Since the overall prediction error of a classifier on imbalanced problems can be potentially misleading and biased, it is commonly evaluated by measures such as G-mean and ROC (Receiver Operating Characteristic) curves. However, for many classifiers, the learning process is still largely driven by error based objective functions. As a result, there is clearly a gap between the measure according to which the classifier is to be evaluated and how the classifier is trained. This paper investigates the possibility of directly using the measure itself to search the hypothesis space to improve the performance of classifiers. Experimental results on three standard benchmark problems and a real-world problem show that the proposed method is effective in comparison with commonly used sampling techniques.
Swipe to navigate through the chapters of this book
Please log in to get access to this content
To get access to this content you need the following product:
- A Measure Oriented Training Scheme for Imbalanced Classification Problems
- Springer Berlin Heidelberg
- Sequence number
Neuer Inhalt/© ITandMEDIA