目的 基于Preferential Attachment-VIKOR算法对中药配伍关系进行挖掘，实现对阿尔茨海默病（Alzheimer’s disease，AD）中药处方与证候对应关系的预测与处方综合评价。方法 通过检索中国知网、维普、万方数据库，筛选2005年8月—2025年8月所有中药治疗AD非重复的经验选方，整理中药处方的证候、组成，导入Neo4j图数据库，加载Graph Data Science插件，使用Preferential Attachment算法结合VIKOR算法进行网络挖掘，分析证候与处方的关系并进行处方综合评价，并基于经典名方对构建的算法准确性进行验证。结果 筛选AD相中医证候51个，中药处方113个，中药171味，形成336个节点，1 297条边的“疾病-证候-处方-中药”网络，网络中处方节点与证候节点间的Preferential Attachment评分即证候评分可以反映证候-处方的相关性，实现“方-证关系”的有效预测，其结果受处方药味数与处方组成影响，借由证候评分、药味数获得的配伍评分可以有效反映处方的配伍合理性。基于VIKOR算法综合证候评分、配伍评分、药味数对处方进行的评价，可以从证候与配伍2个维度实现对处方在具体证候下的综合评价。结论 基于Neo4j图数据库建立了AD的“疾病-证候-处方-中药”网络，并基于Preferential Attachment-VIKOR算法实现对中药处方与证候对应关系的预测与处方证候与配伍的综合评价。

Objective To mine traditional Chinese medicine (TCM) compatibility patterns and predict the “formula-syndrome” relationships in herbal prescriptions for Alzheimer’s disease (AD), as well as to conduct a comprehensive evaluation of prescriptions, based on a hybrid Preferential Attachment-VIKOR algorithm. Methods Non-repetitive empirical prescriptions for AD published between August 2005 and August 2025 were retrieved from three major Chinese databases: CNKI, VIP, and Wanfang. Data on prescription compositions and associated syndromes were extracted and imported into the Neo4j graph database. The Graph Data Science plugin was loaded. A hybrid approach combining the Preferential Attachment algorithm with the VIKOR method was applied for network mining to analyze syndrome-prescription associations and perform comprehensive prescription evaluation. The accuracy of the model was validated using classical renowned formulas. Results A total of 51 AD-related TCM syndromes, 113 prescriptions, and 171 herbs were included, forming a “disease-syndrome-prescription-TCM” network with 336 nodes and 1 297 edges. The Preferential Attachment score between prescription and syndrome nodes—termed the syndrome score—effectively reflected the relevance of “formula-syndrome” relatioships, enabling accurate prediction. This score was influenced by the number of herbs and composition of the prescription. The compatibility score obtained through syndrome scoring and the number of herbs served as an indicator of rationality in herbal pairing. Using the VIKOR algorithm, an integrated evaluation index was developed based on syndrome score, compatibility score, and herb count, enabling a multidimensional assessment of prescriptions under specific syndromes from both syndrome relevance and compatibility rationality perspectives. Conclusion This study established a “disease-syndrome-prescription-TCM” network for AD in a Neo4j graph database and successfully applied a Preferential Attachment-VIKOR hybrid algorithm to predict formula-syndrome relationships and enable comprehensive evaluation of TCM prescriptions, integrating both syndrome correspondence and compatibility quality.

R285;TP18

海河现代中药实验室科技项目(22HHZYSS00015)