledge representation scheme for case-based reasoners. The greatest advantage offered by this approach to traditional schemes such as frames, attribute vectors, and sets of features is that bulk of information on each case can be encapsulated instead of having all features of the case being exposed in detail. Firstly, encapsulation of is-a information in the concept type hierarchy lets two different concepts be matched using only their types without explicitly stating or referring to any detailed properties that they have in common. Secondly, conceptual relations are used to represent relations between features and .let cases be represented with structural information. The framework in building up a case-based problem solver based on conceptual graph formalism is developed in this study with emphasis on two parts. The first part is the design and development of mechanisms for retrieval of cases that have its problem descriptions represented as conceptual graphs . Conceptual graph matching based on graph similarity measure is developed for this task with a mechanism for fast matching from large case storage. The second part is the design of case-based inference processes. A general organizational structure of a case-based problem solver is designed which results in the development of an implementation framework that includes the representation of knowledge and algorithms of main processes. Finally, a demonstration of how the framework is used in constructing a case- based problem solver for a real-world application is given. The case-based problem solver doing advertising media selection task is implemented under the designed framework as a prototype system that chooses an appropriate medium for the advertisement of a product.
Year1990
TypeThesis
SchoolSchool of Engineering and Technology (SET)
DepartmentOther Field of Studies (No Department)
Academic Program/FoSComputer Science (CS)
Chairperson(s)Vilas Wuwongse
Examination Committee(s)Sadananda, Ramakoti ;Phan Minh Dung
Scholarship Donor(s)Government of Japan
DegreeThesis (M.Eng.) - Asian Institute of Technology, 1990


Usage Metrics
View Detail0
Read PDF0
Download PDF0