Based on miRNAs expression profiling data sets , new data mining algorithms— tSVM‐kNN (t statistic with support vector machine‐k nearest neighbor ) is proposed . Firstly , an original selection is made to this set by characteristics using t‐statistic method . After that , both ideas in support vector machine(SVM )and k nearest neighbor (kNN )algorithms are combined as a classifier , i.e. , SVM‐kNN algorithm .Finally ,the classification results as outputs can be obtained .Then ,simulation experiments show that SVM‐kNN algorithm as a classifier can display a stronger ability compared with running SVM and kNN , respectively . As to the aspects of quantity and recognition accuracy with a miRNAs label , tSVM‐kNN algorithm only need five miRNAs but can get a precision of 96.08% in classification . Obviously , compared with some existed methods , the proposed algorithm has more advantages .%基于 miRNA 表达谱数据集,提出了一种新的数据挖掘算法——— tSVM‐kNN (t statistic with support vector machine‐k nearest neighbor).该算法的思想为:首先,采用统计量法对该数据集进行特征初选;其次,将融合了支持向量机和K‐最近邻判别法思想的算法———SVM‐kNN算法作为分类器;最后,输出分类结果.仿真实验表明, SVM‐kNN算法分类器的分类能力比单独运行SVM和 kNN都好;在miRNA “标签”的数量和识别精度方面, tSVM‐kNN算法只需要取5个miRNAs即可获得96.08%的分类准确率.与同类的算法相比,其具有明显的优越性.
展开▼