Urban residential quarter housing occupancy rate prediction method, equipment and medium

Abstract

The invention discloses an urban residential quarter housing occupancy rate prediction method and device and a medium, and relates to the field of housing occupancy rate prediction.The method comprises the steps that an urban residential quarter housing occupancy rate estimation model is built based on urban residential quarter basic information and mobile phone signaling data, and correction is conducted through the actually-collected housing occupancy rate; according to the spatial range boundary, the terrain, the PM2.5 concentration, the POI, the urban road network, the NPP-VIIRS and the urban community house price of the to-be-predicted urban residential community in the set time period, determining characteristic index data, and aligning the spatio-temporal data to obtain processed characteristic index data; and through a mixed multi-head attention mechanism coupled with an efficient routing strategy and a long short-term memory network, constructing an urban residential quarter housing occupancy rate prediction model, inputting the processed characteristic index data, and determining a predicted value of the housing occupancy rate of the urban residential quarter to be predicted. According to the invention, accurate prediction of the housing occupancy rate of the urban residential quarter is realized.

Description

Technical Field

[0001] The present application relates to the field of housing occupancy rate prediction, and in particular to a method, device and medium for predicting housing occupancy rate in urban residential areas. Background Art

[0002] The occupancy rate of urban residential communities focuses on assessing and analyzing trends in the number of households and housing usage within different urban communities. In today's stock-of-stock era, it is not only crucial for governments and urban planners to optimize resource allocation and rational development, but also reveals the economic vitality, population migration trends, and social development potential of urban areas. It plays a vital role in modern urban management and development and serves as a key tool for urban management and development. The occupancy rate generally refers to the proportion of rooms or units that are occupied and used during a specific period of time, while the vacancy rate refers to the proportion of unoccupied and unused units. Currently, research on the occupancy rate of urban residential communities primarily focuses on the occupancy rate of public rental housing in the economic and social sectors and the vacancy rate of buildings in the construction sector.

[0003] In terms of research data, early studies used municipal tax records and basic household registration data to calculate actual housing occupancy rates and, in turn, actual vacancy rates. With the widespread use of nighttime light data, researchers have gradually incorporated this data into urban vacancy rate studies. Subsequently, some scholars have used 100-meter grid population data to improve the accuracy of occupancy rate calculations. As research continues to deepen, emerging geographic big data and municipal service industry data are increasingly being used to identify urban vacancy rates. Resource data such as water and electricity consumption are being used to quantify the degree of vacancy in urban housing at a fine-grained level. Street view data is being used to describe residents' daytime and nighttime behavior and identify the number of houses with lights on at night, thereby estimating actual occupancy rates. In terms of research methods, early researchers used traditional interviews and surveys to obtain information on actual housing occupancy. However, due to the lengthy and costly nature of surveys, the calculated occupancy rates were somewhat affected by the time and spatial scope of the surveys. With the development of machine learning and artificial intelligence, some scholars have begun leveraging existing survey data, demarcating residential areas using land use data, and combining them with remote sensing imagery, socioeconomic statistics, and points of interest to estimate the spatial distribution of urban housing vacancy rates using classic machine learning algorithms. Although the existing research on urban housing occupancy rates has made some progress, there are still some shortcomings, mainly reflected in the following aspects: First, the existing urban housing occupancy rate research is usually based on night light data or municipal service big data, but night light data is easily interfered by noise and the actual accuracy is not high, while municipal service big data is costly and difficult to obtain; Second, most of the current research focuses on static data such as urban landscape design (such as building density), POI (Point of Interest) density, and gross national product, and pays less attention to time series information, ignoring dynamic time series factors such as housing prices and air quality, and insufficient research on the time series characteristics of urban housing occupancy rates; Third, the spatial and temporal scales of occupancy rate predictions are not refined enough, especially at the spatial scale of urban communities and the monthly time scale. The research on the spatiotemporal characteristics of urban community housing occupancy is still not in-depth. Summary of the Invention

[0004] The purpose of this application is to provide a method, device and medium for predicting housing occupancy rate in urban residential areas, so as to achieve accurate prediction of housing occupancy rate in urban residential areas.

