Predictive analysis and quality assessment in statistical learning

统计学习中的预测分析和质量评估

• 批准号: RGPIN-2022-04698
• 负责人: Gweon, Hyukjun
• 金额: $ 1.38万
• 依托单位: University of Western Ontario
• 依托单位国家: 加拿大
• 项目类别: Discovery Grants Program - Individual
• 财政年份: 2022
• 资助国家: 加拿大
• 起止时间: 2022-01-01 至 2023-12-31
• 项目状态: 已结题
• 来源: https://www.nserc-crsng.gc.ca/ase-oro/Details-Detailles_eng.asp?id=765664
• 关键词: Predictive; analysis; quality; assessment; statistical

Predictive modeling is a central topic in the fields of data mining and statistical learning, where researchers and practitioners seek the advanced analytic tools and computational methods for better use of data resources. In many applications, it is of utmost importance to develop effective systems taking different types and characteristics of data into account. My proposed research program is expected to make significant contributions in this research field through the development and application of analytical approaches and computational methodologies in the study of statistical learning and assessment processes. More specifically, my research team and I will concentrate on the following three themes: (1) Assessment/improvement of the reliability of probabilistic classifiers; (2) Sparse-labeled data analysis in insurance; and (3) Application of pre-trained language models to automatic survey coding. Theme (1) studies predictive model assessment with a special focus on probabilistic multinomial classifiers. The outcomes from Theme (1) will contribute to the development of the theory and methodology in assessing prediction uncertainty in supervised classification. In Theme (2), following the recent trend in valuation of large variable annuity portfolios, new efficient data mining frameworks will be designed and examined to mitigate the heavy computational challenge of the problem. Another contribution of my proposed research program, which is the focus of Theme (3), is to offer survey researchers a deeper understanding of recent pre-trained language models and tools for effective survey coding. This research will provide innovative insights for researchers and practitioners in various fields including data mining, insurance/finance and business analytics, and train highly qualified personnel who will be prepared to lead in the related research areas and the high-tech industry.

预测建模是数据挖掘和统计学习领域的中心主题，研究人员和从业者寻求先进的分析工具和计算方法以更好地利用数据资源。在许多应用中，开发考虑不同类型和数据特征的有效系统至关重要。我提出的研究计划预计将通过在统计学习和评估过程的研究中分析方法和计算方法的开发和应用，在该研究领域做出重大贡献。更具体地说，我和我的研究团队将专注于以下三个主题：（1）概率分类器可靠性的评估/改进； (2) 保险中的稀疏标记数据分析； (3)预训练语言模型在自动调查编码中的应用。主题 (1) 研究预测模型评估，特别关注概率多项式分类器。主题（1）的成果将有助于监督分类中评估预测不确定性的理论和方法的发展。在主题（2）中，遵循大型可变年金投资组合估值的最新趋势，将设计和检查新的高效数据挖掘框架，以减轻该问题的繁重计算挑战。我提出的研究计划的另一个贡献是主题（3）的重点，是让调查研究人员更深入地了解最新的预训练语言模型和有效调查编码的工具。这项研究将为数据挖掘、保险/金融和商业分析等各个领域的研究人员和从业者提供创新见解，并培养高素质人才，为相关研究领域和高科技行业做好准备。