VISUAL AND ANALYTICAL TOOLS FOR CLUSTER ANALYSIS

点击翻译按钮获取中文摘要

This subproject is one of many research subprojects utilizing the resources provided by a Center grant funded by NIH/NCRR. The subproject and investigator (PI) may have received primary funding from another NIH source, and thus could be represented in other CRISP entries. The institution listed is for the Center, which is not necessarily the institution for the investigator. The output of a clustering algorithm is a list of clusters and their membership, in some cases supplemented by record relationships, such as proximity values, connections. The goal of this project is to design new analytical and visual tools and techniques to provide insights into the multiple clustering algorithm results. Our work addresses questions in the biomedical field by devising tools that enable scientists to look through multiple clustering methods and analyze the results, in order to aid the data exploration. The biomedical field has undergone strong growth over the past two decades, generating massive data sets that require new approaches. Using visual techniques to harness the experimental scientists intuition seems common sense but the strong dependence on automatic computation is still present in the field. This project is identified by three specific aims: I. Develop, implement and validate analytical measures of similarity of multiple clustering results: - Formally define the similarity and dissimilarity of clustering results - Identify new analytical measures that capture the definitions of similarity and dissimilarity - Identify new analytical measures that capture the definitions of similarity and dissimilarity - Define and implement specific applications of the analytical measures II. Visualizations for identification of similar and dissimilar records, clusters and subsets: - Design and implement visualization techniques that work with the new analytical measures - Create and implement interaction methods that facilitate the exploration III. Investigate and improve the performance of the techniques and algorithms and apply and extend the techniques and tools on large experimental data sets