L.: Classification of sparse time series via supervised matrix factorization

Data sparsity is an emerging real-world problem ob-served in a various domains ranging from sensor net-works to medical diagnosis. Consecutively, numerous machine learning methods were modeled to treat miss-ing values. Nevertheless, sparsity, defined as missing segments, has not been thoroughly investigated in the context of time-series classification. We propose a novel principle for classifying time series, which in contrast to existing approaches, avoids reconstructing the miss-ing segments in time series and operates solely on the observed ones. Based on the proposed principle, we de-velop a method that prevents adding noise that incurs during the reconstruction of the original time series. Our method adapts supervised matrix factorization by pro-jecting time series in a latent space through stochastic learning. Furthermore the projected data is built in a su-pervised fashion via a logistic regression. Abundant ex-periments on a large collection of 37 data sets demon-strate the superiority of our method, which in the ma-jority of cases outperforms a set of baselines that do not follow our proposed principle.