Using Knowledge Discovery in Database & Data Mining to Develop Techniques in Medical Informatics Applied to Surgical Databases

医学信息学领域使用的工具是数据库中的知识发现(KDD)或数据挖掘。KDD涉及多个领域：统计、模式识别、人工智能和数据仓库。然而，它已经成为一门有自己目标的不同学科。数据挖掘的目的是提取以前未知的和潜在有用的信息。KDD与更传统的统计工具不同，它侧重于模型识别，同时最大限度地减少错误分类。统计方法往往侧重于估计和模型识别。然后必须用标准的统计技术来验证知识发现过程产生的假设。越来越清楚的是，由于数据库的规模和复杂性，更传统的技术是不够的。一家拥有450张床位的医院维护着一个数据库，其中包括6个月的实验室测试和药物治疗，以及15个月来患者随访的摘要。计算机的存储容量必须增加12千兆字节。对调查人员检查数据关系的依赖必须让位于自动化过程。知识发现可以产生有趣的和意想不到的假设，这些假设可以用统计方法进行检验。自动假设形成和检验周期可以持续到重要模式出现。该项目涉及研究和开发应用于主要存储在安联保健系统中的几个数据库的数据挖掘技术。其中包括一个数据库，其中包含对肯塔基州1万名女性队列的25年随访，以检查健康和生活方式的模式。此外，这些技术将被应用于学生信息数据库，以检查学生习惯、学生成功和学生学习之间的关系。利用开发的材料和数据库，将向数学系的学生讲授数据挖掘的应用课程。该GOALI项目由MPS多学科活动办公室(OMA)和数学科学处(DMS)共同支持。

The tools used in the field of medical informatics are those of knowledge discovery in databases (KDD) or alternatively, data mining. KDD spans a variety of fields: statistics, pattern recognition, artificial intelligence, and data warehousing. However, it has become a distinct discipline with its own objectives. The purpose of data mining is to extract previously unknown and potentially useful information. KDD differs from the more traditional statistical tools by focusing on model identification while minimizing misclassification. Statistical methods tend to focus on estimation and model identification. The hypotheses generated by the KDD process must then be validated by standard statistical techniques. It has become increasingly clear that more traditional techniques are not adequate because of the size and complexity of the databases. One 450-bed hospital maintained a database with six months of laboratory tests and medications as well as summaries of patient followup for 15 months. The storage capacity of the computer had to be increased by 12 gigabytes. The dependence upon an investigator examining relationships in the data must yield to an automated process. KDD can generate interesting and unexpected hypotheses which can be examined by statistical methods. The automatic hypothesis formation and testing cycle can continue until important patterns emerge. This project involves the study and development of data mining techniques applied to several databases primarily stored in the Alliant Health System. This includes a database containing a 25-year followup on a cohort of 10,000 women in Kentucky to examine patterns of health and lifestyle. In addition, the techniques will be applied to a database of student information to examine relationships between student habits, student success, and student learning. Using the developed materials and databases, a course will be taught to students in the Department of Mathematics on applications of data mining. This GOALI project is jointly supported by the MPS Office of Multidisciplinary Activities (OMA) and the Division of Mathematical Sciences (DMS).