CISC333 Data Mining

CISC333 Data Mining: Fall 2011

Announcements

Introduction

This course is offered in Fall 2011 in slot 14 in Botterell B148. The prerequisites are CISC121 and a stats course. Although the course is numbered at 3rd Year, second-year students who wish to take it may do so.

Course Material

The outline form of the class presentations is available here in the subdirectory announced in class. File names are of the form modulei.ppt, where i starts at 0. You may wish to print these and take notes on them, or load them onto your laptop and bring them to class to follow along. I will add substantial content to these outlines in class.

When each module is completed, the annotated version will also be available in the other subdirectory announced in class, with file names of the form donemodulei.ppt. (Some of the powerpoint slides have been adapted from a college data mining course by Piatetsky-Shapiro and Parker (with permission)).

Tutorials

The tool we will use for most of the practical work in the course is Rapidminer, a development from the Weka toolkit. You can find extensive tutorial material on the Rapid website.

Rapidminer is available for free download at http://rapid-i.com/; you need Rapidminer Community Edition (should be Rapidminer 5).

Rapidminer is also available on the Caslab machines, under either Windows or Linux.

Tutorial on using Matlab (useful for visualization) and the SVD and ICA matrix decompositions.

Exercises and Assignments

Exercises are a chance for you to get some hands-on experience. The exercise questions will often be open-ended. You might expect to spend 3 or 4 hours on these sheets each week. Each one is marked on this scale: acceptable; inadequate; or not seriously attempted. There will be five exercise sheets in the first half of the term.

(Exercise sheets will appear a week or ten days before they are due.)

Exercise sheet 1 (due Friday of Week 2)
Exercise sheet 2 (due Friday of Week 3)
Exercise sheet 3 (due Friday of Week 4)
Exercise sheet 4 (due Friday of Week 5)
Exercise sheet 5 (due Friday of Week 6)
A Project for which preliminary results are due Thursday of Week 9, and the final report is due Thursday of Week 11.

Outline

This table describes what we will cover, keyed to the modules and text.

Module Status Content Text Refs

Module 0 Done Introduction Chapter 1

Module 1 Done Prediction and Exploration Bits of Chapter 2

Module 2 Done Data preparation and model quality Chapter 2

Module 3 Done Simple predictors

Module 4 Done Decision trees Chapter 4

Module 5 Done More decision trees

Module 6 Done Neural networks Chapter 5.4

Module 7 Done Support Vector Machines Chapter 5.5

Module 8 Done Rule based systems Chapter 5.1

Module 9 Done Object selection: sampling, ensemble techniques Chapter 5.6

Module 10 Done Attribute selection Section 2.3.4

Module 11 Done Prediction case studies

Module 12 Done Clustering I: similarities, k-means Chapter 8.2

Module 13 Done Clustering II: Expectation-Maximization Section 9.2.2

Module 14 Done Clustering III: Top-down and bottom-up clustering Chapter 8.3

Module 15 Done Clustering IV: Matrix decompositions

Module 16 Done Clustering V: Clustering Large Datasets Section 9.5.2,Section 8.4.2

Module 17 Done Visualization Chapter 3.3

Module 18 Done Clustering Case Study

Module 19 Done Clustering VI: Biclustering

Module 20 Skipped Biclustering Case Study: Topic Detection

Module 21 Done Mining the Web

Module 22 Done Collaborative Filtering

Module 23 Done Counterterrorism and Fraud

Module 24 Skipped Summary

Avail means that the basic powerpoint is available. Done means that the marked-up powerpoint is available.

Text

Tan, Steinbach, Kumar, Introduction to Data Mining, Addison-Wesley, 2006, ISBN 0-321-32136-7 ($104 at Amazon, $97.95 at Campus Bookstore).

Assessment

There are 3 deliverables:

5 weekly exercise sheets, due in each of weeks 2-6. These exercise sheets are not marked as such. However, for each exercise sheet that is not adequate, your final mark in the course will be reduced by a factor of 0.02. The exercise sheets are useful both for understanding the material and as ground work for the project, so they are worth doing!
A project in which you can apply data mining techniques to a real dataset and report on what you discover, due at the end of week 11, and worth 30% of your final grade.
A final examination, worth the remaining 70% of your grade. You must pass the exam to pass the course. You may bring one 8.5x11 sheet of notes into the exam.

Note that the assessment in the course is backloaded, so please take this into account when planning your procrastination.

Your instruction team

Instructor

David Skillicorn
528 Goodwin Hall
skill cs queensu ca
533 6065

Questions? Try asking me before or after class, or come and find me at my office any time I'm there. I have a schedule posted by my door. I'm teaching this course, another in slot 2, and all Friday afternoon, so there's no use trying around those times.

I have a scheduled office hour from 1-2 on Tuesdays.

Teaching Assistant

Unless the enrollment picks up, we won't have one. Why not encourage your friends to take the course?

Resources

Study skills. You probably know all of the conventional wisdom about how to learn, but perhaps you don't actually use it. Here is an excellent link: Study Hacks.

You may also want to subscribe to, or read: KD Nuggets

Module	Status	Content	Text Refs
Module 0	Done	Introduction	Chapter 1
Module 1	Done	Prediction and Exploration	Bits of Chapter 2
Module 2	Done	Data preparation and model quality	Chapter 2
Module 3	Done	Simple predictors
Module 4	Done	Decision trees	Chapter 4
Module 5	Done	More decision trees
Module 6	Done	Neural networks	Chapter 5.4
Module 7	Done	Support Vector Machines	Chapter 5.5
Module 8	Done	Rule based systems	Chapter 5.1
Module 9	Done	Object selection: sampling, ensemble techniques	Chapter 5.6
Module 10	Done	Attribute selection	Section 2.3.4
Module 11	Done	Prediction case studies
Module 12	Done	Clustering I: similarities, k-means	Chapter 8.2
Module 13	Done	Clustering II: Expectation-Maximization	Section 9.2.2
Module 14	Done	Clustering III: Top-down and bottom-up clustering	Chapter 8.3
Module 15	Done	Clustering IV: Matrix decompositions
Module 16	Done	Clustering V: Clustering Large Datasets	Section 9.5.2,Section 8.4.2
Module 17	Done	Visualization	Chapter 3.3
Module 18	Done	Clustering Case Study
Module 19	Done	Clustering VI: Biclustering
Module 20	Skipped	Biclustering Case Study: Topic Detection
Module 21	Done	Mining the Web
Module 22	Done	Collaborative Filtering
Module 23	Done	Counterterrorism and Fraud
Module 24	Skipped	Summary