Prediction of chlorophyll‐a data based on triple‐stage attention recurrent neural network

Abstract Marine Internet of Things (IOT) is the use technology to connect various sensing devices at sea, so as integrate maritime information and realize monitoring systematic management complex data sea. The marine environment changeable, disasters occur frequently, such red tides. Due sudden destructive nature tide, it plays a pivotal role monitor occurrence tide for IoT, where machine learning has been widely used predict However, they were rarely able catch change chlorophyll‐a, which important practical significance predicting tide. In order deal with above problems, this paper proposes triple‐stage attention‐based recurrent neural network, can enhance representation ability input sequences, selectively capture dynamic spatial correlations between multi‐channel observations in sequence, meanwhile adaptively capturing temporal different time intervals sequence. results show that method outperforms state‐of‐art baseline methods here.