Abstract
A process for generating a unifying motif-based homogeneous feature vector representation is described and evaluated. The motivation was to determine the viability of this representation as a unifying representation for heterogeneous data classification. The focus for the work was cardiovascular disease classification. The reported evaluation indicates that the proposed unifying representation is a viable one, producing better classification results than when a Recurrent Neural Network (RNNs) was applied to just ECG time series data.