With the popularization and application of spatiotemporal data acquisition equipment, a large number of object positional data is created, which is typical big data.The positional data of departure and arrival can reflect the flow regularity of moving object, which can be expressed as a regional flow model. Moreover, regional flow models can be used to improve urban planning, intelligent transportation systems, etc.We analyze the methods for constructing object flow patterns in spatial regions. Due to the randomness of object moving, to find patterns with high prediction precision is a big challenge. We therefore propose a model for constructing object flow patterns including data discretization and serialization, pattern training and evaluation and so on, which can quantitatively represent the regional flow regularity as time sequences. We also present a new hierarchical clustering tree with skewness. Based on the skewness, we design a method for removing abnormal sequences and selecting the patterns automatically, which improves the prediction precisionof patterns. Experimental results on real datasets show that the proposed flow patterns can be used to express regional flow patterns, and the proposed pattern training method has higher prediction accuracy compared with the existing ones.
