You are here:

Design and evaluation of a knowledge-based expert system as a decision support tool in selecting statistical methods
DISSERTATION

, University of Houston, United States

University of Houston . Awarded

Abstract

Expert systems are sophisticated computer programs that attempt to capture enough of the human expert's expertise so that it has the capability of solving problems expertly and reaching the same conclusions that human experts would reach. Unlike conventional computer programs, an expert system can explain itself by describing why some line of questioning is relevant as well as explaining how it arrived at its recommendation.

In this study the researcher designed and evaluated an expert system, Expert-Inquiry, that was used by doctoral students in education to select the appropriate statistical method for a given research design. This study addressed the following four questions: (1) Can a prototype expert system be designed to assist doctoral students in education in selecting the appropriate statistical method? (2) Can the expert system prototype be validated by the faculty who are subject matter experts in methodology against criteria of dependability, efficiency, effectiveness, reliability, portability, value, and user-friendliness? (3) Can the expert system prototype be validated by doctoral students in education against criteria of dependability, efficiency, effectiveness, reliability, portability, value, and user-friendliness? (4) Should expert system technology be considered as a decision-support tool to assist doctoral students in education in learning how to select the appropriate statistical method for a given research design?

Using problem scenarios, participants used Expert-Inquiry to determine the utility of the system in terms of: (1) dependability and effectiveness, (2) efficiency and speed, (3) reliability, (4) user-friendliness, (5) value and (6) portability and versatility. A panel of subject-matter experts reviewed the problem scenarios and evaluation instruments.

Fifty doctoral students participated in this study. Information packets were distributed to volunteers. Participants received an envelope containing directions on how to use the expert system, locations of the computers that contained Expert-Inquiry, scenarios that were to be used to evaluate the expert system, and a questionnaire for feedback. Participants were grouped based on the number of research courses they had completed. This resulted in two groups: Group 1: composed of students with 0 to 3 hours of research courses and Group 2: students with 6 or more hours of research courses. A nonparametric analysis (Kruskal-Wallis) of evaluations scores and accuracy of solutions to the problem scenarios found there were no statistically significant differences between the two groups in their perception of Expert-Inquiry. Additionally, the participants rated the dependability and effectiveness between uncertain and agreed. However, they agreed that Expert-Inquiry met the criteria of efficiency and speed, reliability, user-friendliness, value, and portability and versatility. Using the decision tree embedded in the system, the researcher was able to determine where the student's logic failed.

Results of this study indicate that Expert-Inquiry can be beneficial to both the students and faculty. Students can utilize the system and master concepts at their own pace; they can use the HOW and WHY features to better understand recommendations. They can also use the decision tree to see alternative options and solutions. The faculty can use the system to help them pin-point problem areas and create customized scenarios that will help the students master those concepts. (Abstract shortened by UMI.)

Citation

Hence, B.M. Design and evaluation of a knowledge-based expert system as a decision support tool in selecting statistical methods. Ph.D. thesis, University of Houston. Retrieved November 28, 2020 from .

This record was imported from ProQuest on October 23, 2013. [Original Record]

Citation reproduced with permission of ProQuest LLC.

For copies of dissertations and theses: (800) 521-0600/(734) 761-4700 or https://dissexpress.umi.com

Keywords