目的:以新型冠状病毒肺炎（简称新冠肺炎）湿热蕴肺证为例，探讨证候的症状间关系及其对证候诊断的贡献度，为建立证候诊断依据提供方法学支持。方法:基于654份新冠肺炎患者的临床调查数据，以湿热蕴肺证为例，采用SPSS Modeler 14.1软件，结合关联规则与贝叶斯网络数据挖掘技术，探讨症状间关系并明确症状（群）对证候诊断的贡献度。结果:654份新冠肺炎临床资料中涉及湿热蕴肺证患者121例，其中出现频率>40%的症状有发热（53.72%）、咳嗽（47.93%）、舌质红（45.45%）、脉数（43.80%）、苔腻（42.15%）、苔黄（41.32%）、乏力（40.50%）和纳呆（40.50%）。关联规则分析显示，二项关联关系较强的症状群包括发热、口渴，胸闷、气促，咳嗽、痰黄等；三项关联关系较强的症状群包括咳嗽、痰黄、痰黏稠，纳呆、呕恶、头身困重，发热、口渴、乏力等。以湿热蕴肺证（是＝1，否＝0）为目标变量，以出现频率>15%的症状为输入变量，建立贝叶斯网络模型，得出湿热蕴肺证症状（群）概率分布表，其中发热的父节点（每个输入变量的上级节点）只有1个（湿热蕴肺证），条件概率是0.54；咳嗽的父节点有痰黄、湿热蕴肺证，表示在湿热蕴肺证中咳嗽与痰黄存在直接因果关系，且在有痰黄的条件下，咳嗽的条件概率为0.99。常见症状（群）及其对湿热蕴肺证诊断的贡献度为：发热、口渴（0.47），咳嗽、痰黄（0.49），胸闷、气促（0.46），纳呆、头身困重（0.61），苔黄腻、脉滑数（0.95）。结论:关联规则结合贝叶斯网络在阐释症状间关系及其对证候诊断的贡献度具有一定的可行性和客观性，为建立证候诊断依据提供了方法学支持。“,”Objective:To explore the correlation between symptoms and their contribution to syndrome based on syndrome of lung damp-heat accumulation in coronavirus disease 2019 (COVID-19), thus to provide methodological basis for the syndrome diagnosis.Methods:Based on 654 clinical investigation questionnaires data of COVID-19 patients, a model based on syndrome of lung damp-heat accumulation was set. Using SPSS Modeler 14.1 software, association rules and Bayesian network were applied to explore the correlation between symptoms and their contribution to syndrome.Results:There were 121 questionnaires referring to syndrome of lung damp-heat accumulation in total 654 questionnaires. The symptoms with frequency > 40% were fever (53.72%), cough (47.93%), red tongue (45.45%), rapid pulse (43.80%), greasy fur (42.15%), yellow tongue (41.32%), fatigue (40.50%) and anorexia (40.50%). Association rule analysis showed that the symptom groups with strong binomial correlation included fever, thirst, chest tightness, shortness of breath, cough, yellow phlegm, etc. The symptom groups with strong trinomial correlation included cough, yellow phlegm, phlegm sticky, anorexia, vomiting, heavy head and body, fever, thirst, fatigue, etc. Based on SPSS Modeler 14.1 software, with syndrome of lung damp-heat accumulation (yes = 1, no = 0) as target variable, and the selected symptoms with frequency > 15.0% as input variables, the Bayesian network model was established to obtain the probability distribution table of symptoms (groups), in which there was only one parent node (the upper node of each input variable) of fever, and the conditional probability was 0.54. The parent node of cough had yellow phlegm and syndrome of lung damp-heat accumulation, indicating that there was a direct causal relationship between cough and yellow phlegm in syndrome of lung damp-heat accumulation, and the conditional probability of cough was 0.99 under the condition of yellow phlegm. The common symptom groups and their contribution to syndrome were as follows: fever and thirsty (0.47), cough and yellow phlegm (0.49), chest tightness and polypnea (0.46), anorexia and heavy cumbersome head and body (0.61), yellow greasy fur and slippery rapid pulse (0.95).Conclusions:It is feasible and objective to analyze the correlation between symptoms and their contribution to syndromes by association rules combined with Bayesian network. It could provide methodological basis for the syndrome diagnosis.