Partition-based Cluster Analysis on Time Series Gene Expression Data

• Main Authors: Tsung-Lung Lee, 李宗龍
• Other Authors: Huang-Cheng Kuo
• Format: Others
• Language: en_US
• Online Access: http://ndltd.ncl.edu.tw/handle/r5u5ey

碩士 === 國立嘉義大學 === 資訊工程學系研究所 === 97 === Abstract Gene expression data produced by DNA microarray experiments advance the study of functions of genes. Researchers use microarray to generate large amount of gene expression data and observe the differences among gene expression. Data of gene expression time series describes the trend of gene behaviors. Recently, clustering is often used and is a popular analysis for studying gene expression time series data. In the same cluster, genes have similar behavior. Cluster analysis demonstrates the relativity among genes. In the paper, we use a similarity measurement, named LCSS (Longest Common Subseries Similarity), to overcome the influence of “shift-effect.” This influence is an issue which needs to overcome when using similarity measurement. In clustering, sequential pattern mining technique is used to resolve the number of clusters for partitioning-based clustering algorithm, and several similarity measurements (like Pearson correlation coefficient) are used to determine neighbors in KNN, A mechanism based on nearest neighbor is used as the criteria for objects relocation in the clustering algorithm.