CN-121998445-A - Building group development intensity and municipal matched facility capacity matching analysis method

CN121998445ACN 121998445 ACN121998445 ACN 121998445ACN-121998445-A

Abstract

The invention provides a building group development intensity and municipal supporting facility capacity matching analysis method which comprises the steps of carrying out standardized cleaning, time-space alignment and feature extraction on multi-source data, carrying out development behavior trend prediction based on a time convolution network, realizing key time sequence reinforcement by combining an attention mechanism, introducing dynamic confidence gating shunt straight-through and correction flow, carrying out buffer correction on low confidence prediction by utilizing a generated countermeasure network, integrating an improved fuzzy judgment model, integrating municipal facility capacity indexes, realizing regional supply and demand pressure grading early warning, having the capability of on-line feedback learning and gating parameter self-adaption adjustment, improving prediction accuracy, stability and facility bearing risk pre-judgment effect, and supporting intelligent city planning decision.

Inventors

WANG JIANTAO
CHEN BIN
JIANG BO
Cui Pinqi
GU SIYING
Liang Donggan
MING XIAOHUA

Assignees

广州市白云城市建设投资有限公司

Dates

Publication Date: 20260508
Application Date: 20251205

Claims (10)

1. The method for analyzing the matching of the development intensity of the building group and the capacity of municipal matched facilities is characterized by comprising the following steps: s1, collecting multi-source heterogeneous data of a city area, and carrying out standardized cleaning and space-time alignment preprocessing on the multi-source heterogeneous data to generate a development behavior prediction basic data set; S2, constructing a development behavior trend prediction sub-model based on a time convolution network architecture, inputting a development behavior prediction basic data set, extracting space association features and time evolution rules through a multi-scale expansion convolution layer, and outputting a building increment probability distribution map and a corresponding space unit confidence scoring matrix; s3, constructing a dynamic confidence level gating unit, and based on the space unit confidence level scoring matrix, adopting a piecewise linear function to establish a mapping relation between a confidence level threshold and a shunt weight, and generating a dynamic allocation strategy for developing the strength through flow and the correction flow; S4, designing a reconstruction-resistant buffer correction module, receiving low-confidence correction stream data, constructing a training set by utilizing historical development tracks of similar plots in the region, reconstructing local features of low-confidence prediction by generating an countermeasure network architecture, and outputting corrected development intensity estimated values and residual error compensation coefficients thereof; And S5, establishing an improved fuzzy comprehensive judgment matching degree fusion layer, carrying out weighted combination on the development intensity straight-through flow and the corrected development intensity estimated value according to a confidence degree weight, constructing a supply and demand matching degree feature vector by combining a municipal facility capacity measuring and calculating result, calculating a facility bearing pressure index through a fuzzy membership function, and generating a grading early warning signal.
2. The method for analyzing the matching of the development intensity of the building group and the capacity of the municipal supporting facility according to claim 1, wherein the step S5 further comprises: And S6, deploying an online feedback learning mechanism, periodically collecting actual development progress data and initial prediction deviation values, updating time attention weight entropy value calculation parameters of the development behavior trend prediction sub-model, and optimizing a dynamic adjustment rule of the confidence coefficient threshold value based on deviation distribution characteristics.
3. The method for analyzing the matching of the building group development intensity and the municipal supporting facility capacity according to claim 1, wherein the step S1 specifically comprises: Based on a database and a public data platform of a city planning management department, historical land yielding data in a city area are obtained, and a historical record table of land level development behaviors is formed; Carrying out structural extraction on the regional development planning text to generate a structural planning element table; collecting macro economic index data, and performing standardized normalization processing on the macro economic index data to obtain a time sequence input feature vector reflecting the overall development situation of the city; Performing data cleaning operation on the historical record list of the land parcel level development behaviors, the structural planning element list and the time sequence input characteristic vector reflecting the overall development situation of the city to obtain cleaned land parcel data; And carrying out space coordinate alignment on the cleaned land block data based on a geographic information system platform, and carrying out time sequence alignment on historical yielding data and planning text analysis results in the cleaned land block data by utilizing a time stamp field to generate a multi-source heterogeneous data set with a unified time-space reference.
4. The method for analyzing the matching of the building group development intensity and the municipal supporting facility capacity according to claim 3, wherein the data cleaning operation comprises filling of missing values, abnormal value elimination, repeated record de-duplication and field format unification, and filling of missing parts in time sequence data is performed by adopting a sliding window interpolation method.
5. The method for analyzing the matching of the building group development intensity and the municipal supporting facility capacity according to claim 1, wherein the step S2 specifically comprises: Based on the generated development behavior prediction basic data set, constructing a development behavior trend prediction sub-model framework by adopting a time convolution network, and constructing a space-time joint characteristic input tensor; Performing multi-scale expansion convolution operation on the space-time joint feature input tensor, and extracting neighborhood correlation features between a multi-scale evolution rule of development behaviors in a time dimension and space units by setting convolution kernel stacking layers with different expansion coefficients to obtain high-dimensional space-time feature embedded representation; Performing time attention mechanism weighted fusion processing on the high-dimensional space-time feature embedded representation, and calculating an attention weight matrix based on similarity among feature vectors of each time step to obtain attention-enhanced time sequence feature representation; based on the time sequence characteristic representation of the attention enhancement, generating building increment probability distribution of each space unit in the future by adopting a fully-connected network, and obtaining a probability distribution matrix of the development intensity of each space unit in the future and a time attention weight matrix corresponding to the probability distribution matrix; based on the entropy value of the time attention weight matrix and the historical return testing deviation data, a space unit confidence degree scoring model is constructed, confidence degree scores of development strength prediction results of all the space units are calculated, and a development behavior prediction result matrix containing confidence degree information is output.
