Feature expansion of single dimensional time series data for machine learning classification

In this paper, we propose a feature expansion approach for the lowest one-dimension (1-D) time series data classification problems, where the expanded features include temporal, frequency, and statistical characteristics. We show that the proposed feature expansion can improve the classification accuracy compared to conventional machine learning algorithms for data classification. This is because the expanded features enable classifiers to consider multiple dimensions which are not feasible for low dimension data. Experiment results show that the proposed feature expansion method can improve the classification performance compared to conventional machine learning algorithms for 1-D actual biosensor data.