CS 553
Intelligent Data Anaysis

Description: Differences between data and knowledge, assessing knowledge; Data analysis process, methods, tasks and tools; Practical data analysis; Data understanding, attribute understanding, data quality, data visualization, correlation analysis, outlier detection, missing values; Principles of modeling, model classes, fitting criteria and score functions, model fitting, types of errors; Data preparation, feature selection, dimensionality reduction, record selection, improving data quality; Use of machine learning and data mining techniques in intelligent data analysis. Credit units: 3.

Semester: Fall 2012
Classroom: EA-502
Schedule: Monday 10:40 - 12:30; Thursday 9:40 - 10:30
Office Hours: Wednesday 15:40 - 17:40
Classroom: EA-418
Instructor: H. Altay Güvenir

Main Text Book:
Michael R. Berthold, Christian Borgelt, Frank Höppner, Frank Klawonn Guide to Intelligent Data Analysis: How to Intelligently Make Sense of Real Data, Springer-Verlag, London 2010. Link to the web page at publisher

Related Journals:
Intelligent Data Analysis: An International Journal

Related Papers:

• On Comparing Classifiers Pitfalls to Avoid and a recommended approach
• Statistical Comparison of Classifiers over multiple dataset

Weekly Schedule

1. Introduction, motivation, Difference between data and Knowledge, Assessing knowledge, Data analysis process (CRISP-DM), Methods, Tasks and Tools
2. Practical Data Analysis, Project Understanding
3. Data Understanding, Attribute Understanding, Data Quality, Data Visualization
4. Correlation Analysis, Outlier Detection, Missing Values, Data Understanding
5. Tools: KNIME, R, Weka, RapidMiner, RIMARC
6. Principles of Modeling, Model Classes
7. Model Classes, Types of Errors
8. Feature selection, Dimensionality Reduction, Record selection
9. Improving Data Quality, Missing Values
10. Workshop
11. Workshop
12. Workshop
13. Workshop
14. Workshop

Project Proposal (Due: Oct. 11, 2012)

Project Proposal should include information about the following:

• Domain of the data set to be analyzed,
• Expert who will evaluate the results (project owner),
• Source of the data set, where it is to be obtained,
• Possible list of features (independent parameters), their types, expected values,
• Posssible dependent parameters, their types, expected values,
• Possible challenges for data cleaning (e.g., sources of noise, missing values), planned approaches to ovecome these problems,
• Form of the knowledge expected to be derived from tha data,
• Possible machine learning algorithms that can be used.

Workshop (Nov. 22, 2012 - Dec. 27, 2012)

During the workshop, the students are expected to present the initial results of their analysis. Each presentation should include information about the goal of the work, related parameters, statistics about these parameters, including characteristics, missing feature values, noise in the data. Results of the initial experiements should be presented using public domain data mining tools. Finally, the further analysis planned should be described. Each presentation is expected to last about 40 minutes. We will have in class discussion about the presentation for about 10 minutes. The workshop grade will be based on parameters such as data collection, data cleaning, techniques used, tools used, interpretation of results and timing.

Final Report (Due: Jan. 10, 2013, 5:00 PM.)

The final report summarizes your term project. Although there are no rules about the format, it should have two parts. The first part is the cover page, which includes the title of the project, your name, the name of the course and the date of submission. The scond part is the report itself. The report should be written in a way that it can be submitted to a academic journal or conference. That is, it should have sections such as abstract, introduction, conclusion and references. The main criteria for the evaluation of your report are listed below:

• Organization, readability, methodology, completeness
• The domain of the data, features, their domain, the way they are collected
• The objective of the project
• Quality of the data; data analysis and preperation steps
• Tools used
• Statistical analysis and machine learning methods used
• Results: Knowledge extracted from the data, plausability, validity, interestingness

Grading Policy: Grades

• Participation: 10%
• Presentation, Term Project: 40%
• Report, Term Project: 50%

Lecture Slides:

• Chapter 1
• Chapter 2
• Chapter 3
• Chapter 4
• Rob Schapire's slides
• Chapter 5
• Interestingness
• Chapter 6
• Chapter 7

Free Packages:

• KNIME
• The R Project
• Kickstarting R
• simpleR - Using R for Introductory Statistics
• RapidMiner
• Weka
• RuleQuest