Research On Construction Methods Of Object Flow Patterns In Spatial Region

With the popularization and application of spatio-temporal data acquisition equipment,a large amount of object positional data is created,which is typical big data.The positional data contained in moving regularity can be used effectively to improve urban planning,intelligent transportation systems,urban safety management,etc.The positional data of moving object can be reflected by the object to departure or arrival a place at a certain moment,such as the credit card records of the public transportation,the record of the taxi and the stay point of the trajectory data,etc.At the same time,through analyzing the positional data of departure and arrival data,we can find the moving object regularity,which can be expressed as the regional flow model.Usually people use the method of qualitative description flow pattern of a region,such as residential areas leave eraly and return late,but commercial areas are different.However,there should be such a kind of demand,which requires the quantitative description of the moving objects in a given region,and the quantitative regularity can be predicted by the numerical prediction of the moving objects in the region.On the other hand,the quantitative flow regularity can be more precise to describe the regional characteristics.According to the real demand,we study the methods for constructing spatial object flow patterns in regions.The flow regularity of the region is expressed by the time sequence,and the model is summarizing by the number of arrival and departure objects in a certain region over a period of time.The object flow pattern is the dynamic property of the spatial region,and the different spatial regions have different flow patterns,and a spatial region can show multiple modes.The object flow pattern is a description of the moving object regularity in the region,which is different from the real time monitoring.At the same time,compared with the qualitative description of moving regularity,this paper presents the flow model is quantitative,that is,the spatial region scope is more flexible and the time granularity is finer.Due to the randomness of object moving,to find patterns with the high prediction precision is an important challenge.Such as patterns constructing,patterns training and patterns evaluation.For constructing patterns,we proposed a model for constructing the object flow patterns including data discretization and serialization,pattern training and evaluation and so on.For patterns training,we also present a method for removing abnormal sequences and selecting of the patterns automatically which improves the prediction precision.Using real datasets we performed the experimental evaluations of the model training method,the experimental results show that the created flow patterns can be used to express the personal flow pattern.With experiments and analysis of the proposed algorithms,the result of the proposed training model method compared with the existing training methods has higher prediction accuracy.Moreover,the effectiveness of the algorithm is verified.