Affiliations: [a] School of Mathematics and Information Science and Technology, Hebei Normal University of Science and Technology, Qinhuangdao, Hebei, China | [b] Yanching Institute of Technology, Langfang, Hebei, China
Correspondence:
[*]
Corresponding author: Yan Kang, School of Mathematics and Information Science and Technology, Hebei Normal University of Science and Technology, Qinhuangdao, Hebei 066000, China. E-mail: kangyan_1983@163.com.
Abstract: In order to accurately predict noctiluca density, a new prediction model PSO-GRNN was constructed according to the relationship between noctiluca density and environmental factors. Particle swarm optimization (PSO) algorithm was used to obtain the optimal smoothing factor of generalized regression neural network (GRNN) in this model, that could reduce human subjective influence and improve the generalization ability and performance of generalized regression neural network. The dissolved oxygen, water temperature, total nitrogen, salinity, phytoplankton density and soluble inorganic phosphorus were taken as the model input, and the noctiluca density was taken as the model output. Finally, the new model and other traditional models were both tested, and the prediction results of the improved model were compared with other traditional models. The experimental results showed that the improved new model had high accuracy in the prediction of noctiluca density, and could indirectly realize the early prediction of red tide.
Keywords: Noctiluca density prediction, PSO, GRNN, Smoothing factor
