摘要
精确预测剪接位点是真核基因系统研究的第一步。为了取得更加精确的预测结果,本文采用了一个新的标识序列识别方法HM-SVM对剪接位点进行识别。依据剪接位点附近存在的序列保守性,将联合核函数学习融入最大边缘分类器,结合HM-SVM工作集最优化算法,构建并生成了健壮分类器。实验结果表明,该方法在对于剪接位点的识别中,较目前常用的机器学习方法,获得了更高识别率。
Accurate prediction for splice sites is the primary step in the system research for eukaryotic genes. A novel discriminative learning technique for label sequences, named by Hidden Markov Support Vector Machines, is adopted for better prediction performance of splice sites. According to the conservation feature in the vicinity of splice sites, the joint kernel learning is syncretized into the maximum margin classifier, and combining the HM-SVM training set optimization algorithm,a robust classifier is designed and generated here. The experimental results show that the HM-SVM approach acquires higher rate in the splice site identification than popular machine learning techniques.
出处
《微计算机信息》
北大核心
2006年第12S期240-242,共3页
Control & Automation
关键词
隐马尔可夫支持向量机
剪接位点
识别
Hidden Markov Support Vector Machines
Splice Site
Identification
