NSF Conference in the Mathematical Sciences on Data Mining and Bioinformatics; January 8-10, 2004; Gainesville, FL

DMS-0337163 PI：George Casella2004年1月8日至10日在佛罗里达大学举行的关于数据挖掘和生物信息学的数学科学会议对帮助我们分析和理解大量数据的统计工具有很大的需求。传统的统计方法往往无法处理这类数据集的潜在复杂性。一些潜在的统计问题是模型选择，包括在模型空间中搜索的算法、稳健性、数据质量和采样、多重性问题、高维、小样本(“大p小n”)问题的推断、数据的适当缩放，以及基于来自医学图像、微阵列或环境监测的复杂数据集的推断。这类数据固有的生物学问题包括根据DNA序列确定蛋白质的三维结构，以及从微阵列收集的数据确定数千个基因的差异表达水平。数据挖掘(DM)是针对海量数据集的方法的总称。生物和基因组数据的数据挖掘通常被称为生物信息学(Bio)。数据挖掘和生物信息学的主题非常适合召开NSF地区性会议。它吸引了具有不同统计兴趣的广泛领域的研究人员，从对互联网交易感兴趣的人到欺诈检测到微阵列再到蛋白质结构。根据这些海量数据集做出推论的挑战将吸引那些对理论和方法统计感兴趣的人。最后，接受在现有统计方法不令人满意的情况下分析非正统数据的精细挑战的兴奋，无疑将吸引关注统计学应用的研究人员。希望这次会议将提供对DM/Bio的工作和使用的最新水平的评估，提出开放的问题，并促进学术界、工业界和政府的研究工作者之间的合作，努力为这些问题提供解决方案，并回答对科学和社会都非常重要的问题。我们预计这次会议将引起全国研究人员(特别是年轻研究人员)、佛罗里达大学和邻近大学的教职员工和研究生对这一主题的兴趣，并促进初级和高级研究人员之间的互动。

DMS-0337163 PI: George CasellaABSTRACTNSF Conference in the Mathematical Sciences on Data Mining and Bioinformatics at the University of Florida, January 8-10, 2004There is a large demand for statistical tools to help us analyze and understand massive amounts of data. Traditional statistical approaches often fail to cope with the underlying complexity of such datasets. Some of the potential statistical issues are model selection, including algorithms to search through model spaces, robustness, data quality and sampling, multiplicity issues, inference in high-dimensional, small sample ("large p small n") problems, appropriate scaling of data, and inference based on complex datasets from medical images, microarrays or environmental monitoring. Biological questions inherent in such data include determining the three dimensional structure of proteins based on DNA sequences and determining the differential expression levels of thousands of genes from data collected on microarrays. Data mining (DM) is the generic term that encompasses such methods for massive datasets. Data mining on biological and genomic data is often called Bioinformatics (Bio). The topic of Data Mining and Bioinformatics is ideal for a NSF regional conference. It appeals to a wide spectrum of researchers with diverse statistical interests from those interested in internet trafficking to fraud detection to microarrays to protein structure. The challenge of drawing inferences based on these massive datasets will appeal to those interested in theoretical and methodological statistics. Finally, the excitement of accepting the fine challenge of analyzing unorthodox data where existing statistical methodology is not satisfactory will undoubtedly fascinate researchers concerned with applications of statistics.It is hoped that this conference will provide an assessment of the current state of the art in the workings and use of DM/Bio, bring up open problems, and foster collaboration among research workers in academia, industry and government in an effort to provide solutions to these problems and answer questions of great importance to both science and society. We expect the conference to generate interest in this topic among researchers nationwide (particularly young researchers), among faculty and graduate students at the University of Florida and neighboring universities, and promote interactions between junior and senior researchers.