【摘要】 生物信息学领域的微阵列分析、质谱数据分析等大规模机器学习问题的不断涌现,对已有的特征选择算法提出了严峻的挑战,迫切需要研究适应于高维小样本特征的、准确性和运行效率等综合性能较好的、新的特征选择算法。将基于量子粒子群算法(QPSO)与SVM结合,并将建立的诊断模型用于生物标记物的选择。实验结果表明,新的基于量子粒子群算法建立的模型不仅具有良好的预测精度,而且在速度上有大幅的提高。更多还原

【Abstract】 Continuous emergence of large-scale machine learning problems in bioinformatics sector covering from microarray analysis to mass spectral data analyses,etc.,poses serious challenge on existing feature selection(FS) techniques,thus there is an urgent need in the study of a novel FS algorithm which adapts to the feather of high dimension and small sizes sample and with better comprehensive properties in both accuracy and operation efficiency.In this paper we combined QPSO with SVM and applied the diagnosis model built to the selection of biomarkers.Experiments showed that the new model built on the basis of QPSO achieved high prediction accuracy and raised the speed conspicuously as well.更多还原