Use this URL to cite or link to this record in EThOS:
Title: Combining case-based reasoning with neural networks in diagnostic systems
Author: Reategui, Eliseo Berni
ISNI:       0000 0001 3510 5441
Awarding Body: University of London
Current Institution: University College London (University of London)
Date of Award: 1997
Availability of Full Text:
Access from EThOS:
Full text unavailable from EThOS. Please try the link below.
Access from Institution:
This thesis presents a new approach for integrating Case-Based Reasoning (CBR) with a Neural Network (NN) in diagnostic systems. In the hybrid NN-CBR approach, the neural network makes hypotheses and provides remindings that are used in guiding the search for similar experiences in a library of previous cases. CBR is responsible for the selection of a most similar match for a given problem, as to support a particular hypothesis made by the neural network, or to decide among hypotheses. Items called diagnosis descriptors have been created in order to represent in an intelligible way the knowledge stored in the neural network. These descriptors are used for consultation purposes, for confirming or refuting a final result, and for building explanations. The NN-CBR architecture has been used in the development of a system for the diagnosis of Congenital Heart Diseases (CHD). The system has been evaluated using two cardiological databases with a total of 214 CHD cases. The cases of the first database were collected with the supervision of an expert in CHD, while the cases of the second database were collected without expert supervision. The hybrid system has shown a good performance when diagnosing cases of the first database. Additionally, its diagnosis descriptors created to interpret the knowledge of the neural network have been found to be similar to the knowledge elicited from experts for the same diagnostic problems. A drop in performance could be observed when the system was trained to diagnose cases of the second database, whose cases were collected without the supervision of an expert. Three other well-known databases have been used to evaluate the NN-CBR approach further. The hybrid system manages to solve problems that cannot be solved by the neural network on its own. Additionally, its indexing mechanism based on remindings provided by the neural network introduces a significant reduction in the number of case comparisons, if contrasted with a standard nearest-neighbour procedure. By using these alternative knowledge representation and reasoning schemes, the hybrid NN-CBR approach suggests some solutions for common CBR problems (e.g. indexing and retrieval), as well as for neural network problems (e.g. the interpretation of the knowledge stored in a neural network and explanation of reasoning).
Supervisor: Not available Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: Bionics