原始数据集

In this study, machine learning algorithms94 was used instead of traditional statistical analysis methods to process the clinical95 characteristics data of patients with IDC, and a machine learning-based predictive96 model was developed to discover the relationship between certain independent97 variables and the likelihood of patients with IDC developing distant organ metastasis.98 Focusing on the predictive models of four algorithms: Logistic Regression, SVM,99 Random Forest and XGBoost, and using them as base models for integration to100 construct hybrid models. A ten-fold cross-validation was used to compare the101 accuracy, recall, precision, and AUC value of each model to predict the risk of cancer102 metastasis in patients with IDC, thereby assisting clinicians in making more rational103 clinical decisions and enabling patients to receive treatment earlier.

本研究采用机器学习算法替代传统统计分析方法，对浸润性导管癌（Invasive Ductal Carcinoma，IDC）患者的临床特征数据进行处理，并构建基于机器学习的预测模型，以探究特定自变量与此类患者发生远处器官转移的风险之间的关联。本研究聚焦逻辑回归（Logistic Regression）、支持向量机（SVM）、随机森林（Random Forest）以及XGBoost四种算法的预测模型，以其作为基模型进行集成，构建混合模型；采用十折交叉验证法对比各模型预测浸润性导管癌患者癌症转移风险的准确率、召回率、精确率以及曲线下面积（Area Under Curve，AUC）值，以此辅助临床医师做出更合理的临床决策，使患者能够更早接受治疗。