中醫治療上所使用的中藥材數以萬計。而這些中藥材進一步組成更多的方劑,方劑的組成及用途並沒有一定的科學量化依據。因此,本論文提出一種中草藥作用的自動化推論方法,以建立中藥方劑作用的科學根據。 自適應共振理論(Adaptive Resonance Theory; ART)是一種可以自動訓練的類神經網路。其特點是簡單,只要設定輸入及輸出資料的維度,便可以進行資料的分類。我們根據中藥的藥物歸經理論,將已知功能的中藥依據其四性、五味及歸經等屬性組成中藥的特徵向量。以這些特徵向量訓練一個ART類神經網路系統,來推論未知功能的其他中藥藥物。
Traditional Chinese medicine includes thousands of herbs as remedies for sick people. These herbs make up more complex prescriptions further. In the past, the usage of herb prescriptions was decided by doctors’ experience. There was not any qualitative method for the purpose of understanding the application of herb prescriptions. Therefore, we propose a method to infer that function of herb, and build up scientific rules of functions for traditional Chinese medicine prescriptions. Adaptive Resonance Theory (ART) is a model of neural network. This network just use a data set composed of input vector and output vector. It’s simpleand stable. We use it to classify herb prescriptions according to the theory of channel entry. The feature vector of Chinese medicine consists of the attributes including nature, flavour and channel entries. The ART model is trained with known feature vector. Then it is used to infer the application of unknown medicinal herbs.