Multivariate times series classification through an interpretable representation

Multivariate time series classification is a machine learning task with increasing importance due to the proliferation of information sources in different domains (economy, health, energy, crops, etc.). Univariate methods lack ability capture relationships between variables that compose multivariate and therefore cannot be directly extrapolated environments. Despite good performance competitive results proposals published date, they are hard interpret their high complexity. In this paper, we propose method based on an alternative representation series, composed set 41 descriptive features, order improve interpretability obtained. Our proposal uses traditional classifiers over extracted features look for form series. We have selected four state-of-the-art algorithms as base evaluate our method. tested complete University East Anglia repository, obtaining highly interpretable capable explaining achieving statistically indistinguishable from best state-of-the-art.