The Trend Prediction Of Red Tide Biomass In Zhejiang Coastal Ship-data Based On Deep Learning

The red tide occurs frequently in Zhejiang coastal waters, which causes water environment ecological system imbalance, aquaculture disaster and even harmful to our human health. So, the research of how to effectively predict the trend of red tide biomass is a difficult problem we face, at the same time it is of great importance for research significance in complex and sensitive Zhejiang coastal waters. The whole process of the growth, reproduction, extinction of the red tide algal has some connection with environmental factors. The correlation is complex, and the environmental factors have the time-series characteristics, which brings difficulties for the prediciton of red tide biomass. The neural network method has been widely used in the trend prediction of red tide biomass. Deep learning is a deep neural network, which successfully solves the shortcomings of the traditional neural network. The main research contents and results are as follows:(1) In this paper, a new non-linear deep learning model CRBM-DBN, which combines Conditional Restricted Boltzmann Machines(CRBM) and Deep Belief Networks(DBN), is constructed to predict the red tide biomass in Zhejiang coastal waters, processes the input vectors with Gaussian distribution, samples with Contrastive Divergence and effectively improves the accuracy and fitting of the red tide prediction, compared with the traditional neural network.(2) The optimal values including the depth of CRBM network and training parameters are solved based on Particle Swarm Optimization(PSO) algorithm. The experiment adopts ship observation data in Zhejinag coastal waters as input data training CRBM-DBN model, and compares our model to the classical deep learning model and the shadow model. The results show that the CRBM-DBN model is superior to predict the red tide biomass with less prediction error, higher fitting and prediction accuracy, and effective to predict the trend of red tide biomass in the next four frame time.(3) With multi-source, heterogeneous, massive red tide monitoring data in Zhejiang coastal, this paper combines GIS spatial database engine with Oracle database, integrates the prediction model constructed in this paper, and develops a platform of the red tide prediction and warning in Zhejiang coastal waters.The research results show that the CRBM-DBN model constructed in this paper has a better prediction results in Zhejiang coastal waters, which demonstrates the feasibility and practicability of our model, and has a certain reference value to the prediction and warning of the red tide in Zhejiang coastal waters.