6. The method for analyzing the matching of the building group development intensity and the municipal supporting facility capacity according to claim 1, wherein the step S3 specifically comprises: normalizing the space unit confidence score matrix output by the trend predictor model to obtain a normalized confidence matrix; Based on historical prediction return deviation distribution data, determining an initial value range of a confidence coefficient threshold value by adopting a statistical analysis method, dividing a confidence coefficient level interval by combining regional development stage characteristics, and generating a piecewise linear function mapping rule table; Performing confidence coefficient-shunt weight mapping calculation on each space unit in the standardized confidence coefficient matrix according to the piecewise linear function mapping rule table to generate a shunt weight matrix in the space dimension; Based on the distribution weight matrix, performing dynamic distribution operation on the development intensity predicted value to form two development intensity output streams with different processing paths; and executing channel identification marking processing on the branched through flow and the corrected flow output to generate a development intensity prediction data packet of the channel attribute tag.
7. The method for matching and analyzing the development intensity of the building group with the capacity of municipal supporting facilities according to claim 6, wherein the step S3 further comprises the steps that a dynamic confidence degree gating unit sets a confidence degree threshold according to historical prediction return deviation and development stage characteristics by means of the space unit confidence degree scores, a split weight is calculated for each space unit through a piecewise linear function, predicted values higher than the threshold are classified into straight-through flows, classified into correction flows lower than the threshold, and channel attribute labels are respectively added.
8. The method for analyzing the matching of the building group development intensity and the municipal supporting facility capacity according to claim 1, wherein the step S4 specifically comprises: Constructing a training set of low-confidence prediction samples based on historical development track data of similar plots in the region; based on the generation of the countermeasure network architecture design generator model, low-confidence development intensity prediction data are input, local development mode features are extracted by using a convolutional neural network, and development intensity correction candidate diagrams with regional consistency are generated; Constructing a discriminator model, performing countermeasure training on the development intensity correction candidate graph and the real historical development data, and optimizing the spatial feature distribution output by the generator by minimizing the identification accuracy of the discriminator on the generated sample; calculating a residual compensation coefficient based on the residual between the development intensity correction candidate map and the original low-confidence prediction; And carrying out fusion processing on the development intensity correction candidate graph and the residual error compensation coefficient, outputting a standardized corrected development intensity estimated value, and transmitting the standardized corrected development intensity estimated value to a matching degree fusion layer.
9. The method for analyzing the matching of the building group development intensity and the municipal supporting facility capacity according to claim 1, wherein the step S5 specifically comprises: Carrying out confidence coefficient normalization processing on the output development intensity straight-through flow and the output correction flow, constructing a weighted fusion coefficient matrix based on confidence coefficient scores of all space units, and outputting a fused development intensity feature map; Based on water supply, power supply, drainage and traffic bearing capacity data provided by the municipal facility capacity measuring and calculating module, constructing a standardized municipal facility capacity feature vector, and performing space alignment and attribute mapping with the development intensity feature map to form a regional supply and demand matching degree feature vector set; Performing fuzzification processing on the regional supply and demand matching degree feature vector set by adopting an improved fuzzy comprehensive evaluation model, setting a multi-level fuzzy membership function based on facility types, calculating a bearing pressure fuzzy membership matrix of each facility type, and outputting a fuzzy evaluation set of facility bearing states; based on the maximum membership principle in the fuzzy comprehensive judgment result, combining with a preset facility bearing pressure grade dividing standard, carrying out grading judgment on the municipal facility bearing state of each space unit, and generating a bearing pressure grade label matrix; And carrying out joint analysis on the bearing pressure grade label matrix and the space unit development strength prediction confidence score, dynamically correcting the early warning grade of the low confidence region based on a confidence weighted grading rule, generating a final facility bearing pressure grading early warning signal diagram, and outputting the final facility bearing pressure grading early warning signal diagram to a visual display module.
10. The method for analyzing the matching of the building group development intensity and the municipal supporting facility capacity according to claim 1, wherein the online feedback learning mechanism periodically collects and cleans the actual development progress, calculates a prediction error time sequence, models uncertainty distribution by adopting a sliding window method, dynamically adjusts attention weight entropy parameters and confidence threshold self-adaptive adjustment rules of a time convolution network, and feeds back correction residual errors to a buffer correction module to optimize training loss function weights.

