基于多任务多核学习的蓝藻爆发预测模型  

A Multi-Task Multiple Kernel Learning Model for Cyanobacterial Bloom Forecasting

作　　者：黄逸彬 蒋鹏 刘晓[1] HUANG Yibin;JIANG Peng;LIU Xiao(Department of Industrial Engineering and Management,School of Mechanical Engineering,Shanghai Jiao Tong University,Shanghai 200240,China)

机构地区：[1]上海交通大学机械与动力工程学院工业工程与管理系,上海200240

出　　处：《工业工程与管理》2021年第4期109-116,共8页Industrial Engineering and Management

基　　金：国家自然科学基金面上项目(71673188);中国博士后科学基金面上项目(2018M640397)。

摘　　要：相对于水表层蓝藻爆发预测,水体垂直方向的蓝藻分布预测能更好地辅助决策者管理水体。传统数据驱动的蓝藻爆发预测模型大多将不同水层的蓝藻爆发视作独立任务。该类模型忽略了来自多个水层蓝藻爆发预测任务间的关联性,从而导致其无法突破预测精度的瓶颈。为此,提出一种基于多任务多核学习的蓝藻爆发预测模型,并试图共享来自不同水层蓝藻爆发预测任务之间的信息来提升蓝藻爆发预测精度。实际案例研究表明所提出的模型能够利用来自不同水层监测数据中的共性知识更准确地预测蓝藻爆发,进而有利于缓解该类水环境灾害带来的影响。Compared to cyanobacterial bloom forecasting in the surface water layer,the forecasting along water columns could assist decision-makers in managing water bodies better.Traditional datadriven cyanobacterial bloom forecasting models tended to treat cyanobacteria forecasting at different water depths as independent tasks. Such models ignored relationshipsamong different water depths,which resulted in the bottleneck of forecasting accuracy. Amulti-task multiple kernel learning model was proposed to improve the cyanobacterial bloom forecasting by sharing information from different water depths.Computational results based on a real-world case study show that the proposed model can forecast cyanobacterial blooms more accurately through sharing knowledge among water depths and help to mitigate the impacts of such water environment disasters.

关 键 词：蓝藻爆发 预测建模 多任务学习 多核学习 信息共享 

分 类 号：C931.1[经济管理—管理学]