Research And Design Of Mining Method For Moving Targets Activity

With the rapid development of the existing target tracking technology,the spatiotemporal information of moving objects such as pedestrians,vehicles and ships can be effectively recorded.These spatial and temporal data contain a large number of potentially valuable patterns and knowledge,which are of great value in urban planning,defense and military,location-based services and so on.In order to find the activity law of moving targets from mobile data,spatio-temporal co-occurrence method is used to analyze the co-occurrence characteristics of moving targets.On this basis,this paper analyzes the partial periodicity of co-occurrence model,and the research contents as follows.Aiming at the problem of how to extract effective patterns from spatio-temporal data,this paper puts forward a mining spatio-temporal co-occurrence patterns algorithm based on double network.Mining temporal co-occurrence patterns from spatio-temporal data set of moving targets needs to calcuate the spatio-temporal interest degree.In order to simplify the calculation method of spatio-temporal interest degree,this paper puts forward a modeling method of double layer spatio-temporal network based on the characteristics of spatio-temporal data.The spatio-temporal network can effectively preserve the temporal and spatial relations between the temporal and spatial objects and the temporal and spatial elements.When calculating the spatio-temporal interest degree,the time and space frequency can be calculated quickly only by reading the time series between the corresponding objects.Due to the preservation of the spatio-temporal relationships among the elements in the double layer spatio-temporal network,a large number of redundant candidate patterns can be reduced by the element network layer.Thus,the computationalcost is reduced and the memory space is reduced.Based on the time-space network,this paper proposes a spatio-temporal co-occurrence patterns mining algorithm,which takes the effective period of elements into account In the algorithm,the weighted eigenvalues are used to represent the spatio-temporal frequency of the pattern,and the pattern list is used to save the spatio-temporal co-occurrence pattern set The experimental results show that compared with the Celik's algorithm and Wang's algorithm,with the same data set and results,the proposed algorithm has higher running efficiency.A periodic co-occurrence pattern mining algorithm for moving targets is proposed to solve the periodic problems of moving objects in this paper.The activities of pedestrians,vehicles and boats moving targets are usually part of the cyclical,this will be part of a cyclical pattern of co-occurrence rules used in the moving target,and then this paper puts forward some periodic patterns of co-occurrence adaptive mining algorithm The algorithm adds the model validity according to the effective period of the element,and improves the computing method of the co-occurrence frequency.The algorithm is adaptive in the determination of the period span and the confidence parameter.Give a time frame based on the spatio-temporal data and initialization method to determine the span of adaptive,and on the basis of the maximum span and retain all partial periodic patterns of co-occurrence principle,gives the method of determining the parameters of adaptive confidence.In this paper,the partial periodic pattern analysis is used in the study of spatio-temporal co-occurrence patterns,and a mining algorithm of partial periodic spatio-temporal co-occurrence patterns is put forward.According to the effective period of elements,the pattern efficiency is used to improve the calculation of co-occurrence frequency.Based on the periodic prior property,this algorithm determines the partial periodicity of long patterns first and then considers partial periodicity of its subsets.Experiments show that this algorithm reduces the co-occurrence frequency calculation of spatio-temporal co-occurrence pattern set,and improves the efficiency of the algorithm compared with the Apriori-like algorithm and Naive algorithm.