A spatio-temporal LSTM model to forecast across multiple temporal and spatial scales

This paper presents a novel spatio-temporal LSTM (SPATIAL) architecture for time series forecasting applied to environmental datasets. The framework was three different ocean datasets: current speed, temperature, and dissolved oxygen. Network implementation proceeded in two directions that are nominally separated but connected as part of natural system – across the spatial (between individual sensors) temporal dimensions sensor data. Data from twenty sensors were used train model. Results compared against four baseline models: machine learning algorithms generated by robust autoML frameworks, deep neural networks based on CNN LSTM, respectively. demonstrated ability accurately replicate complex signals provide comparable performance state-of-the-art benchmarks. Learning multiple simultaneously increased robustness missing addresses fundamental challenges related applications learning: 1) data sparsity, particularly challenging environment, 2) datasets inherently while classical ML approaches only consider one these at time. Furthermore, sharing parameters all input steps makes SPATIAL fast, scalable, easily-parameterized framework. • A spatiotemporal forecasting. We present experimental results AutoML models. Describes processing learning.