An Event-Triggered Deep Belief Network for Forecasting Coastal Water-Quality

Forecasting coastal water-quality is an indispensable step to prevent the marine pollution. In this article, we design an event-triggered deep belief network (ET-DBN) to forecast coastal water-quality. In theory, the main contribution is ET-DBN model, which is trained by the event-triggered algorithm with a selective parameter update mechanism. In practice, the main contribution is that ET-DBN model can accurately approximate the non-stationary characteristics of coastal water-quality via the event-triggered strategy on the spatio-temporal data. Finally, the experimental results show that the designed ET-DBN can adapt to the non-stationarity characteristics of coastal water-quality and achieve the accurate forecasting performance.