LID 2015 - Special Session on Learning from Imbalanced Data

Machine learning techniques have shown tremendous progress in recent years, which has allowed it become commonly used in the real world. Many techniques have been introduced to discover different representations of knowledge from data in numerous fields. It is in this context that the importance of certain problems that some researchers were beginning to glimpse is of paramount importance. One of such problem is the imbalanced data, where one class contains much smaller number of examples than the remaining classes. The imbalanced distribution of classes constitutes a difficulty for standard learning algorithms and calls for specialized approaches. This problem is extensive in many real-world applications: fraud detection, risk management, face recognition, text classification, and many others. The aim of this special session is to provide a forum for international researchers and practitioners to present and share their original works addressing the new challenges, research issues and novel solutions in imbalanced data. Topics of interest include but not limited to:
Sampling techniques for imbalanced data
High dimensional and class-imbalanced data
Ensembles for imbalanced data
Pre-processing, structuring and organizing complex data
Imbalanced classes in noisy environments
Skewed data and difficult classes
Imbalanced data for regression
Imbalanced data and semi-supervised learning
Imbalanced in multi-class problems
Performance evaluation of classifiers in imbalanced domains
Handling class imbalance by modifying inductive bias and post-processing of learned models
Theoretical aspects of constructing combined imbalanced learning systems
Imbalanced learning in changing environments
Incremental online learning algorithms
Cost-sensitive learning
Real applications