论文部分内容阅读
为了提高智利竹筴鱼渔场预报水平和满足渔业捕捞生产的需要,利用2002—2008年的东南太平洋公海海域捕捞的中国大型拖网渔船共计15艘的生产统计资料,以及海洋环境数据(包括海表温度、叶绿素a浓度、表温距平、叶绿素a浓度距平、海表温度梯度强度和海面高度异常等数据),基于CART的算法,构建了智利竹筴鱼渔场决策树预报模型。用含1 114条记录的数据集对模型进行训练,并采用ROC方法对该模型诊断中心渔场的准确性进行了分析。最后将该模型应用于2009年各月份的智利竹筴鱼中心渔场预报,并与实际渔场位置进行了对比,结果显示预报渔场与实际生产位置基本一致,表明利用CART决策树方法建立智利竹筴鱼渔场预报模型是可行的。
In order to improve the forecast of fishing ground for Chilean catfish and to meet the needs of fishery production, a total of 15 large-scale Chinese trawlers fishing in the high seas of the Southeast Pacific from 2002 to 2008 were used to produce production statistics and marine environmental data including sea surface temperature , Chlorophyll a concentration, surface temperature anomaly, chlorophyll a concentration anomaly, sea surface temperature gradient intensity and sea surface height anomaly). Based on the CART algorithm, a decision tree prediction model of the Chilean catfish fishing ground was constructed. The model was trained with a dataset containing 1 114 records, and the accuracy of the model diagnostic center fishery was analyzed by ROC method. Finally, the model was applied to forecast the fishing ground of the Chilean Penaeus monocytogenes in each month of 2009 and compared with the actual fishing ground. The results show that the predicted fishing ground and the actual production position are basically the same, indicating that the use of CART decision tree method to establish Chilean mackerel Fisheries forecasting model is feasible.