Study On The Detection Of Abnormal Vessel Behavior Based On Association Rules

In order to fully mining the hidden vessel behavior rules from the massive Automatic Identification System(AIS)data,this study comprehensively considers vessel’s position(longitude and latitude),Speed Over Ground(SOG),Course Over Ground(COG),time,vessel type,length and other factors that affect the vessel’s behavior.Then,use the association rule mining technology to connect the above factors in series,and find the objective rules of the normal vessel behavior contained in the AIS data,using this rules as a standard,we can detect the abnormal behavior of a small number of vessels in complex waters,and provide new ideas for ensuring the safety of vessels’ water navigation.This study uses historical AIS data,takes vessel behavior as the research object,combines the knowledge of vessel behavior,and improves the mining algorithm of association rules to study and analyze the problem of abnormal vessel behavior detection.The specific work content is as follows:(1)Combined with the data requirements of association rule mining technology,data preprocessing is performed on the historical AIS data of Laotieshan waters for one year,including data cleaning,data selection and data discretization,etc.,and the influence of vessel type,length and registry on vessel behavior is analyzed separately.On this basis,the factors affecting vessel behavior are divided into vessel motion factors(SOG,speed change,COG and heading change),navigation environment factors(vessel position,sailing season and sailing period)and vessel attribute elements(vessel type,length and vessel’s nationality),and a set of Multi-element Vessel Behavior Representation Model(MVBRM)suitable for association rule mining is proposed.(2)The association rule mining algorithm is used to mine the objective laws of vessel behavior hidden in the historical AIS data.Frequent pattern mining based on trajectory points,and the water is divided into several grid cells using grid division technology,then,using Frequent-Pattern Growth(FP-growth)algorithm to mine the vessel behavior knowledge contained in the trajectory points in each grid cell.Mining frequent sequential patterns based on trajectories,and recode all trajectories within the water.The improved FP-growth algorithm is used to mine the vessel behavior knowledge contained in the trajectories in the waters.Finally,mining experiments were carried out using AIS historical data in Laotieshan waters to verify the effectiveness and feasibility of the method.(3)A framework for detecting abnormal vessel behavior based on association rule mining method is constructed.This framework regards the excavated objective laws of vessel behavior as a standard of normal vessel behavior and establishes a library of normal vessel behavior rules,and the data to be tested is preprocessed,behavioral feature extraction and association rule mining are performed.Through the CBA classifier,the corresponding class association rules are extracted from the normal behavior rule base for rule matching,and the detection of abnormal vessel behavior is completed.Finally,import the test set data to verify the effectiveness of the method.The results show that the method has a 100% detection rate for known abnormal behaviors in a small amount of test set data in Laotieshan’s complex waters.The detection rate of a large number of test set data is affected by the proportion of abnormal behavior trajectories to normal trajectories.As the proportion of abnormal behavior trajectories increases,the detection rate gradually increases,and when the proportion of abnormal behavior trajectories is 20%,the detection rate reaches 92% and tends to be stable.The results show that the method proposed in this thesis can effectively detect the abnormal behavior of vessels in complex waters.