# 26:198:622:01 Index: 36745

A Ph.D. course offered in the fall of 2009

Accounting, Business Ethics, and Information Systems
Rutgers Business School - Newark and New Brunswick
Rutgers University

Prof. Alexander Kogan
One Washington Park #924 (Newark), (973) 353-1064




Newark Campus; 1WP-534


1:00 p.m. - 3:50 p.m.

Overview: Many successful applications of machine learning exist already, including systems that analyze past sales data to predict customer behavior, identify fraudulent credit card transactions, and recognize faces or spoken speech. This course will cover supervised learning, Bayesian decision theory, parametric methods, multivariate methods, dimensionality reduction, clustering, nonparametric methods, decision trees, linear discrimination, multilayer perceptrons, hidden Markov models, assessing and comparing classification algorithms, and combining multiple learners.

Required Textbook:

The course will utilize the Blackboard online facilities, which can be found at:

For a student to gain access to our Blackboard system, they must be enrolled and they must have a NETID (pegasus or eden account). Once an enrolled student obtains a NETID, they will be added to the roster within 2 business days. Students should also check their email account in the system and if it is not correct, they need to update their official student record. If they do not have a valid email associated with their official student record, their email address will show up as Students who do not have a NETID, can create one online using this link:

Coursework: The coursework includes attendance of lectures and participation in class discussions, writing a research paper and presenting it in class, and completing a computational project in machine learning and submitting its results for evaluation.

Research Paper: Every student is required to write a research paper devoted to an important topic in machine learning. While it is expected that most students will choose a topic devoted to applications of machine learning in business, other topics may work as well. Every student is required to prepare a three page long proposal for the research paper, and submit this proposal for instructor's evaluation by October 22, 2009. The Word document must be posted to the appropriate Blackboard discussion forum, and its printout should be submitted in class. The research paper should be presented during the last meeting of the class on December 17, 2009. Both the research paper and the presentation should also be posted to the appropriate Blackboard forum before the last meeting of the class. The research articles to be covered in the research paper can be found in the following theoretical and applied journals publishing relevant articles:

Most of these journals are available through the Rutgers University Library subscriptions, and can be accessed from campus computers or from home through the library Web site:

The following online research tools can be useful in conducting bibliographic searches for your research paper: Important guidance on writing machine learning papers can be found in the following manuscript:

Computational Project: Every student is required to carry out a computational project focused on experimental comparison of several machine learning methods on different datasets. The computational experiment should compare at least FIVE different machine learning methods on at least FOUR different datasets. Every student is required to prepare a computational project proposal and submit this proposal for instructor's evaluation by November 12, 2009. The Word document describing the machine learning methods and the datasets to be utilized in the experiment, as well as the proposed experimental methodology, must be posted to the appropriate Blackboard discussion forum, and its printout should be submitted in class. The description of results of the computational experiment should also be posted to the appropriate Blackboard forum and its printout should be submitted during the last meeting of the class on December 17, 2009. It is recommended that the computational experiments be conducted using the public domain Machine Learning software package called Weka (using the most recent book version - currently 3-4-15). Here are some useful Weka links:

Another good public domain Machine learning software package that can be use in experiments in addition to (or instead of) Weka is RapidMiner:

The datatsets for use in computational experiments can be obtained from the UCI Machine Learning Repository:

It is absolutely essential to start working on the research paper and the computational project as soon as possible.

Grading: The research paper and the computational project evaluations will be the basis for the course grade:


Research Paper


Computational Project

Communications and Course E-Mailing List:

The best way to contact me is via email. The course is supported by the RAMS e-mailing list ml-phd-list. The list membership is automatically synchronized with the current class roster. Make sure that your current e-mail address is available in the Rutgers online directory. To post a message to the list, e-mail it to

All the postings to this list are permanently archived and available from

Please note that your postings should be appropriate for this course.

Every student is responsible for maintaining the current e-mail address in the Rutgers Online Directory. The directory record update page can be found at:

Preliminary Schedule:

  1. 09/03/2009
    Chapter 1 - Introduction to Machine learning

  2. 09/10/2009
    Chapter 2 - Supervised Learning

  3. 09/17/2009
    Chapter 3 - Bayesian Decision Theory

  4. 09/24/2009
    Chapter 4 - Parametric Methods

  5. 10/01/2009
    Chapter 5 - Multivariate Methods

  6. 10/08/2009
    Chapter 6 - Dimensionality Reduction

  7. 10/15/2009
    Chapter 7 - Clustering

  8. 10/22/2009
    Chapter 8 - Nonparametric Methods
    • Research Paper Proposal is due

  9. 10/29/2009
    Chapter 9 - Decision Trees

  10. 11/05/2009
    Chapter 14 - Assessing and Comparing Classification Algorithms

  11. 11/12/2009
    Chapter 10 - Linear Discrimination
    • Computational Project Proposal is due

  12. 11/19/2009
    Chapter 11 - Multilayer Perceptrons

  13. 12/03/2009
    Chapter 13 - Hidden Markov Models

  14. 12/10/2009
    Chapter 15 - Combining Multiple Learners

  15. 12/17/2009
    Research Paper Presentations
    Computational Projects are due