Application-Independent Feature Construction from Noisy Samples

When training classifiers, presence of noise can severely harm their performance. In this paper, we focus on “non-class” attribute noise and we consider how a frequent fault-tolerant (FFT) pattern mining task can be used to support noise-tolerant classification. Our method is based on an application independent strategy for feature construction based on the so-called δ-free patterns. Our experiments on noisy training data shows accuracy improvement when using the computed features instead of the original ones.

• Type: Conference
• Year: 2009
• Where: PAKDD
• Authors: Dominique Gay, Nazha Selmaoui, Jean-François Boulicaut
• Keywords: Attribute Noise, Data Mining, PAKDD 2009, Pattern Mining Task, So-called δ-free Patterns