COMPARISON OF RULE-BASED AND NEURAL NETWORK SOLUTIONS FOR A STRUCTURED SELECTION PROBLEM
1993; SAGE Publishing; Issue: 1399 Linguagem: Inglês
ISSN
2169-4052
Autores Tópico(s)Infrastructure Maintenance and Monitoring
ResumoAdvantages and disadvantages are compared of using a rule-based paradigm versus a neural-network-based paradigm for developing expert systems involving structured selection problems. For comparison purposes, two knowledge-based expert systems were developed using the two alternative paradigms to solve the same specific problem: selection of pavement sections that would benefit most from the routing and sealing maintenance treatment. Each expert system used commercially available microcomputer software costing less than $1,000. The two programs have been compared in terms of the results achieved, software and hardware requirements, system development and programming effort, knowledge processing, how uncertainty is dealt with, and other parameters. Neural networks provide an efficient and appropriate computational tool for solving structured selection problems. They can be implemented faster and updated more easily than rule-based systems. However, neural networks do not encode knowledge in any useful form whether used for future reference, explanation of reasoning, or knowledge-based updating.
Referência(s)