Deep Learning for Haemodialysis Time Series Classification

In this paper, we propose a deep learning approach to deal with time series classification, in the domain of haemodialysis. Specifically, we have tested two different architectures: a Convolutional Neural Network, which is particularly suitable for time series data, due to its ability to model local dependencies that may exist between adjacent data points; and a convolutional autoencoder, adopted to learn deep features from the time series, followed by a neural network classifier. Our experiments have proved the feasibility of the approach, which has outperformed more classical techniques, based on the Discrete Cosine Transform and on the Discrete Fourier Transform for features extraction, and on Support Vector Machines for classification.