Author: Lee Seng-whye
Title: Verification and validation of a life underwriting knowledge-base
Degree: M.Sc.
Year: 1996
Subject: Expert systems (Computer science) -- Verification
Expert systems (Computer science) -- Validation
Hong Kong Polytechnic University -- Dissertations
Department: Department of Computing
Pages: 131 leaves ; 30 cm
Language: English
Abstract: The increasing application of artificial intelligence principles to realistic problems has created a demand for efficient construction and maintenance of Knowledge Based System (KBS) . Systematic and continual verification and validation of these systems are expected to ease the development and foster user acceptance. Verification refers to getting the KBS right, and validation means getting the right KBS. New methods of verifying and validating of KBSs are necessary to ensure its reliability. However, at the moment there is no single method of verification and validation has evolved that has gained general acceptance. The integration of several methods may help elevate the problem. Without proper verification and validation procedures, current commercial knowledge base that underwrites insurance policies can cause dramatic adverse impacts on human lives and corporate bottoms lines. In a conventional system maintaining task, it is usually a costly mistake to defer the activity of detecting and correcting errors until late in the development stage. Also, by not identifying specification errors at the initial stage worsens the productivity of the development process and the quality of system developed. Moreover, the fixing of one error may have the additional effect of introducing more errors into the system. KBSs are not immune from this danger. An increasingly competitive environment led the life underwriting department of ABC Life insurance Co., Ltd to consider introducing automation into the underwriting process. An Underwriting and Issue Work-Station Project was set up to solve the problem. The project started with a specification study, during which a complete ABC Life Insurance KBS Life Underwriting Specifications manual was produced. At this stage no one is sure that the Specifications are error-free. The purpose of this dissertation is directed at achieving comprehensive verification and validation of the ABC's Life Underwriting KBS. A review of verification literature has shown that there are a number errors and anomalies that can occur in a KBS. Generally, they can be classified as redundancy, subsumption, conflicting rules or circular rules. To detect the errors and anomalies in the KBS we can adopt many of the methods or tools used for verification of KBS. As it is impossible to consider all methods and tools, we reviewed some of them that are appropriate for verifying the ABC's Life underwriting KBS. The methods reviewed are the Normal Form Approach, the Decision Table Method and the Incidence Matrix Method. An example of each method is given to illustrate the verification technique. The three methods are then evaluated based on the representation schemes, ease of automation and simplicity of its algorithms. It is found that the Incidence Matrix Method satisfied all three criteria and thus it is selected to verify the ABC'S Life Underwriting KBS. Next, we look at how the Incidence Matrix Method can be automated. Borland C++ is chosen as the programming tool to implement the automation. Two programs A and B (see appendix 1) are written. Program A checks rules for redundancy and Program B checks rules for complex subsumption. As the rules in the ABC Life Insurance KBS Life Underwriting Specifications Manual were written in decision table form, we first have to translate them into production rules and then into the matrix form. The matrices are fed into the programs. Both Programs produce a commentary as output. The results of the verification of the Life Underwriting KBS are then provided. The second part of the dissertation deals with the validation of the Life Underwriting KBS. Again a literature review of the different types of validation techniques is carried out. The techniques reviewed are the Informal Validation, Validation by Testing, Field Tests, Subsystem Testing and Sensitivity Analysis. The technique "Validation by Testing" is chosen because it is the most comprehensive and the most practical of all the techniques available. To validate the ABC's Life Underwriting KBS, a test plan is written. The test plan includes the testing mission, purposes and test areas. Test cases are then generated to test three areas. They are Data Elements Validation, General Underwriting and Impairment Underwriting. The results of the testing are then compared to that of the underwriters' suggestions. Issues such as problems encountered and limitations are raised in the concluding part of the dissertation. The notions of certainty factors and the incremental verification of KBS are introduced. Finally, the dissertation recommends further development or research in verification of KBS to include certainty factors, incremental verification to match continual knowledge maintenance and verification problem to cover other forms of knowledge representations such as frames, semantic nets etc.
Rights: All rights reserved
Access: restricted access

Files in This Item:
File Description SizeFormat 
b12350394.pdfFor All Users (off-campus access for PolyU Staff & Students only)3.41 MBAdobe PDFView/Open

Copyright Undertaking

As a bona fide Library user, I declare that:

  1. I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
  2. I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
  3. I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.

By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.

Show full item record

Please use this identifier to cite or link to this item: