面向乳腺肿瘤的诊前问答系统决策模型构建研究
ecision-making model construction of a pre-clinical Q&A system for breast tumors
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C4.5经典分类算法和悲观剪枝法对调研收集的177条病例数据，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C4.5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。以乳腺肿瘤单一病种展开研究，基于乳腺科门诊录音构建的C4.5决策树，将医生临床医学实践与机器学习算法结合对未就医患者进行诊前风险预测效果较好。
[Purpose/Significance] To predict the disease risk of potential or existing breast tumor patients using a decision tree classification model to simulate the expert consultation idea. [Methodology/Procedure] A C4.5 classical classification algorithm and a pessimistic pruning method were used to predict the outcome of patient pre-consultation for 177 case data collected from the study. [Results/Conclusions] A C4.5 decision tree with 76 leaf nodes and "whether postoperative chemotherapy or radiotherapy ends in the hospital" as the root node was generated with 95% prediction accuracy and classified into 3 risk levels according to the classification labels.The C4.5 decision tree was constructed based on a single breast tumor disease, and it is a combination of clinical practice and machine learning algorithm to predict the risk of unattended patients.
郑群、曹旭晨、李一凡、王世文
肿瘤学医学研究方法计算技术、计算机技术
乳腺肿瘤4.5算法决策树模型构建
Breast tumor4.5 algorithmdecision treemodel construction
郑群,曹旭晨,李一凡,王世文.面向乳腺肿瘤的诊前问答系统决策模型构建研究[EB/OL].(2023-03-23)[2025-08-16].https://www.biomedrxiv.org.cn/article/doi/bmr.202303.00029.点此复制
评论