目的：运用人工神经网络算法探讨“方”与“证”之间的非线性映射关系。方法：整理《伤寒论》中有症有方的条文，利用人工神经网络对245 条数据进行训练，建立方证知识库，先选择100 条作为测试样本，然后分别以桂枝汤证、麻黄汤证、小柴胡汤证、白虎汤证的症状表现输入模型，进行“症状”到“药物”的预测分析。结果：模型预测正确率79%，对桂枝汤证及麻黄汤证各药物的预测误差均在0.1 以内；对小柴胡汤、白虎汤证中主药的预测误差在0.1 以内，对其它药物的预测误差在0.3 以内。结论：利用人工神经网络能较好的模拟方证之间的非线性映射，该思路可进一步应用于方证辨证论治体系中，为方证规范化及中医信息化服务。

This study was aimed to explore the nonlinear relationship between syndromes and prescriptions by using the artificial neural network (ANN) algorithm. The clauses composed of syndromes and prescriptions in Shanghan Lun (Treatise on Cold Damage and Miscellaneous Diseases) were sorted out. A total of 245 clauses were trained with ANN to establish the knowledge base model of syndromes and prescriptions. After that, 100 clauses were tested. And then,the symptoms of Gui-Zhi-Tang (GZT) syndrome, Ma-Huang-Tang (MHT) syndrome, Xiao-Chai-Hu-Tang (XCHT)syndrome and Bai-Hu-Tang (BHT) syndrome were input into the model to predict the Chinese herbs. The results showed that the test accuracy of the model was 79%. For the herbs to GZT and MHT syndrome, the test error was less than 0.1. For the main herbs to XCHT and BHT syndrome, the test error was less than 0.1. While for other herbs, the test error was less than 0.3. It was concluded that the ANN algorithm can simulate the nonlinear relationship between syndromes and prescriptions, which can be applied to study the syndrome differentiation, in order to contribute to the syndromes and prescriptions standardization and the traditional Chinese medicine (TCM) information.

江苏省自然科学基金委面上项目（BK2009429）：心系基础证的研究，负责人，吴承玉；江苏省教育厅研究生培养创新工程项目（CXZZ13_0607）：基于软计算方法的心系基础证诊断知识库的构建及应用，负责人，杨涛。