面向乳腺肿瘤的诊前问答系统决策模型构建研究

Study on the Construction of a Decision-making Model of the Pre-clinical Q&A System for Breast Tumor

目的/意义运用决策树分类模型模拟专家问诊思路,预测潜在或已有乳腺肿瘤患者的疾病风险。方法/过程采用C 4.5经典分类算法和悲观剪枝法,对调研收集的病例数据进行患者预问诊的结果预测。结果/结论生成一棵以“术后化疗or放疗在院是否结束”为根节点、拥有76个叶子节点的C 4.5决策树,预测准确率达95%,并根据分类标签划分为3个风险等级。

Purpose/Significance To predict the disease risk of potential or existing breast tumor patients by using a decision tree classification model to simulate the expert consultation idea.Method/Process A C 4.5 classical classification algorithm and a pessimistic pruning method are used to predict the outcome of patient pre-consultation for case data collected from the study.Result/Conclusion A C 4.5 decision tree with 76 leaf nodes and“whether postoperative radiotherapy or chemotherapy ends in the hospital”as the root node is generated with 95%prediction accuracy and classified into 3 risk levels according to the classification labels.