大西洋热带海域长鳍金枪鱼渔场预报模型的比较

为提高大西洋热带海域长鳍金枪鱼(Thunnus alalunga)渔场预报的准确率,对K最近邻(k nearest neighbor,KNN)、逻辑斯蒂回归(logistic regression,LR)、决策与分类树(classfication and regression tree,CART)、梯度提升决策树(gradient boosting decision tree,GBDT)、随机森林(random forest,RF)、支持向量机(support vector machine,SVM)和Stacking集成(stacking ensemble learning,STK)共7个模型的预报性能进行了对比分析。该7个模型利用2016～2019年在大西洋公海海域(19°16′S～16°21′N;46°27′W～2°09′E)作业的13艘中国远洋延绳钓渔船的渔业数据,结合0～500 m不同水层的温度、盐度、溶解氧、叶绿素a浓度、海表面风速、涡动能和混合层深度数据建立。各模型取75%数据作为训练数据,25%为测试数据,采用预报准确率(accuracy,ACC)与接受者操作特征曲线下面...

COMPARISON ON FISHING GROUND FORECAST MODELS OF THUNNUS ALALUNGA IN THE TROPICAL WATERS OF ATLANTIC OCEAN<

To improve the accuracy of the forecast model for albacore tuna (Thunnus alalunga) fishing ground in the tropical waters of Atlantic Ocean, seven fishery forecast models, e.g. k-nearest neighbor (KNN), logistic regression (LR), classfication and regression tree (CART), support vector machine (SVM), random forest (RF), gradient boosting decision tree (GBDT), and stacking ensemble learning (STK) model were used and compared based on the data of 13 tuna longliners of Chinese fishing enterprises from 2016 to 2019 in the high seas of the Atlantic Ocean (19°16''S~16°21''N; 46°27''W~2°09''E). Using environmental factors (temperature, salinity and dissolved oxygen) at different water layers from 0 to 500 m, as well as chlorophyll-a concentration, sea surface wind speed, eddy kinetic energy, and mixed layer depth, the relationship between albacore tuna CPUE and the environmental factors were analyzed. Seventy-five percent of the data were taken as training data and 25% as test data. The performance of each model was evaluated by prediction accuracy (ACC) and area under receiver operating characteristic curve (AUC). Relationships between CPUE (catch per unit of effort) and marine environmental factors were established. Results show that: (1) the prediction performance of STK model was obviously better compared with other models and its ACC and AUC is 75.92% and 0.742, respectively; (2) the areas of central fishing ground predicted by STK model for albacore tuna is consistent with the actual fishing ground generally; (3) the marine environmental factors that affect the distribution of albacore tuna fishing grounds in the Atlantic Ocean included mainly temperature and salinity of 100 m layer, and dissolved oxygen at 100, 150, and 500 m layer. The accuracy and the prediction performance of the STK model is high for albacore tuna fishing ground forecast in the tropical waters of Atlantic Ocean.