从大数据角度建立中药化合与症状之间的关系具有良好的应用前景，然而这种药物-症状之间的关联研究目前相当缺乏。在这里，1441个美国FDA批准的药物被收集，并用主成分分析法提取122个描述符，解释了91%的变异。然后，k-均值+把药物数据集划分成3类，三种对应的SVDD模型（药物相似的筛选模型）使其高达95.6%的准确性。此外，6878种中药分子通过以上3个SVDD模型得到5309个候选药物分子，准确性达到77.19%。为了评估SVDD模型的准确性，8559中药分子症状挖掘自pubmed摘要，涉及697个中药分子和314个症状。697中药分子中的大多数在SVDD数据中发现（5309分子），显示筛选SVDD候选药物的潜力。此外，建立了中药分子-中药分子网络和分子-症状网络。以SVDD为核心的人工智能方法能够有效的分析中药新药的数据。

From the perspective of big data to establish the relationship between traditional Chinese medicine combined with symptoms has a good application prospect. However, studies of the association between the drug - symptoms at present is lack of. Here, 1441 American FDA approved drugs were collected, and 122 descriptors were extracted using principal component analysis, explaining 91% of the variance. Then, the k- mean and the drug data sets are divided into 3 categories, three corresponding SVDD models (drug similar screening model) to make it as high as 95.6% of the accuracy. In addition, 6878 kinds of traditional Chinese medicine molecules through the above 3 SVDD model to get 5309 candidate drug molecules, the accuracy of 77.19%. In order to evaluate the accuracy of the SVDD model, 8559 Chinese medicine molecular symptoms were excavated from the pubmed summary, involving 697 Chinese medicine molecules and 314 symptoms. 697 most of the traditional Chinese medicine molecules found in the SVDD data (5309 molecules), showing the potential of screening SVDD drug candidates. In addition, the establishment of traditional Chinese medicine molecules - molecular network of Chinese medicine and molecular - symptom network. SVDD as the core of the artificial intelligence method can effectively analyze the data of traditional Chinese medicine.