基于贪心EM的模体预测算法

摘要

论文主要研究了Motif预测的算法,在贪心EM预测算法基础上进行分析优化,并制定新的预测方法.工作重点是在参数的初始化,对参数模型的重新划分并引入Kd-树的层次聚类的方法,建立新算法.预测结果表明,在预测较大数据集方面PKGE算法有一定的优势,尤其是对同一物种的序列预测具有更强的搜索和分类能力,在没有改变时间复杂度的前提下显著地提高了搜索的效率.%Motif finding algorithm was studied as the key point in this paper. Optimization was based on the EM-baced greedy algorithm and then predicting method was established. Parameters initialization,the re-division of the parameter model and intro?duction of Kd-tree hierarchical clustering method,and the establishment of PKGE algorithm were paid more attention in this paper. The results indicated PKGE algorithm has some advantages in predicting motifs in large data sets,especially in the prediction of se?quences in the same species. With advantages in sequence search and classification capabilities,the search efficiency was improved by PKGE algorithm significantly without changing complexity in the time.