An effective method for imbalanced time series classification: Hybrid sampling

Publication Type:
Conference Proceeding
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), 2013, 8272 LNAI pp. 374 - 385
Issue Date:
Filename Description Size
Thumbnail2013001549OK.pdf235.2 kB
Adobe PDF
Full metadata record
Most traditional supervised classification learning algorithms are ineffective for highly imbalanced time series classification, which has received considerably less attention than imbalanced data problems in data mining and machine learning research. Bagging is one of the most effective ensemble learning methods, yet it has drawbacks on highly imbalanced data. Sampling methods are considered to be effective to tackle highly imbalanced data problem, but both over-sampling and under-sampling have disadvantages; thus it is unclear which sampling schema will improve the performance of bagging predictor for solving highly imbalanced time series classification problems. This paper has addressed the limitations of existing techniques of the over-sampling and under-sampling, and proposes a new approach, hybrid sampling technique to enhance bagging, for solving these challenging problems. Comparing this new approach with previous approaches, over-sampling, SPO and under-sampling with various learning algorithms on benchmark data-sets, the experimental results demonstrate that this proposed new approach is able to dramatically improve on the performance of previous approaches. Statistical tests, Friedman test and Post-hoc Nemenyi test are used to draw valid conclusions. © Springer International Publishing 2013.
Please use this identifier to cite or link to this item: