Study On Shipping Status Distribution Of Anomaly Detection Method Based On Data Mining

With the advent of the 5G era and the rise of the Internet of Things,massive amounts of data is coming,and the amount of data from Automatic Identification System(AIS)is increasing.At the same time,along with the development strategy of the Belt and Road,the complex data presents more challenges for marine supervision department.In addition,with the complex marine environment,there are various risks and hidden dangers,so the safety of vessel navigation has become an urgent problem to be solved.Therefore,when the abnormal behavior is discovered at the first time with the alarm,it will greatly reduce the harm of abnormal behavior and ensure the safety of navigation.With the rapid growth of the number of large vessels,abnormal identification,not only for vessel operators,shipping supervisors,but also for the whole shipping industry,is a real problem.How to solve these problems,AIS data came into being.Based on the existing AIS data,data mining technology is used to identify abnormal vessels based on the characteristics of the Beibu Gulf.The research content and innovation point are as follows:1)According to the multi-state parameter machine learning method,based on the longitude and latitude position of the vessel trajectory,the course over ground(COG)and the speed over ground(SOG),the constraint model is established by extracting multiple features of the vessel.The improved Density-Based Spatial Clustering of Applications with Noise(DBSCAN)algorithm is combined with the isolation forest(i Forest)algorithm to obtain the outlier points under the normal distribution of vessels.2)A method for the identification and analysis of abnormal behaviors of vessels under multi-vessel environment is proposed.Most of the recognition and detection algorithms has been proposed and implemented adopt the line identification of single vessel track points,which is,the modeling of single vessel motion behavior,without fully considering the state distribution under multiple vessels and the interaction between vessels.In order to avoid collision,crossing and other dangerous behaviors,the position and distribution of vessels exist certain laws.The core of this paper is to treat the vessel as a particle,then detect outliers points of the collective state distribution and behavior of the vessels,without the constructing a complex model,combined with the i Forest algorithm,the outliers are first screened out.The experiment was introduced into the Beibu Gulf to analyze the status distribution of multiple vessels and complete the abnormal detection.3)Study the improved DBSCAN algorithm.In view of DBSCAN's poor clustering effect and difficult selection of parameter Eps for uneven density,large clustering distance,the internal longitude and latitude relationship of AIS data was used to construct adjustable threshold to improve parameter selection,and obtain this parameter through data driving.After the improvement,the ship's position(longitude and latitude)and speed and heading characteristics are respectively clustered,so as to find the ship with abnormal position and speed.The effect of the improved DBSCAN algorithm was verified in the AIS dataset of the Beibu Gulf.4)Analyze abnormal points in the experiment based on geographical knowledge.Considering the characteristics of Beibu Gulf area from the distribution of ports and surrounding anchorages,the paper focuses on the identification of coastal high-speed areas and land ship points,and verifies the feasibility and effectiveness of the experiment in the Beibu Gulf under the interpretability of machine learning.5)The label of anomaly definition based on major project "Research and Application Demonstration of Key Technologies of Smart Vessel Networking in the Beibu Gulf" is used to verify the detection effect.The accuracy rate reached 91.73%,which is better than other mainstream measurement methods.The detection effect of vessel outliers and vessel speed anomalies are significant.The purpose of this paper is to perceive the maritime traffic situation and identify abnormal vessels through the analysis of AIS data,so that the maritime department and shipping company can find the abnormal as soon as possible,and give early warning and quickly troubleshoot the vessels.By combining the improved DBSCAN and i Forest algorithm,anomalies can be quickly identified through shipping status distribution in the Beibu Gulf.The method is simple,with few parameter adjustment,high evaluation index and good robustness..