Data-driven Behavior Patterns Mining For White Stork

As an indispensable part of maintaining the ecological balance of nature,wild animals play a vital role in the survival and development of human beings.With the continuous development of sensor technology and wireless network technology and the gradual maturity of data mining technology,it has created unprecedented convenient conditions for wildlife research and protection.This thesis takes the endangered species of white stork as the research object,mainly from the four aspects of staying behavior,migration behavior,community behavior and migration state to study its behavior patterns.(1)In the aspect of trajectory stay point analysis,it aims at the existing problems of large amount of data calculation and high time complexity.Trajectory stay point analysis algorithm based on trajectory compression and continuous time-space distance clustering is proposed,which uses trajectory compression and considers the spatio-temporal continuity between trajectory points,only calculates the distance between time adjacent position points.Finally,the space-time distance was combined with compactness and separation to evaluate the clustering results of each algorithm.The results show that the proposed algorithm can effectively identify the stork's stay area while reducing the amount of data calculation,and the distance within the class is the minimum,the distance between,classes is the maximum.(2)In the aspect of frequent trajectories mining,aiming at the problems of different order of magnitude of trajectory similarity measurement methods and the clustering process complicated,A frequent trajectory mining model based on complex network clustering is proposed,which combines the similarity measurement method of the order of magnitude with complex networks,and on the real movement trajectories to validate the feasibility and validity of the model.The results show that the frequent track mining model,which combines the track similarity measure method with the complex network,can effectively identify the frequent track of white stork population,and the clustering process is more convenient and efficient.(3)In the aspect of social network research,in view of the one-sided problem of using discrete location relations to represent the social strength between moving objects,The trajectory segment similarity value between the white stork individuals represents the strength of the relations between them,and the visual analysis software is used to study its social network relationship graph.The result shows that the method of measure the relations strength with trajectory segment instead of local location can depict social network relations more accurate and reliable,and the network structure is more stable.(4)In the aspect of stork migration status prediction,aiming at the problems of low data resolution and low prediction accuracy,by using high-resolution data sets,on the basis of accurately extracting feature information with high influence on migration state from historical data,adding future environmental feature data,test experiments are carried out on a variety of prediction models,and finally use the accuracy rate to evaluate the prediction effects of each model.The experimental results show that the addition of future environmental features can generally improve the prediction accuracy of each model,and the prediction effect is stable.