Constructive Induction for Classifying Time Series

We present a method of constructive induction aimed at learning tasks involving multivariate time series data. Using metafeatures, the scope of attribute-value learning is expanded to domains that contain instances that have some kind of recurring substructure, such as strokes in handwriting recognition, or local maxima in time series data. These substructures are used to construct attributes. Metafeatures are applied to two real-world domains: sign language recognition and ECG classification. Using a very generic set of metafeatures we are able to generate classifiers that are either comprehensible or accurate, producing results that are comparable to hand-crafted preprocessing and comparable to human experts.