DATA MINING RESEARCH GROUP

Data Mining has been established as a vital technology of the New Economy. Made possible by the large amount of data already captured in computers (an amount which grows daily) and powered by the union of technologies ranging from Statistics to Artificial Intelligence to Databases, Data Mining is an indispensable tool for any organization that wants to be competitive in its market segment. By leveraging the amount of data collected, and extracting information from this data, organizations can make more informed decisions with respect to every aspect of their dynamics, from supplying to customer management to internal production.

The Data Mining group at the Computer Engineering and Computer Science Department of the Speed Scientific School, University of Louisville, was officially created on its first meeting on February 22, 2001. The creation of the group is not only an answer to the need for R&D on the area of Data Mining, but also a realization that relevant knowledge and experience in the area or related areas was already possessed by many of the faculty members of the department. The goals of the group are:

To leverage the individual knowledge and practical and research experience of its members to achieve results beyond those at the reach of individuals though collaboration and pooling of resources. \item To advance the state of the art in the theory and practice of Data Mining, establishing the department as a national center for research on the area.
To serve the community of high-tech companies in Louisville and the state of Kentucky by creating mutually beneficial relationships with any interested industry partner. The group will offer very flexible arrangements to industrial partners, ranging from simple sponsorships or temporary consulting services to support for long-term projects.

MEMBERS

At the present, the group counts as members Dr. Patricia Cerrito , Dr. Ahmed Desoky , Dr. Adel Elmaghraby , Dr. James Graham , Dr. Mehmet Kantardzic , Dr. Anup Kumar , Dr. Ramohan Ragade , and Dr. Antonio Badia . They are all faculty in the Computer Engineering and Computer Science department at the Speed Scientific School , University of Louisville , except Dr. Cerrito, who is with the Mathematics department in the College of Arts and Sciences. Together, they bring in expertise not only on data mining, but also in related areas like knowledge discovery, machine learning, neural networks, distributed systems, data visualization, data quality and databases. Together, they have published more than fifty scientific papers in data mining and relevant areas in international conferences and journals. Besides this academic activity, they have participated in several projects with local industry, hospitals and UofL School of Medicine (see a description of some projects below) and they bring more than twenty years of collective real-world experience.

Past Projects

The members of the group have considerable practical experience in applying Data Mining techniques to real problems, as well as experience in related areas. The following is just a summarized selections of some past projects of the members of the group.

Machine Learning for Drug Design This project, developed jointly with the Institute for Molecular Diversity and Drug Design at the University of Louisville, and the Department of Biostatistics and Medical Informations at the University of Wisconsin, involves the use of machine learning for structure-activity prediction in drug design. In its simplest form, structure activity prediction is an ideal example of an inductive machine learning problem. Active molecules comprise the set of positive examples, while those molecules with little, or no, activity comprise the set of negative examples. The machine learning task is to learn a structured description that will distinguish the positive examples from the negative ones. Concurrently there have been new advances in combinatorial chemistry, in which mixtures of thousands of chemical compounds can be synthesized and tested for bio-activity. Applying the machine learning, structure-activity prediction methodologies to the results of combinatorial chemistry experiments promises to be a major advance in pharmacophore identification and drug design.
Heuristic Search in Planning Sequences of Views for 3D Object Recognition, a contract with Space and Naval Warfare Systems Center, San Diego, CA. This ongoing project uses sequences of images with the objective of finding comprehensible rules and relations in data sets. Often these rules are abstractions and require the application of machine learning techniques such as decision tree learners, artificial neural networks and/or sparse correlation kernel analysis. Usually two steps are performed: Data Preprocessing (feature extraction) and Data Analysis (prediction, modeling and performing).
Improvements in Diagnostics using Data Mining Technologies. This project, joint with the H. Lee Moffitt Cancer Center and Research Institute of the University of South Florida, studied the way that Polycythemia Rubra Vera is diagnosed and discovered improved methods of diagnosis which use fewer (but more relevant) tests.
Model of Decision Making in a Selection of Laproscopic Methods, a joint project with the Department of Gynecology, University of Louisville.

Materials

Note: These tutorials were not made by the members of this group. The authors are indicated in the documents. They are here only as a reference and as a service to whoever visits this page.

A tutorial on Web Mining for Personalization from PKDD 2001 is here on PDF format.

A tutorial on Text Mining, also from PKDD 2001 is here on PDF format.

A tutorial on Data Mining for Bio-Informatics is here on PowerPoint format.

How to Contact Us

If you are interested in contacting the group, please send e-mail to abadia@louisville.edu or call 852-0478 working days from 9:30 am to 4:30 pm. Please do not hesitate to get in touch with us if you would like further information!