基于空间自回归和空间聚类的渔情预报模型  

A fishery forecasting model based on a spatial auto-regressive model and spatial clustering

作　　者：袁红春[1] 谭明星[1] 顾怡婷[1] 陈新军[2] 

机构地区：[1]上海海洋大学信息学院,上海201306 [2]上海海洋大学海洋科学学院,上海201306

出　　处：《海洋科学》2015年第12期165-172,共8页Marine Sciences

基　　金：上海市教育委员会科研创新重点项目(12ZZ162);上海市科学技术委员会科技支撑项目(12510502000;14391901400)

摘　　要：作者针对远洋渔场渔情预报精度偏低的问题,提出一种基于空间自回归和空间聚类的渔情预报模型。该模型利用空间自回归对收集到的渔业历史数据进行预处理,然后通过空间聚类将所有数据样本根据地理位置分划成若干个区域,最后研究每个区域中环境数据与渔获数据之间的数学关系,各自建立栖息地适宜性指数模型(Habitat Suitability Index,HSI),并以印度洋大眼金枪鱼(Thunnus obesus)为例进行验证。结果表明,本模型的均方差为0.1742,与传统线性回归方法的均方差0.2363相比,能更好地表达海洋环境数据与渔获量之间的关系,预测精度显著提高。In order to improve the predictive accuracy of pelagic fishing grounds, we have proposed a fishery-forecasting model based on a spatial autoregressive model and spatial clustering. In our model, the spatial autoregressive method is employed to first preprocess historical fishery data. Using the spatial clustering method, all data samples are then divided into several regions based on their geographical locations. By analyzing the mathematical relationships between environmental data and fishing data in the same region, a habitat suitability index model was built, with a follow-up experiment on bigeye tuna（Thunnus obesus） in the Indian Ocean. The results of the experiment showed that compared with the mean square error of 0.2363 in a traditional linear regression method, the model proposed in this paper had a mean square error of 0.1742. Therefore, our model can better demonstrate the relationship between marine environmental data and fishing quantity, and the predictive accuracy has been significantly improved.

关 键 词：大眼金枪鱼(Thunnus obesus) 渔情预报 空间自回归 空间聚类 非线性回归 

分 类 号：S934[农业科学—渔业资源]