基于主成分和BP神经网络的智利竹筴鱼渔场预报模型研究

Application of BP neural network based on principal component analysis in fishing grounds of chilean jack mackerel (Trachurus murphyi) in the southeast Pacific Ocean

东南太平洋智利竹筴鱼Trachurus murphyi是我国大型拖网渔船队的重要捕捞对象。准确预报中心渔场是提高渔业生产能力的重要工作。本文根据2003—2009年我国船队在东南太平洋海域捕捞智利竹筴鱼的渔捞日志数据,结合海洋遥感获得的海表温度(SST)和海面高度(SSH)等海洋环境因子,利用主成分和BP神经网络方法对智利竹筴鱼中心渔场预报模型进行了研究。研究利用主成分分析法(PCA)得到累计贡献率在90%以上样本的主成分,综合考虑模型测试的精度与速度,基于原始样本和经PCA处理后的主成分分别建立了BP模型,其最优BP模型结构分别为5∶10∶1和3∶7∶1。研究结果表明,经PCA处理后的主成分所建立的BP神经网络模型在训练结果和测试结果上均要优于用原始样本建立的BP神经网络模型,两者的预报准确率分别为67%和60%。

Chilean jack mackerel （Trachurus murphyi ）is an important target species for Chinese factory trawler fleet in the southeast Pacific Ocean,and the accurate forecasting of fishing ground can provide better scientific guid-ance for fishing operation.In this paper,we built the forecasting models by using the methods of principal compo-nent analysis （PCA）and BP neural networks according to the catch data from the logbooks and fishing yield statis-tics from Chinese factory trawler fleets,the sea surface temperature （SST）and sea surface height （SSH）obtained by satellite remote sensing from 2003 to 2009.Based on the PCA,we got the principal components of different fac-tors.We also determined the two suitable model structures by using the original-samples and PCA-processed-sam-ples combined with the accuracy of models,respectively.It is found that the model used by PCA-processed-sam-ples is better than that model used by original-sampled based on the results of training and test,and their accuracy rates were 67% and 60% respectively.