Mining Spatiotemporal Sub-prevalent Co-location Patterns Based On Star Model

With the development of information technology,spatial data is growing rapidly,and spatial data mining is becoming hot.Spatial co-location pattern mining is an important research area,and it is widely used in real life.Traditional co-location pattern mining only considers the feature instances that form clusters,and may ignore some important spatial relationships among features that are not clusters.For this reason,sub-prevalent colocation pattern proposed a star model and star instances to mine the important spatial relationships among features that do not form clusters.In spatial data,time is an important dimension.Spatial instances will appear or disappear with the change of time,and their position also changes over time.Therefore,time should be taken into account in co-location pattern mining.However,the existing sub-prevalent co-location pattern mining does not properly consider the time factor.Therefore,based on the star model,this paper studies spatiotemporal sub-prevalent colocation pattern mining,and the main work is as follows:1.By analyzing the spatial examples of its position changing with time,spatiotemporal sub-prevalent co-location pattern is studied.Firstly,analyzing how to properly introduce teme into sub-prevalent pattern mining.We measure spatiotemporal sub-prevalent pattern by temporal sub-frequency.Then,the downward inclusion property of temporal sub-frequency is proven,the search space is reduced by using the property.And we propose an effective algorithm,STSCP(SpatioTemporal Sub-prevalent Colocation Patterns).Finally,experiments on eight different data sets verify that STSCP has high efficiency under the influence of different factors such as number of instances,number of time slots,time sub-prevalence threshold,spatial sub-prevalence threshold.In addition,compared with PTBA(Prefix-Tree-Based Algorithm),the proposed STSCP can find the spatiotemporal sub-prevalent patterns with different occurrence frequencies according to time sub-prevalent threshold set by users.STSCP can mine patterns that are more in line with user needs and have more rich semantic and practical value.2.The time continuous sub-prevalent pattern based on star model is presented by considering the continuity of time.Firstly,the relevant concepts are given and the index of measuring spatiotemporal continuous sub-prevalent patterns is defined:spatiotemporal participation ratio and spatiotemporal participation index.Then,an algorithm is proposed,TCSCP(Time Continuous Sub-prevalent Co-location Patterns).Finally,experiments are carried out on three datasets of different sizes,and it is verified that the proposed TCSCP has good efficiency under the influence of different distance threshold,time threshold and spatiotemporal sub-prevalence threshold.In addition,compared with PTBA and STSCP,TCSCP mines the least number of patterns,which can narrow the range of patterns and more accurately find the patterns required by users.