Research On The Prediction Model Of Urban Human Flow Based On Multi-source Data

Nowadays,the world is in the process of exponential expansion.The large-scale growth of urban population brings not only scale gain,but also many global challenges.In the face of the great transformation period of the city,the research on the spatial distribution of the urban population develops with the deepening of the understanding of the city.In this process,artificial intelligence technology of computer science is involved in the field of planning,which provides a powerful analysis tool for all aspects of research.The research of urban population distribution changes from static population to dynamic flow,from macro coarse particle size to micro fine particle size,from simple function fitting to machine learning and prediction.In this context,on the basis of summarizing the relevant research on urban population distribution,this paper uses multi-source data such as mobile phone signaling to establish an indicator system based on three categories: spatial form,location potential and functional business.The location model based on SGD algorithm(two ideas of multiple linear regression analysis and multiple nonlinear regression analysis)and the prediction based on xgboost algorithm are constructed respectively The model is used to predict the density of people flow in Nanjing.Combined with the prediction results of the two models,the mechanism research,problem summary and application prospect are further carried out from the dimensions of hierarchical clustering characteristics,distribution of people flow frequency,time curve of people flow and application prospect of the model.Specifically,the structure of this paper is as follows:The first chapter is the introduction,which introduces the research background,research significance,research content and research objectives of the urban flow density prediction,and summarizes the related research methods and the technical route of the full text.The second chapter is the research review and theoretical framework.In the research review,the relevant researches on the spatial distribution of urban population by scholars since the 19 th century are sorted into three types: the location model based on urban center,the gravity model based on interaction,and the algorithm model based on data learning.Combined with the previous research results,the theoretical framework of this paper is put forward,and the internal mechanism of urban people flow is summarized into three categories: spatial form,location form and functional business form.The third chapter is the model preprocessing,which defines the research scope,spatial unit and basic data information,and fills in the trajectory of the signaling data of the mobile phone to meet the prediction target of the traffic density.The index system is refined,and a hierarchical system of 3 categories,6 middle categories and 36 small categories is constructed.The definition of error and accuracy is adjusted according to the data characteristics.Finally,the paper analyzes the current spatial distribution of the density of people flow in Nanjing.The fourth chapter is empirical research,from three aspects of space,location and function to explore a reasonable location model.Based on correlation analysis and index selection,this paper constructs the prediction model of urban human flow from two ideas of multiple linear regression model and multiple nonlinear regression model respectively,and optimizes the parameter solution process of large-scale samples by using stochastic gradient descent(SGD).Finally,two regression models are evaluated according to the prediction accuracy,and the limitations and experience of location model in the study of people flow prediction are summarized.In the fifth chapter,in order to solve the problem of uneven distribution of the actual population density in space,the top level of the model is designed as C4.5 algorithm decision tree,the control samples are classified into two categories: 0 and 1,and xgboost algorithm(extreme gradient)is adopted Boosting)targeted learning is carried out for each branch category,and finally a prediction model of the flow density corresponding to the underlying characteristics is formed.The prediction model based on xgboost algorithm has good accuracy in the global and high value range of the test set,and the prediction performance is greatly improved compared with the location model.The sixth chapter is the representation and mechanism analysis.Combined with the prediction results of location model and algorithm model,it discusses the complexity of the internal mechanism of urban human flow,and makes further mechanism research,problem summary and application prospect from multiple dimensions such as hierarchical clustering characteristics,distribution of human flow frequency,time curve of human flow and application prospect of the model.The seventh chapter is the summary of the research.On the basis of the full text,it summarizes the shortcomings of the research and the prospect of the development direction.This paper consists of 81665 words,70 illustrations and 29 tables.