Analysis And Prediction Of Pirate Attack Hotspots Based On Spatio-temporal Characteristics

At present,Maritime piracy still seriously threatens the safety of maritime transportation,threatens the safety of personnel and property of ships at all times,and disrupts the normal international trade and maritime order,which has always been the focus of international shipping.Through previous research on piracy,it is found that the scope of pirate activities and historical activities are traceable,and its temporal and spatial distribution is not random.By studying the characteristics of pirate attacks in time and space,combined with the analysis of influencing factors,the reasons for the formation of pirate attack hotspots and the characteristics of hot spots can be further explained.At the same time,by identifying and extracting the spatiotemporal distribution characteristics of pirate activities,predicting the temporal and spatial distribution of pirate attack hotspots at the microscopic scale,and indepth understanding of the occurrence pattern and future development trend of pirate attacks,it is conducive to the International Maritime Organization and other relevant departments to rationally optimize anti-piracy resources,and formulate long-term and short-term anti-piracy plans to achieve effective prevention and combating of pirate attacks.This thesis first gives a basic introduction to the current situation of global piracy,takes the Strait of Malacca as the research area,and selects the pirate attacks from 2007 to 2022 as the research object,including the following research content:(1)Through the characteristic analysis of the temporal distribution and spatial distribution of pirate attacks in the study area,it is found that the pirate attack incidents in the study area have obvious aggregation in time and space,which proves the predictability of pirate attack incidents in the study area.On this basis,the temporal and spatial interaction of pirate attacks is analyzed,and the optimal temporal and spatial thresholds are calculated,which provides a theoretical basis for the analysis and prediction of subsequent pirate attack hot spots.(2)On the basis of spatio-temporal near repeat analysis,the causes and characteristics of pirate attack hotspots are analyzed by nuclear density estimation,case chain model,consistency analysis and other methods,and the structural characteristics of pirate attack hotspots are further revealed,which provides a theoretical basis for the microscopic spatiotemporal scale of spatiotemporal distribution prediction of pirate attack hotspots.(3)Through the construction of spatio-temporal risk matrix,the design of ConvLSTM network,and the evaluation method of prediction results,a spatio-temporal distribution prediction model of pirate attack hotspots based on the ConvLSTM model is established.The model can effectively identify and extract the spatial distribution characteristics of sparse matrix,and remember the dependence problem of long-term series,so as to predict the distribution of pirate attack hotspots in time and space,and comprehensively evaluate the accuracy of model prediction from two aspects: model fit and prediction result evaluation.The results show that the spatiotemporal distribution prediction model of pirate attack hotspots based on ConvLSTM model constructed in this thesis is due to the prediction model based on CNN network and the prediction model based on FC-LSTM.