基于FCM-ANFIS模型的船舶交通流预测  

Vessel Traffic Flow Prediction Based on FCM-ANFIS Model

作　　者：惠子刚[1] 肖建兵[1] HUI Zigang;XIAO Jianbing(Department of Navigation,Qingdao Ocean Shipping Mariners College,Qingdao 266071,China)

机构地区：[1]青岛远洋船员职业学院航海系,山东青岛266071

出　　处：《中国航海》2018年第2期73-76,101,共5页Navigation of China

基　　金：中国远洋海运集团有限公司2017科研计划项目(2017-H-011)

摘　　要：提出一种基于模糊C均值聚类自适应神经模糊推理系统(Fuzzy C-Means Clustering Algorithm with Adaptive Neural-Fuzzy Inference System,FCM-ANFIS)的船舶海上交通流预测模型,采用相关分析法预测网络模型的输入变量个数;采用模糊C均值聚类算法对仿真数据进行分析和模糊聚类,从而确定模型的聚类中心,进而建立ANFIS预测网络结构。该模型采用BP算法与最小二乘算法相结合的混合学习算法进行网络参数训练,可克服传统预报模型收敛速度慢和局部最优等问题,进而提高预报精度。同时,选用船舶自动识别系统(Automatic Identification System,AIS)记录的船舶交通流数据进行仿真预报,仿真结果验证了该模型的可行性和有效性,并取得良好的效果。An accurate ship traffic flow prediction approach is proposed based on FCM-ANFIS（ Fuzzy C-Means Clustering Algorithm with Adaptive Neural-Fuzzy Inference System）. The input parameter set of prediction model is defined through correlation analysis. The FCM is introduced to perform clustering of fuzzy data. The hybrid algorithm combining BP algorithm and least squares estimation algorithm is used for training to achieve better rate of convergence and to ensure global optimization. The prediction model is verified with data from AIS（ Automatic Identification System）. Simulation results show that the proposed method can predict coastal ship traffic with high accuracy and satisfactory stability.

关 键 词：船舶交通流预测 模糊C均值聚类 自适应模糊推理系统 相关分析 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] U692[自动化与计算机技术—控制科学与工程]