基于Stacking集成学习的乳腺癌智能诊断预测模型

Intelligent Diagnosis and Prediction Model of Breast Cancer Based on Stacking Ensembled Learning

摘要:

整合创新数据预处理方法和机器学习算法，根据乳腺癌威斯康星州诊断数据集构建了智能预测模型。首先，采用以LightGBM（Light Gradient Boosting Machine）为基模型的特征递归消除法进行特征选取；其次，使用结合ADASYN（adaptive synthetic sampling）过采样和OSS（one-sided selection）欠采样的综合采样进行数据不平衡处理，得到均衡的训练数据集；最后，以MLP（multilayer perception）、LightGBM、CatBoost（categorical boosting）作为基学习器，逻辑回归模型作为元学习器构建了基于Stacking集成学习的智能诊断模型，并通过5折交叉验证和准确率、敏感度、受试者操作特征曲线的下方面积等多项分类预测指标进行评估。实验结果显示所提出的模型能够达到98.2%的预测准确率，具备稳定且优秀的分类预测性能，能够为乳腺癌的临床诊断提供强有力的决策支持。

Abstract:

Integrating innovative data preprocessing methods and machine learning algorithms， an intelligent prediction model is constructed based on the Breast cancer Wisconsin diagnostic dataset. Firstly， the feature recursive elimination method based on light gradient boosting machine （LightGBM） model is used for feature selection. Secondly， the integrated sampling combined with adaptive synthetic sampling （ADASYN） oversampling and one-sided selection （OSS） undersampling is used to deal with data imbalance， and a balanced training data set is obtained. Finally， with multilayer perception （MLP）， LightGBM and categorical boosting （CatBoost） as the base learner and logistic regression model as the meta-learner， an intelligent diagnosis model based on Stacking ensembled learning is constructed. It is evaluated by 5 folds cross-validation and classification prediction indicators such as accuracy， sensitivity， and area under receiver operating characteristic curve. The experimental results show that the proposed model can achieve a prediction accuracy of 98.2%， and has stable and excellent classification prediction performance， which can provide strong decision support for clinical diagnosis of breast cancer.