Description

Building group development intensity and municipal matched facility capacity matching analysis method Technical Field The invention relates to the technical field of urban development supply and demand matching analysis and model robustness optimization, in particular to a matching analysis method for building group development intensity and municipal supporting facility capacity. Background At present, supply and demand matching analysis of urban building development intensity and municipal supporting facility capacity belongs to the leading research direction in the field of urban planning and intelligent evaluation. With the acceleration of the urban process, a method of combining the development demand calculation and the infrastructure capacity calculation in the early stage of land yielding is generally adopted in all places to guide the urban space structure layout and the municipal facility configuration. The main flow technical system mostly adopts multi-source heterogeneous data support such as historical land yielding, planning text, macroscopic economic indexes and the like, quantitatively predicts future development intensity through a rolling prediction or machine learning model, is linked with a municipal facility capacity model, calculates supply and demand matching degree or facility bearing pressure in an area, outputs grading risk early warning, and provides data support for planning, yielding and construction decision. Meanwhile, trend prediction analysis, multi-source data automatic processing and intelligent matching evaluation methods based on deep learning are explored step by step in the industry, and the precision, usability and foresight of the model are improved. However, in practical application, the following prominent problems generally exist in the prior art: (1) The development behavior prediction model and the facility matching degree evaluation module are mostly of strong coupling and integrated feedforward structures. The output of the development intensity prediction is directly used for matching degree calculation, and flexible error isolation and compensation means are lacked among models. Once the prediction part has local errors, abnormal fluctuation or inaccurate data extrapolation, the evaluation result is easily amplified or distorted, the overall robustness of the model is reduced, and even the final early warning decision is influenced; (2) Most current methods do not deal well with prediction uncertainty. The mainstream model only depends on the fitting goodness or the reliability of the single confidence index judgment result, and an adaptive mechanism cannot be formed to dynamically adjust the influence weight of the predicted value on the matching evaluation. This directly results in unstable model output and expanding error accumulation in situations where development activities are expected to fluctuate strongly, data distribution deviates from statistical assumptions or uncertainty rises, impairing system look-ahead and practicality; (3) Some technologies attempt to introduce feedback learning to correct prediction bias, but are often limited to periodic data backtracking, so that it is difficult to timely perceive real-time changes of model stability, and quick and fine-grained error correction cannot be performed on low-confidence regions. The lack of split-flow grading processing and local feature reconstruction capability makes it difficult for the model to combine overall trend grasping and local extremum correction, and obvious short plates exist in a high-dynamic and heterogeneous city development scene. Disclosure of Invention The invention aims to solve the technical problems and provides a matching analysis method for building group development strength and municipal supporting facility capacity. The technical scheme of the invention is realized in that the building group development strength and municipal matched facility capacity matching analysis method comprises the following steps: S1, collecting historical land yielding data, regional development planning text and macro economic indexes of an urban region, and carrying out standardized cleaning and space-time alignment preprocessing on the multi-source heterogeneous data to generate a development behavior prediction basic data set; S2, constructing a development behavior trend predictor model based on a Time Convolutional Network (TCN) architecture, inputting the multisource data preprocessed by the S1, extracting space correlation features and a time evolution rule through a multiscale expansion convolutional layer, and outputting a building increment probability distribution diagram of 3-5 years in the future and a corresponding space unit confidence score matrix; S3, constructing a dynamic confidence coefficient gating unit, and based on the confidence coefficient scoring matrix generated in the S2, adopting a piecewise linear function to establish a mapping relation between a confidenc