Pattern-discovery and related algorithms

模式发现和相关算法

• 批准号: 8042-2009
• 负责人: Chiu, David
• 金额: $ 1.38万
• 依托单位: University of Guelph
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2010
• 资助国家: 加拿大
• 起止时间: 2010-01-01 至 2011-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=450214
• 关键词: Pattern; discovery; related; algorithms

Identifying significant relationships in an uncertain environment, using data from multiple sources, is extremely important in discovering novelty in pattern discovery. For example, linking genotype and phenotype attributes plays a vital role in understanding the biological processes, under normal or novel conditions involving genetic modifications and changes, in problems such as the engineering of novel bioproducts. We propose to study biological processes as an information system that includes identifying relationships of functional segments and molecular sites of biomolecules, and aim at understanding their effects on biomolecular structure and functionality. We study the problem as a data-mining problem using the paradigm of granular computing and propose analysis methods based on different types of multiple-value (multi-attribute) association patterns extracted from an alignment of data. The patterns are based on analysis of statistical testing from a prior model. The goal is to provide a very flexible analysis/synthesis through selection, partitioning and attribute clustering etc. of the data. The paradigm incorporating hierarchical and modular analysis aims at a more generalized model than contiguous sequential pattern analysis and is more descriptive than multiple-variable analysis. After detecting significant patterns, further pattern-mining methods to understand the inter-pattern relationships will be developed.

使用来自多个来源的数据来识别不确定环境中的重要关系对于发现模式发现中的新颖性极其重要。例如，在涉及遗传修饰和变化的正常或新条件下，在新型生物产品工程等问题中，连接基因型和表型属性对于理解生物过程起着至关重要的作用。我们建议将生物过程作为一个信息系统来研究，包括识别生物分子的功能片段和分子位点的关系，并旨在了解它们对生物分子结构和功能的影响。我们使用粒度计算范式来研究该问题作为数据挖掘问题，并提出基于从数据对齐中提取的不同类型的多值（多属性）关联模式的分析方法。这些模式基于对先前模型的统计测试的分析。目标是通过数据的选择、分区和属性聚类等提供非常灵活的分析/综合。结合分层和模块化分析的范式旨在建立比连续序列模式分析更通用的模型，并且比多变量分析更具描述性。在检测到重要模式后，将开发进一步的模式挖掘方法来理解模式间的关系。