Knowledge Discovery and Data Mining

Yoav Benjamini
Department of Statistics and Operations Research
972-3-6408756 (voice)
972-3-6409357 (fax)
ybenja@post.tau.ac.il
Office Hours: Tuesday 1500-1600.

Lecture Hours:

Wednesday 16-19 Schrieber 08

Important Announcement

NONE

Books:

Inteligent Data Analysis - an Introduction

Data Analysis and Regression

Computing:

The statistical software we shall use is Splus. Tutorial material is available in the following documents (which can also be printed).

Notes on S: A programming environment for Data Analysis and Graphics Splus by Venables and Smith.
Introductory Guide to Splus by Ripley.

Other books on S and Splus which we shall use are:

Statistical Models in S
by Chambers and Hastie.
S-plus Guide

Guide to Statistical and Mathematical Analysis
The New S Language

by Becker Chambers and Wilks.

Regarding the SAS software:

Mina thinks the online help that is accesible from the program is more helpful for beginners. You can get her help by writing to mina@post.tau.ac.il. This year introductory talk by Vered can be viewed, but you may benefit from Ronen's Introductory talk given 4 years ago about SAS which is available as PowerPoint Presentation.

Topics:

Introduction
Data we get and data we analyse (Types of Data; Symmetry Additivity and Linearity; Data Cleansing) (Splus program for analysing skewness)
Visualization of Data (Principles; Foundamental Problems; Parallel Coordinates; Manet) Power Point Presentation .
Mining of Association Rules.
The pitfalls of multiplicity and the False Discovery Rate Power Point presentation (including material on model selection)
Multiple Linear Regression
Jacknifing and cross validation (olar)
Model Selection in high dimensions
Logistic Regression
Nearest Neighbor and Case Base Reasoning
The curses and blessings of high dimensionality
Decision Trees
Bootstrapping (retsu'at hamagaf) and Bagging
Neural Nets Models

Please keep on looking at this list as the course progresses. Some of these will be activated later.

Requirements:

Homework sets will be assigned and graded.

Targil 1 Handout
Targil 2 using data which can also be read from math.tau.ac.il/~ybenja/segal/seg.dat . Now available comments on solutions .
Targil 3 using data which can also be read from math.tau.ac.il/~ybenja/kdd/ex2.txt
Targil 4: For the country data for year 2000, to be found at http://pwt.econ.upenn.edu/php_site/pwt_index.php, build a model explaining Per Capita Gross National Product by the other variables. Summarize your results in 2 pages (including figures).
Targil 5

A final independent group project of analysing a real data set is to be handed in by the end of the exam break. The project might consist of participating in the a data mining competition. Submitting in time for the competition will be specially awarded by appropriate grade. Participation is in groups.