《中醫證候學》是蒐集及整理古今中醫臨床診斷文獻的巨著。它先依病因為綱將疾病分為15個證門,又依病因的不同性質將15個證門再細分為45個證類。接著,它依病位為目將45個證類再分為281證型。最後,它再依病機為科將281個證型分為2344個證候。《中醫證候學》的辨證體系非常龐雜,一般醫生難以掌握,我們希望運用計算機龐大的記憶能力及迅速的分析能力,研製一個中醫自動辨證系統,提供研究中醫辨證體系的平台。 本論文根據《中醫證候學》虛證門的症狀資訊研製一個中醫自動辨證系統,分成臟腑辨證、證門辨證、證類辨證、證型辨證及證候辨證五個部分,逐步判斷疾病的病因、病位及病機。本論文以兩種不同的方式進行辨證:平權辨證及加權辨證。平權辨證將每個症狀之權重視為相同,亦即假設證候中每個症狀之發生機率相同。實驗顯示平權辨證在許多情況的辨識率偏低。加權辨證依症狀之發生機率給予不同的權重。實驗顯示加權辨證在大多數的情況都可以提高辨識率,改善平權辨證在許多情況辨識率偏低的問題。
The book Syndromatology of Traditional Chinese Medicine is a monumental work that collects and organizes ancient and modern literatures on the theory and practice of clinical diagnosis in traditional Chinese medicine. This book categorizes diseases into 15 syndrome divisions according to the causes of diseases. Based on the properties of the causes of diseases, these 15 syndrome divisions are further divided into 45 syndrome subdivisions. These 45 syndrome subdivisions are then divided into 281 syndrome classes according to locations of diseases. Finally, these 281 syndrome classes are divided into 2344 syndromes according to mechanisms of diseases. Since the system of syndrome differentiation is huge and complex, it is very difficult for doctors to master it. We wish to utilize the huge memorizing capability and the speedy analyzing capability of computers to develop an automatic syndrome differentiation system to provide a platform for the research of the system of syndrome differentiation. This thesis develops an automatic deficiency syndrome differentiation system according to the symptom information in the system of deficiency syndrome differentiation in Syndromatology of Traditional Chinese Medicine. This system consists of five parts: visceral syndrome differentiation, syndrome division differentiation, syndrome subdivision differentiation, syndrome class differentiation, and syndrome differentiation to gradually differentiate the causes, locations, and mechanisms of diseases. This thesis uses two different syndrome differentiation approaches: plain syndrome differentiation and weighted syndrome differentiation. In plain syndrome differentiation, every symptom in syndrome has the same weight. This implies the probability of occurrence of every symptom is the same. Experiments shows that the syndrome differentiation rate in plain syndrome differentiation is very low in many situations. In weighted syndrome differentiation, every symptom in syndrome has a (maybe different) weight that depends on the probability of occurrence of this symptom. Experiments shows that the syndrome differentiation rate in weighted syndrome differentiation is raised in most situations and alleviates the problem of the low syndrome differentiation rate in plain syndrome differentiation.