集成优化核极限学习机的冠心病无创性诊断

冠心病的早期无创性诊断一直是医疗诊断领域的研究热点，为了提高冠心病诊断的准确率和诊断效率，提出了一种新颖的局部Fisher判别分析(LFDA)特征提取方法和集成核极限学习机(KELM)相结合的冠心病诊断模型(LFDA-EKELM)。首先使用LFDA方法剔除不相关特征和冗余特征，找出对分类结果贡献度较高的特征子集，产生不同的训练集以训练粒子群优化的KELM分类器PSO-KELM，并基于旋转森林(RF)构建集成分类器，实现冠心病的智能诊断。实验结果表明，与基于ELM、SVM和BPNN方法相比，提出方法有效提高了冠心病诊断准确率，提升了诊断效率，且分类结果高于已有方法和相似方法，是一种有效冠心病诊断模型。

Optimized kernel extreme learning machine based on ensemble method for diagnosis of heart diseases

The early diagnosis of heart diseases is one the most important medical research areas, in order to further improve the accuracy and efficiency of heart diseases, this article proposed a novel model LFDA-EKELM, which was based the combination of Local Fisher Discriminant Analysis (LFDA) for feature extraction and kernel extreme learning machine (KELM) by ensemble method, for heart disease diagnosis. In the proposed method, LFDA firstly used to eliminate the irrelevant features, and selected the discriminate feature sets, and then to generate the diverse training subsets, ensemble classifiers were constructed based on Rotation Forest (RF) which optimized by particle swarm optimization approach by means of these subsets. Experiment on heart diseases dataset, in terms of experimental comparison with ELM, SVM and BPNN, the proposed method not only greatly improves the diagnosis accuracy, but also increase the training efficiency, the classification rate is higher than those of existing methods, it proves the effective and validity of the proposed model. those of existing methods, it proves the effective and validity of the proposed model.