栉孔扇贝夏季大规模死亡的神经网络（ANN）预测模型

根据2002和2003年对山东荣成桑沟湾栉孔扇贝养殖海区的水温、盐度、pH、氨氮浓度、亚硝氮浓度等环境因子和扇贝血清中的蛋白浓度、酸性磷酸酶活力、碱性磷酸酶活力、超氧化物歧化酶活力和过氧化氢酶活力等免疫学指标及栉孔扇贝养殖密度和死亡率的监测数据，运用人工神经网络（ANN）的原理和误差反相传播（BP）网络的方法，利用MATLAB软件初步建立养殖栉孔扇贝夏季大规模死亡的BP人工神经网络预测模型。预测模型经过300次的学习训练，误差平方和由67.46下降至0.0091。该预测模型对未参与模型构建的样本预测的结果与实际监测结果的符合率达到87.5％。本研究首次将人工神经网络与水产动物病害死亡的预测相结合，建立的预测模型具有对数据适应能力强，可适时学习，预测结果准确等突出优点，为水产养殖动物病害死亡程度的预测提供了一个新的研究方法。

Prediction Model (ANN) on Massive Death of Scallop Chlamys farreri in Summer

A model was developed primarily for the prediction on massive death of cultured scallops Chlamys farreri in summer using the principle of Artificial Neural Network (ANN) and the method of Back Propagation (BP) network by MATLAB software. Data were acquired from a two-year (2002 & 2003) investigation on temperature, salinity, pH, NH3-N, NO2-N of seawater and protein concentration, acid phosphatase activity, alkaline phosphatase activity, superioxide dismutase activity, catalase activity of scallops serum in Sanggou Bay, Rongcheng, Shandong province. After 300 times of studies and trainings, the sum-squared error of the prediction model decreased from 67.46 to 0.0091. The model was tested, and the prediction accuracy was 87.5%. It is the first time that ANN was used in aquaculture. The present study presented a new way for disease prediction and control of aquaculture animals.