The primary objective of our research is to solve challenging Bioinformatics, Geoinformation Science and Computational Finance problems by developing and integrating distributed and parallel data mining and statistical learning technology for an efficient knowledge discovery at large sequence, spatial, temporal and spatial-temporal databases.
Data Mining in Bio Medical Sciences
- Predictive Modeling of Patient State and Therapy Optimization
- Bioinformatics, Medical Informatics
Data Mining in Geo Sciences
- Discriminative Modeling Framework for Mining of Spatio-Temporal Data in Remote Sensing
- Auxiliary System Sensor Fusion
- Spatial and Spatial Temporal
- Parallel and Distributed Data Mining
Data Mining in Social Sciences
- Center on Intersystem Regulation by Drugs of Abuse
- Time Series Analysis for Financial Model
- General Machine Learning Topics
- Hybrid Knowledge Systems: Integrating Dynamic Learning and Expert Knowledge
Neural Networks Theory: Optimization and Application
- Neural Networks Application to ATM Networks Control
- Computing with Nonmonotone Multivalued Neurons
Zoran Obradovic, who has been with the College since 2000, is Director of the Center for Data Analytics and Biomedical Informatics and Professor of Computer and Information Sciences. Obradovic's research focuses on developing data mining and statistical learning methods for knowledge discovery in large databases.
Obradovic is currently working on one project supported by the Defense Advanced Research Projects Agency (DARPA), several projects supported by the National Science Foundation (NSF) and by National Institute of Health (NIH).
One of Obradovic's NIH project is a continuation of his most well-known work. Obradovic was part of a team of biochemists and computer scientists who overturned a long-accepted paradigm of protein structure and function. Previously, it had been thought that "disordered" proteins - proteins in one of two forms thought to be derivatives of the true form - were not functional. Obradovic's team, by using data mining techniques on existing data, showed that the "disordered" forms of proteins are in fact normal and functional, in many cases. Furthermore, they showed that some proteins, and even parts of proteins, can move from one form to another and retain functionality. Obradovic is currently focused on better characterizing the structure and functionality of proteins in different forms; his DisProt program that predicts disordered protein regions was rated best model at three consecutive Critical Assessments of Structure Prediction meetings from 2004-2006.
One of the NSF projects seeks to make better use of atmospheric data collected from multiple sensors of different types. NASA's Terra and Aqua satellites and its AERONET ground detector network are collecting massive amounts of atmospheric data of several types and in several formats. Obradovic is working on methods for making the information collected more useful by improving the integration, accuracy and retrievability of the data. The project has implications beyond better analysis of existing atmospheric data, as the problem of dealing with large datasets consisting of several different kinds of related but un-linked information is increasingly common.
The highly collaborative Center for Data Analytics and Biomedical Informatics currently includes research-active, eight tenure-track faculty, including Obradovic, and 29 PhD students. Its mission is advanced research and education aimed at solving challenging data mining, machine learning, pattern recognition and optimization problems related to efficient knowledge discovery in large databases.
Obradovic received the 2009 Temple University Faculty Research Award and the College of Science and Technology's Dean's Distinguished Award for Excellence in Research in recognition of his work. He has published over 200 articles on data mining and is on the editorial board of six journals. Obradovic earned his PhD at Pennsylvania State University in 1991.