摘要
针对船舶横摇运动时序的小样本、非线性、随机性等特点,提出了一种改进支持向量机(improved support vectormachine,ISVM),采用鲁棒损失函数和小波核函数可以有效压制横摇时序的多种噪音和奇异点,具有良好的鲁棒性及泛化能力;引入单松弛变量使得ISVM具有更简洁的对偶问题及约减的寻优范围,减小了算法运行的时间.建立基于ISVM的船舶横摇运动姿态实时预报模型,对某船横摇运动姿态进行了预报,仿真结果表明该模型是行之有效的.
In considering the time series of ship rolling which displays many special characteristics,including a small sample,nonlinearity,and randomness,this paper proposes an improved support vector machine（ISVM） which utilized the robust loss function and the Marr wavelet kernel function.The ISVM could effectively deal with various kinds of noises and outliers of a rolling time series and had good robustness and generalization ability.Also,using a single relaxation variable allowed the ISVM to possess a concise dual problem,smaller optimal range,and faster run time than a standard support vector machine.Finally,a real-time prediction model showing ship rolling movement attitude was designed to forecast the time series of ships based on the ISVM.Simulation results are provided to validate the effectiveness of the proposed scheme.
出处
《哈尔滨工程大学学报》
EI
CAS
CSCD
北大核心
2011年第5期607-612,共6页
Journal of Harbin Engineering University
基金
国家自然科学基金资助项目(51079033)
关键词
船舶横摇运动
改进支持向量机
Marr小波核
鲁棒损失函数
实时预报
ship rolling movement attitude
improved support vector regression
Marr wavelet kernel
robust loss function
real-time forecasting
