The Application Of Neural Network Algorithm To HAB Predicting System

In recent years, Harmful Algal Blooms (HAB) break out frequently and constantly expand in scale, which leads to great damage to the living environment of all marine organisms and disturbs the ecological balance, resulting in huge economic losses to fishery resources and aquiculture. In addition, human's lives are also threatened by cyanobacterial toxins caused by HAB. Therefore, HAB predicting system with high accuracy is in great need. The paper gives a preliminary study on predicting modeling of HAB based on NNA (Neural Network Algorithm) which is getting mature these years, and then designs and realizes the HAB predicting system based on B/S mode. Specifically speaking, the paper includes the following aspects:Firstly, the model for HAB prediction using the BP algorithm with momentum term based on PCA algorithm are estabilished, based on the brief analysis of the relationship between the environmental factors and the HAB. The causes for HAB are complex and are closely related to several environmental factors, which lead to difficult choices of the model's inputs. To solve this problem, the model adopts PCA (Principal Component Analysis) algorithm to extract the useful information, which is extracting the principal components by giving linear transformation to the large input data matrix. Then, the principal components extracted are regarded as the inputs of the neural network model. As a result, the complexity of the neural network model is effectively reduced and the convergence speed is greatly improved. The simulation results show that the model has better fitting accuracy and predicting results comparing with the traditional model which adopt BP algorithm with momentum term.Secondly, the prediction model for HAB at Yantai Sishili Bay based on LMBP Algorithm is established though the analysis of causes for the HAB of the bay. The model decides the inputs by designing the experimental groups and comparing the results, that is designing different groups of experiments according to the different influence coursed by different inputs and then choosing the inputs according to the experimental results. The paper also describes the procedure of establishing a predicting model in detail with LMBP algorithm. The problems, such as how to decide the number of hidden neurons, how to improve the prediction accuracy of neural network model with the best utilization of the available data, are discussed. The entire process of modeling is clear and understanding, which gives new solutions for HAB predicting modeling with neural network. In addition, LMBP algorithm is the improvement of the BP algorithm with the advantages of faster convergence and higher accuracy of approaching. So, there is great superiority and feasibility of applying the LMBP model to the real-time forecast of HAB, and the LMBP algorithm can be a new solution for HAB prediction.Finally, the HAB predicting system based on B/S mode is realized with several neural network algorithms as the core module. A JSP website is established based on LMAT (Linux+MySQL+Apache+Tomcat) construction envirement. The system can display the prediction results in the forms of histograms and line charts. So, the visibility of the predicting results is greatly improved. The B/S mode, which allows the users to query the monitoring data of marine envirement, to analyse and to obtain the predicting results of the model just through the explorer, also improves the system in utility. Since the system integrated sevel neural network algorithms, it provides a new thought of algorithms for HAB predicting system.