Induction and Uncertainty Management Techniques Applied to Veterinary Medical Diagnosis

TL;DR

This paper explores induction and Bayesian uncertainty techniques to improve veterinary medical diagnosis by extracting knowledge from hospital data and combining evidence more effectively.

Contribution

It introduces a Bayesian evidence combination method using fuzzy events and demonstrates its advantages over traditional statistical approaches.

Findings

More significant variables identified than with classical methods

Improved diagnostic accuracy through induction techniques

Effective evidence combination using Bayesian and fuzzy logic

Abstract

This paper discusses a project undertaken between the Departments of Computing Science, Statistics, and the College of Veterinary Medicine to design a medical diagnostic system. On-line medical data has been collected in the hospital database system for several years. A number of induction methods are being used to extract knowledge from the data in an attempt to improve upon simple diagnostic charts used by the clinicians. They also enhance the results of classical statistical methods - finding many more significant variables. The second part of the paper describes an essentially Bayesian method of evidence combination using fuzzy events at an initial step. Results are presented and comparisons are made with other methods.