基于PSO的双向聚类算法研究
发布时间：2018-12-24 13:24
【摘要】：生物信息学是一门结合了生物学、计算机科学、数学和化学等领域知识的交叉学科。随着科技的飞速发展,基因测序技术的研究取得了重大突破,人们逐渐开始对基因的功能和内在机理展开了探索研究。目前,每天都会产生海量的基因信息数据,生命科学的研究重点也从如何获取生物数据转移到了怎样对这些数据进行有效的分析上面。目前,对基因表达数据的分析处理,主要采用的方法是聚类。一般的聚类只能从基因矩阵的行或者列单一方向进行,这种方法只能找到基因表达数据矩阵中的全局信息。而大量有价值的生物学信息往往就隐藏在这些局部信息中,双向聚类是一种能有效解决该类问题的新兴手段。随着双向聚类得到越来越多的应用,现存算法的缺点与不足也逐渐暴露了出来,因此研究双向聚类问题是很有必要的。本文的研究目的是利用粒子群算法解决双向聚类问题,并通过一系列实验对比说明了结合粒子群优化的双向聚类算法的优越性。本文主要做的工作如下:(1)双向聚类算法是一种局部搜索算法,对于繁杂的基因数据矩阵,直接对其整体进行双向聚类,计算量大且聚类效果不理想。本文在粒子群算法的基础上,使用总体类间差异先对整个基因矩阵全局寻优,找出有一定相似性的基因子矩阵,再对其进行添加或删除行列的操作。使得到的双向聚类结构更加规整,避免了基因表达数据不均衡分类的情况。(2)双向聚类算法是一种多目标优化算法,FLOC算法作为经典双向聚类算法之一,却不能很好的同时优化多个目标。结合PSO算法,对FLOC算法的目标函数做出修改,提出了PSO-FLOC聚类算法,通过实验对比发现,PSO-FLOC算法对多目标优化问题表现更佳,并对算法中参数的取值进行了讨论。(3)在粒子群算法中,粒子只能沿着特定的轨迹搜索,从而不能保证以概率1收敛到全局最优,甚至不能收敛到局部最优。为了提高算法的全局搜索能力,结合具有量子行为的粒子群优化算法,形成了QPSO-FLOC聚类算法,并通过实验与PSO算法进行了分析比较,证明QPSO-FLOC算法能取得更好的聚类效果。
[Abstract]:Bioinformatics is an interdisciplinary subject that combines knowledge in biology, computer science, mathematics and chemistry. With the rapid development of science and technology, great breakthrough has been made in gene sequencing technology. At present, huge amounts of genetic data are produced every day, and the research focus of life science has shifted from how to obtain biological data to how to effectively analyze these data. At present, clustering is the main method to analyze and process gene expression data. The general clustering can only be carried out in the single direction of the row or column of the gene matrix. This method can only find the global information in the gene expression data matrix. However, a large amount of valuable biological information is often hidden in these local information. Bidirectional clustering is a new method to solve this kind of problem effectively. With more and more applications of bidirectional clustering, the shortcomings and shortcomings of the existing algorithms are gradually exposed, so it is necessary to study the bidirectional clustering problem. The purpose of this paper is to solve the bidirectional clustering problem with particle swarm optimization (PSO), and the superiority of bidirectional clustering algorithm combined with PSO is illustrated by a series of experiments. The main work of this paper is as follows: (1) Bidirectional clustering algorithm is a kind of local search algorithm. On the basis of particle swarm optimization (PSO), the global optimization of the whole gene matrix is carried out by using the difference of the whole class, and the gene submatrix with certain similarity is found, and then the operation of adding or deleting the column and column is carried out. The bidirectional clustering structure is more regular, and the unbalanced classification of gene expression data is avoided. (2) Bidirectional clustering algorithm is a multi-objective optimization algorithm. FLOC algorithm is one of the classical bidirectional clustering algorithms. But not very good at the same time optimization of multiple goals. Combined with the PSO algorithm, the objective function of the FLOC algorithm is modified and the PSO-FLOC clustering algorithm is proposed. Through the comparison of experiments, it is found that the PSO-FLOC algorithm performs better for the multi-objective optimization problem. The parameters of the algorithm are discussed. (3) in PSO, the particle can only be searched along a specific trajectory, thus the probability 1 can not converge to the global optimal, or even to the local optimal. In order to improve the global searching ability of the algorithm, the QPSO-FLOC clustering algorithm is formed by combining the particle swarm optimization algorithm with quantum behavior. The experimental results show that the QPSO-FLOC algorithm can achieve better clustering effect compared with the PSO algorithm.
【学位授予单位】：重庆理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13
[Abstract]:Bioinformatics is an interdisciplinary subject that combines knowledge in biology, computer science, mathematics and chemistry. With the rapid development of science and technology, great breakthrough has been made in gene sequencing technology. At present, huge amounts of genetic data are produced every day, and the research focus of life science has shifted from how to obtain biological data to how to effectively analyze these data. At present, clustering is the main method to analyze and process gene expression data. The general clustering can only be carried out in the single direction of the row or column of the gene matrix. This method can only find the global information in the gene expression data matrix. However, a large amount of valuable biological information is often hidden in these local information. Bidirectional clustering is a new method to solve this kind of problem effectively. With more and more applications of bidirectional clustering, the shortcomings and shortcomings of the existing algorithms are gradually exposed, so it is necessary to study the bidirectional clustering problem. The purpose of this paper is to solve the bidirectional clustering problem with particle swarm optimization (PSO), and the superiority of bidirectional clustering algorithm combined with PSO is illustrated by a series of experiments. The main work of this paper is as follows: (1) Bidirectional clustering algorithm is a kind of local search algorithm. On the basis of particle swarm optimization (PSO), the global optimization of the whole gene matrix is carried out by using the difference of the whole class, and the gene submatrix with certain similarity is found, and then the operation of adding or deleting the column and column is carried out. The bidirectional clustering structure is more regular, and the unbalanced classification of gene expression data is avoided. (2) Bidirectional clustering algorithm is a multi-objective optimization algorithm. FLOC algorithm is one of the classical bidirectional clustering algorithms. But not very good at the same time optimization of multiple goals. Combined with the PSO algorithm, the objective function of the FLOC algorithm is modified and the PSO-FLOC clustering algorithm is proposed. Through the comparison of experiments, it is found that the PSO-FLOC algorithm performs better for the multi-objective optimization problem. The parameters of the algorithm are discussed. (3) in PSO, the particle can only be searched along a specific trajectory, thus the probability 1 can not converge to the global optimal, or even to the local optimal. In order to improve the global searching ability of the algorithm, the QPSO-FLOC clustering algorithm is formed by combining the particle swarm optimization algorithm with quantum behavior. The experimental results show that the QPSO-FLOC algorithm can achieve better clustering effect compared with the PSO algorithm.
【学位授予单位】：重庆理工大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP311.13
【参考文献】
相关期刊论文 前10条
1 梁妍;;基于多目标的双聚类算法[J];信息通信;2016年07期
2 孙杰;吴陈;;相似性度量在基因表达聚类分析中的应用研究[J];现代电子技术;2012年06期
3 张敏;戈文航;;双聚类的研究与进展[J];微型机与应用;2012年04期
4 周，
本文编号：2390679
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2390679.html
