Dynamic prediction method and device for urban fine population distribution based on depth learning

Abstract

The invention discloses a dynamic prediction method of urban population refinement distribution based on depth learning, which comprises the following steps: 1, grid processing is carried out on the research area under the scale required by the research, the sample set is constructed, and the corresponding training set and test set are divided; 2, construct a prediction model of urban population distribution, and training that prediction model accord to historical data; 3, input that data of the distribution state of the urban population at the first j times as an input variable into the trainmodel in the step 2, and predicting and obtaining the distribution state of the urban population at a certain time in the future; The prediction model can predict all the regions divided by the urbangrid simultaneously. Convolution long-short memory network extracts spatial features through convolution structure, and acquires temporal features through long-short memory network structure.The bottom layer of the algorithm combines convolution neural network with long-short memory network to learn the high-dimensional spatio-temporal features of population distribution changes, and effectivelyfuses the spatial dimensions of time dimension, so as to improve the prediction accuracy.

Description

technical field [0001] The invention relates to the fields of spatio-temporal big data mining and spatio-temporal analysis and modeling, in particular to a method for dynamic prediction of refined urban population distribution. Background technique [0002] Urban population distribution refers to the spatial distribution of population in a city at a certain moment. my country's economy is developing rapidly, industrialization and urbanization are accelerating, and the population of large cities is constantly gathering, resulting in environmental degradation, traffic congestion, and increasingly serious public safety hazards, which have brought severe challenges to modern urban management and development. Mastering the urban population distribution and development dynamics at a fine spatio-temporal scale can provide important scientific basis and timely and effective services for the exploration of urban residents' travel activities, the optimization of urban public resource ...

AI Technical Summary

Problems solved by technology

[0014] (1) Most of the existing studies on population distribution prediction focus on the problem of population development model at a large scale, and there are limitations in the prediction of population distribution at a fine scale;

[0015] (2) Most of the existing methods only consider the dependence of population distribution change in the time dimension or the correlation of space dimension, but do not consider the temporal-spatial coupling characteristics of population distribution change at the same time. Predicting the population distribution in blocks at a fine scale not only fails to take into account The spatio-temporal correlation characteristics between different blocks, and the prediction efficiency is also seriously affected;

[0016] (3) Existing methods rarely consider the impact of external factors such as working days, legal holidays, meteorological conditions, and major events on the urban population distribution prediction model

Images