Semantic Mining Based Urban Functional Areas Recognition

As the basic urban planning and construction unit,the urban functional area describes the urban geographical spatial structure and functional pattern and reflects the expansion and change of the city.Therefore,it is essential to grasp the function of urban zoning and management accurately.The purpose of this study is to extract and identify urban functional areas on a small scale by considering the context of Point of Interest(POI)data in geospatial space and combining POI data with the word2 vec language model.Firstly,on the premise of using POI proximity information,a functional corpus of the urban interior is built based on the search radius.The POI data and traffic analysis zone(TAZ)is regarded as word documents in natural language.Secondly,by implementing the Skip-gram model in word2 vec,the semantic information of POI is mapped into a high-dimensional space vector.Then the TAZ semantic vector representing the functional features of partition is calculated.Finally,the TAZ is clustered based on the K-means algorithm,and the functional areas are identified according to the distribution characteristics of POI in the cluster.The main conclusions are as follows:.Compared with the results obtained by TF-IDF,LDA and BERT models,the word2 vec model has the highest accuracy rate(69%),followed by the BERT model,which is also based on POI semantic information does not contain spatial location information(63%).The LDA model based on POI frequency(56%)and TF-IDF model(52%)are relatively poor in urban functional area recognition.In this paper,we combine the semantic similarity of POI based on a hierarchy with mean average precision(MAP)and achieve better results in evaluating the semantic accuracy of POI types.Compared with mean reciprocal rank(MRR),the MAP can use more hierarchical information.The semantic vector and negative sampling parameter dimension significantly influence the accuracy of the semantic vector.This study sets larger semantic vector dimensions,and smaller negative sampling parameters can obtain more accurate POI semantic representation.In this paper,the research framework of combining POI data with word2 vec model can make full use of the spatial location and proximity information of POI data,effectively capture the potential semantic relationship between POI data in geospatial,and provide a reference for the identification of urban functional areas and other city research based on POI semantics.This paper effectively combines GIS technology,natural language processing technology and POI data,makes full use of the spatial information and neighborhood information of geographic data,and excavates the potential semantic relationship between POI data,which provides a way of thinking and framework for identifying urban functional areas.The research results can provide a reference for the analysis and Research of urban spatial layout.25 figures,9 tables and 109 references are included in this dissertation.