Abstract:

The expert system of disease diagnosis is subjected to the bad efficiency, the low accuracy and the lack of contrast, by its only one time use of the field knowledge in a reasoning process, which is built on the base of classical model. Making example of goat, this paper designs the architecture of diagnosis system, and introduces the ideology of multi-mode composite reasoning scheme, constructs both Bayesian reasoning supporting learning by self which is based on probability. The method of measuring semantic is based on pattern recognition, and with different theoretical background. Experimental results show that the composite reasoning scheme enables to improve the utilization rate of knowledge, its accuracy of diagnosis reaches 85%, increasing the contrast, and achieves with an accepted macro effect of diagnosis.

Key words: composite reasoning Bayesian reasoning learning by self, semantic distance, disease diagnosis, pattern recognition

摘要：

基于经典专家系统模型建立的疾病诊断系统仅单次性地推理和知识运用，知识资源利用效率低、准确度低、对比度缺失。为此，研究以确诊山羊疾病为例，引入多模式组合推理确诊机制，设计诊断系统体系架构，构造源于概率的自学习贝叶斯推理和基于模式识别的语义距离测度的不同理论背景的多模式组合推理诊断算法。实验结果表明，组合推理诊断模型提高知识库利用率，增加了对比度，准确率达到85%，取得较好的综合诊断效果。

关键词: 组合推理, 自学习式贝叶斯推理, 语义距, 疾病诊断, 模式识别

CLC Number: 

• TP183