Workshop on Class Imbalances: Past, Present, Future
(CIPPF'2012) is organized in conjunction with the
11th International Conference on Machine Learning and Applications ICMLA-2012
Boca Raton Marriott Hotel, Boca Raton, Florida USA
December 12-15, 2012

Submission deadline - extended till August 20th

Workshop Chairs : Nathalie Japkowicz, Jerzy Stefanowski and Nitesh Chawla


Many real-world applications have revealed difficulties in learning from imbalanced data, where at least one of the target classes contains a much smaller number of examples than the other classes. The class imbalance problem occurs in such domains as: fraud/intrusion detection, risk management, medical data analysis, technical diagnostics/monitoring, image recognition, text categorization or information filtering. Class imbalances constitute a difficulty for most learning algorithms and as a result many classifiers are biased toward the majority classes and fail to recognize examples of the minority class. The challenging issue in learning from imbalanced data has received growing research interest in the last decade and a number of specialized methods have already been proposed. However due to the inherent complex characteristics of imbalanced data sets, new methods and studies on the fundamental properties of this problem are still needed. This motivates us to organize the workshop, which could be a forum for discussing current trends and recent advances in learning from imbalanced data as well as new promising research directions.

Suggested topics include (but are not limited to) the following aspects of learning from imbalanced data:
- Sampling techniques (over-, under- or hybrid)
- Modifying the inductive bias of learning algorithms
- Transforming data distributions in ensembles dedicated for class imbalances
- Using simulated data to tackle or study the class imbalance problem
- Detecting and adapting to distributional shift and concept drift in evolving imbalanced data
- Semi-supervised and active learning from imbalanced data
- Applications in fields such as medicine, technical diagnostics, text processing, economy, bio-informatics
- Evaluation challenges arising with class imbalanced data, such as when injecting synthetic data in the training set (the testing set distributions are then materially different)

We encourage submissions of long or short research papers (6 or 4 pages) as well as extended abstracts of work in progress (2 pages). The papers must be submitted in the form of PDF files and should conform to the IEEE CS Press Conference Paper Format

More information at the web link