改善贝叶斯公式判别效果的探讨——计算机辅助预测上消化道再出血

A STUDY TO IMPROVE DISCRIMINATORY ABILITY OF INDEPENDENT BAYES-THEOREM COMPUTER-AIDED PROGNOSIS OF UPPER GASTROINTESTINAL BLEEDING

本文基于条件独立的贝叶斯公式,计算机辅助预测上消化道再出血。同时对症状变量间的交互作用给予适当的考虑,以探讨不同的方法改善贝叶斯公式在计算机辅助诊断中的判别效果。结果表明,结合聚类分析的条件独立的贝叶斯公式和最佳依赖树的贝叶斯公式,其判别符合率分别比条件独立的贝叶斯公式提高5％(P>0.2)和10％(P<0.05)左右。证明利用最佳依赖树的贝叶斯公式对改善贝叶斯公式的判别效果具有一定的成效。

Independent Bayes-Theorem has been used as computer-aided prognosis of upper gastrointestinalbleeding. For improving disriminatory ability of independent Bayes-Theorem in computer-aided dia-gnosis. the interaction of symptoms-variables have been appropriately taken into account. The result shows that the strategy is successful. The accuracy of discriminant in independent Bayes-Theorem combined with cluster analysis and Bayes with optimum tree dependence increased by 5%(P>0.2) and 10% (P<0.05) respectively compared with that in independent Bayes-Theorem.