TSP算法在时序基因分类中的研究与应用

发布时间：2018-03-20 17:35

本文选题：基因分类　切入点：时间序列基因　出处：《郑州大学》2017年硕士论文　论文类型：学位论文

【摘要】：在生物信息学中,从大量数据中挖掘对研究有价值的信息越来越受到研究人员的重视。时序基因表达数据是生物生长过程中不同时间点采集的基因数据,作为研究热点之一的时间序列基因数据挖掘也越来越受到研究者们广泛关注和研究。时间序列基因数据分类是一个非常重要的课题,它包含了大量与时间相关的生物信息。由于基因表达数据的“高维、小样本”特点。传统的机器学习方法在处理高维数据上不仅会建立复杂的分类模型,而且得到的分类结果很难用生物学的角度理解。最高得分对算法(Top Scoring Pairs,TSP)是利用少量特征对对样本进行预测的分类器,此算法在处理基因数据上达到了很好的效果,但还不能处理时间序列基因表达数据。为了充分利用TSP算法优势,本文提出动态最高得分对算法(Dynamic Top Scoring Pairs,DTSP),将TSP算法改进为能处理时序基因数据的分类器。本文的主要工作如下:1.本文提出DTSP算法,将趋势的思想融入TSP算法中,使其能够处理时序基因。DTSP算法不仅考虑了相邻时间点的差异趋势,并且将差异的值的大小考虑进去,提高了分类准确率。同时算法也改进了预测模型,用趋势的规则进行预测。本文使用时序基因数据集进行实验,将DTSP算法和支持向量机、K近邻算法做横向对比,并将DTSP算法中包含的三种算法做纵向对比。实验证明,改进后的分类器得到了很高的分类效果,选出了最具分类能力的特征对。选出的特征对将对生物学基因研究提供了新的思路。2.在动态最高得分对算法理论的基础上,本文设计了基于DTSP算法的时序基因分类系统。该系统不仅实现了DTSP算法对时序基因的分类,同时兼容传统的机器学习分类算法,能够为用户提供二次开发的接口,集成自定义的时序基因分类算法。该系统不仅能对时序基因分类,而且可以将DTSP计算出的最高特征对展示出来,可以用于医疗生物学的进一步研究。
[Abstract]:In bioinformatics, researchers pay more and more attention to mining valuable information from a large amount of data. As one of the research hotspots, time series gene data mining has been paid more and more attention by researchers. Classification of time series gene data is a very important subject. It contains a lot of time-related biological information. Because of the "high dimension, small sample" characteristics of gene expression data, traditional machine learning methods not only build complex classification models in processing high-dimensional data. The results obtained are difficult to understand from a biological point of view. The highest score pair algorithm, Top Scoring Pairsp, is a classifier that uses a small number of features to predict samples, and the algorithm achieves good results in the processing of genetic data. In order to make full use of the advantage of TSP algorithm, In this paper, a dynamic Top Scoring pairsn DTSP algorithm is proposed, and the TSP algorithm is improved to be a classifier capable of processing temporal gene data. The main work of this paper is as follows: 1.The DTSP algorithm is proposed in this paper, which integrates the trend idea into the TSP algorithm. The algorithm not only considers the trend of the difference between adjacent time points, but also takes the value of the difference into account, which improves the classification accuracy, and also improves the prediction model. In this paper, the DTSP algorithm and the support vector machine (SVM) nearest neighbor algorithm are compared horizontally, and the three algorithms contained in the DTSP algorithm are compared longitudinally. The improved classifier has obtained a high classification effect and selected the most capable feature pairs. The selected feature pairs will provide a new way of thinking for the study of biological genes. 2. Based on the theory of dynamic maximum score pair algorithm, In this paper, a sequential gene classification system based on DTSP algorithm is designed, which not only realizes the classification of sequential genes by DTSP algorithm, but also is compatible with the traditional machine learning classification algorithm, and can provide a secondary development interface for users. The system can not only classify sequential genes, but also display the highest characteristic pairs calculated by DTSP, which can be used in the further study of medical biology.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：Q811.4

【相似文献】