论文部分内容阅读
基分类器之间的差异性和单个基分类器自身的准确性是影响集成系统泛化性能的两个重要因素,针对差异性和准确性难以平衡的问题,提出了一种基于差异性和准确性的加权调和平均(D-A-WHA)度量基因表达数据的选择性集成算法。以核超限学习机作为基分类器,通过D-A-WHA度量调节基分类器之间的差异性和准确性,最后选择一组准确性较高并且与其他基分类器差异性较大的基分类器组合进行集成。通过在UCI基因数据集上进行仿真实验,实验结果表明,与传统的Bagging、Adaboost等集成算法相比,基于D-A-WHA度量的选择性集成算法分类精度和稳定性都有显著的提高,且能有效应用于癌症基因数据的分类中。
The differences between base classifiers and the accuracy of single base classifiers are two important factors that affect the generalization performance of integrated systems. Aiming at the problem that the difference and accuracy are difficult to balance, a new method based on difference and accuracy Selective Weighted Averaging (DA-WHA) Measure Gene Expression Data Selective Integration Algorithm. The kernel overrun learning machine is used as a base classifier to adjust the difference and accuracy of the base classifiers by DA-WHA metric. Finally, a set of base classifiers with high accuracy and large difference with other base classifiers Combination of devices for integration. The simulation results on UCI gene datasets show that compared with the traditional algorithms such as Bagging and Adaboost, the classification accuracy and stability of the selective integration algorithm based on DA-WHA metric are significantly improved, Effectively applied to the classification of cancer gene data.