From ceginfo@u.washington.edu Tue Feb 1 16:20:18 2000 Received: from jason03.u.washington.edu (root@jason03.u.washington.edu [140.142.77.10]) by lists.u.washington.edu (8.9.3+UW99.09/8.9.3+UW99.09) with ESMTP id QAA42632 for ; Tue, 1 Feb 2000 16:20:17 -0800 Received: from homer09.u.washington.edu (ceginfo@homer09.u.washington.edu [140.142.77.13]) by jason03.u.washington.edu (8.9.3+UW99.09/8.9.3+UW00.01) with ESMTP id QAA19572 for ; Tue, 1 Feb 2000 16:20:15 -0800 Received: from localhost (ceginfo@localhost) by homer09.u.washington.edu (8.9.3+UW99.09/8.9.3+UW99.09) with ESMTP id QAA45102 for ; Tue, 1 Feb 2000 16:20:14 -0800 Date: Tue, 1 Feb 2000 16:20:14 -0800 (PST) From: Civil and Environmental Engineering To: cegrads@u.washington.edu Subject: Stat 593B (fwd) Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII DEPARTMENT OF STATISTICS COURSE ANNOUNCEMENT Statistics 593B Spring Quarter 2000 DATA MINING Instructor: R. Douglas Martin (doug@statsci.com) Time: 3:30 4:45 Thursday and Friday Place: Loew 118 Credits: 3 Credits, Graded CR/NC Enroll. Limit: 20 COURSE SUMMARY The subject of "Data Mining" has been attracting intense interest in recent years, primarily because of the promise it appears to hold for payoff in a wide variety commercial applications such as credit card fraud detection, cross-selling, WEB-based real-time database marketing, etc. Data mining means different things to different people, but certainly a central theme of data mining is the exploration and analysis, by automatic or semiautomatic means, of large quantities of data in order to discover meaningful patterns and decision rule. As such, the collection of commonly used data mining methods comprise a grab bag of algorithms and techniques borrowed from statistics, computer science machine learning and artificial intelligence. The purpose of this course is threefold: (a) provide an introduction to the collection of techniques that seem to represent the current grab bag, and some of their applications, (b) introduce some recently developed methods that should prove to be quite useful data mining tools, and (c) discuss some of the open computational challenges of dealing with extremely large data sets, including real-time data such as financial tick data and WEB click-stream data. TEXTBOOK Data Mining Techniques (1997) by M. J. A. Berry and G. Linoff, J. Wiley. This is a very clearly written high-level treatment that emphasizes marketing, sales and customer support applications. The above textbook will be supplemented by handout notes, papers and other recommended reading that will provide deeper coverage of the technical details of data mining methods and open problems. COURSE OUTLINE Part 1: Introduction to Data Mining (5 weeks) Applications motivations and what is a "very large" data set The "Data Mining Process" and overview of data mining methods Market basket analysis Memory-based reasoning Automatic cluster detection Link analysis Decision trees Artificial neural networks Genetic algorithms Part 2: Cutting Edge Data Mining Methods (23 weeks) Bagging, boosting, etc. Data squashing Bayesian methods Robust methods for outlier detection and data cleaning Part 3: Computational Challenges in Data Mining (2-3 weeks) Data mining methods that are amenable to block recursive updating Clustering for very large data sets Decision trees for very large data sets To request disability accommodations, contact the Office of the ADA Coordinator ten days in advance of the event. 543-6450 (voice); 543-6452 (TDD); 685-3885 (FAX); access@u.washington.edu (email). .