User Tools

Site Tools


Action unknown: copypageplugin__copy
projects:year6:6a.022.ul

6a.022.UL - Mitigating Concept Drift for Time-Varying Domains Through Adaptive Learning

Project - Summary

Adaptive techniques for concept drift detection are a necessary topic of exploration, as it is guaranteed that as time passes, underlying conditions will shift. As these shifts occur, they can cause major impacts on predictions for the future both short and long term. Some examples of drifting conditions are:

  1. People seeking to hoodwink anomaly detection systems in order to commit fraud.
  2. People deliberately exploiting discovered knowledge for hostile purposes, i.e. discoveries fount via hotspot analysis and association mining.
  3. A sudden increase/decrease in the number of patients in hospitals in a specific region
  4. A sudden increase/decrease in the spending of clients in a specific area or type of purchase item

Typically, changes to underlying conditions are rarely advertised and it is unknown when the impact of such changes will be felt. One of the more difficult portions of this research is determining when and how to update knowledge, in a computationally inexpensive manner. This is due to the “expiration date” that exists on all learned data.

Project - Team

Team Member Role Email Phone Number Academic Site/IAB
Jian Chen PI Not available Not available UL Lafayette
Jennifer Lavergne PI Not available Not available UL Lafayette
Raju Gottumukkala Researcher raju@louisiana.edu (337) 482-0632 UL Lafayette
Satya Katragadda Researcher satya@louisiana.edu (337) 482-0625 UL Lafayette
Adeola Olaleye Siwoku Graduate Student c00301223@louisiana.edu (337) 735-0219 - Cell UL Lafayette
Ryan Benton Researcher rbenton@southalabama.edu (251) 460-6298 University of South Alabama
Tom Johnsten Researcher tjohnsten@southalabama.edu (251) 461-1599 University of South Alabama
Mike Lucito Project Mentor Not available Not available Schumacher Clinical Partners

Project - Deliverables

Deliverables
1 Algorithm and prototypical system for the prediction of emerging and declining trends in spatio-temporal data.
2 Algorithm and prototypical system for the discovery of concept drift in order to update the prediction model using adaptive learning.
3 Algorithm and prototypical system for a prediction outlook of trends across time, based upon given IAB data and/or in-house datasets.
4 Project report and case studies using our current datasets and any suggested by the IAB.

Project - Presentation Video

Project - Documents

projects/year6/6a.022.ul.txt · Last modified: 2021/06/02 15:52 by sally.johnson