Adapting ELM to Time Series Classification: A Novel Diversified Top-k Shapelets Extraction Method

摘要

ELM (Extreme Learning Machine) is a single hidden layer feed-forward network,where the weights between input and hidden layer are initialized randomly. ELMis efficient due to its utilization of the analytical approach to computeweights between hidden and output layer. However, ELM still fails to output thesemantic classification outcome. To address such limitation, in this paper, wepropose a diversified top-k shapelets transform framework, where the shapeletsare the subsequences i.e., the best representative and interpretative featuresof each class. As we identified, the most challenge problems are how to extractthe best k shapelets in original candidate sets and how to automaticallydetermine the k value. Specifically, we first define the similar shapelets anddiversified top-k shapelets to construct diversity shapelets graph. Then, anovel diversity graph based top-k shapelets extraction algorithm named as\textbf{DivTopkshapelets}\ is proposed to search top-k diversified shapelets.Finally, we propose a shapelets transformed ELM algorithm named as\textbf{DivShapELM} to automatically determine the k value, which is furtherutilized for time series classification. The experimental results over publicdata sets demonstrate that the proposed approach significantly outperformstraditional ELM algorithm in terms of effectiveness and efficiency.