深度森林在乳腺癌检测中的应用

Application of Deep Forest in Breast Cancer Detection

对于医疗诊断领域传统机器学习分类算法效果不理想的情况,引入深度森林算法,应用于乳腺癌肿瘤的分类问题。该算法使用随机抽样方式对乳腺癌原始特征进行变换增强其特征表征能力,通过级联随机森林对增强特征做逐层表征学习,最后经过分类器输出分类结果。实验结果表明深度森林较支持向量机和决策树算法在乳腺癌良恶性分类性能更好,最高分类精度可达96.2%。该应用在提高了医师在乳腺癌诊断中的准确率。

For the situation that the traditional machine learning classification algorithm in the field of medical diagnosis is not ideal,this paper introduces the deep forest algorithm and applies it to the classification of breast cancer.In this algorithm,the original features of breast cancer are transformed by random sampling method to enhance the feature representation ability.The enhanced features are represented and learned layer by layer through cascaded random forest.Finally,the classification results are output by the classifier.The experimental results show that the performance of deep forest in breast cancer classification is better than SVM and decision tree algorithm,and the highest classification accuracy can reach 96.2%.This application improves the accuracy of doctors in the diagnosis of breast cancer.