[0005] To achieve the above objectives, this application provides the following solutions: In a first aspect, the present application provides a method for predicting housing occupancy rates in urban residential areas, comprising: Determine the characteristic indicator data of the urban residential area to be predicted in the set time period based on the spatial boundary information, terrain data, near-surface pollutant PM2.5 concentration, POI, urban road network data, NPP-VIIRS data, and urban residential housing prices of the urban residential area to be predicted in the set time period; Preprocessing the characteristic index data of the urban residential area to be predicted in a set time period to obtain processed characteristic index data; According to the processed characteristic indicator data, the housing occupancy rate prediction model of the urban residential area is used to determine the housing occupancy rate of the urban residential area to be predicted; wherein, the urban residential area housing occupancy rate prediction model is obtained by training the initial model using a training data set; the training data set includes the characteristic indicator data of the urban residential area used for training in a historical set period and the corresponding true value of the housing occupancy rate; the true value of the housing occupancy rate is calculated using the urban residential area housing occupancy rate estimation model; the urban residential area housing occupancy rate estimation model is constructed based on the basic information of the urban residential area and mobile phone signaling data; the basic information of the urban residential area includes spatial range boundary information, the number of households in the urban residential area and the average number of people per household; the initial model includes a feature embedding coding module, a position coding layer, a hybrid multi-head attention mechanism module coupled with an efficient routing strategy, a residual connection module, a layer normalization module, a long short-term memory network and a second fully connected layer connected in sequence; the feature embedding coding module includes a feature embedding coding layer and a first fully connected layer.

[0006] In a second aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement any of the above-described methods for predicting housing occupancy rates in urban residential areas.

[0007] In a third aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-mentioned methods for predicting housing occupancy rates in urban residential areas.

[0008] According to the specific embodiments provided in this application, this application has the following technical effects: The present application provides a method, device, and medium for predicting the housing occupancy rate of an urban residential area. The method comprises obtaining spatial boundary information, mobile phone signaling data, terrain data, near-surface PM2.5 concentration, POI, urban road network data, NPP-VIIRS data, and housing prices of the urban residential area to be predicted during a set time period; determining characteristic indicator data of the urban residential area to be predicted during the set time period based on the above data; preprocessing the characteristic indicator data of the urban residential area to be predicted during the set time period to obtain processed characteristic indicator data; and determining the housing occupancy rate of the urban residential area to be predicted using an urban residential area housing occupancy rate prediction model based on the processed characteristic indicator data. The urban residential area housing occupancy rate prediction model is obtained by training an initial model using a training data set; the initial model comprises a sequentially connected feature embedding coding module, a position encoding layer, a hybrid multi-head attention mechanism module coupled with an efficient routing strategy, a long short-term memory network, and a second fully connected layer; the feature embedding coding module comprises a feature embedding coding layer and a first fully connected layer. The present application utilizes the urban residential area housing occupancy rate prediction model to predict the housing occupancy rate of the urban residential area to be predicted, thereby achieving accurate prediction of the housing occupancy rate of the urban residential area. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0010] Figure 1 A flowchart of a method for predicting housing occupancy rates in urban residential areas provided in one embodiment of the present application; Figure 2 A schematic flow chart of a method for predicting housing occupancy rates in urban residential areas provided in another embodiment of the present application; Figure 3 A schematic diagram of a process for calculating the actual housing occupancy rate in a city based on mobile phone signaling data provided by an embodiment; Figure 4 A schematic diagram of the structure of the urban residential area housing occupancy rate prediction model provided in this embodiment; Figure 5 A schematic diagram of occupancy rate prediction for 55 residential areas in a city provided for this embodiment; Figure 6 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0011] The following will be combined with the drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only part of the embodiments of this application, not all of the embodiments. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0012] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0013] This application constructs an urban residential community housing occupancy rate estimation model based on mobile phone signaling data through the boundary information of the spatial range of urban residential communities, mobile phone signaling grid data, and the actual housing occupancy rate of sampled urban residential communities; by constructing an urban residential community occupancy characteristic indicator system that describes the occupancy situation of urban residential communities, aligning data according to time scale and spatial scale, and establishing urban residential community characteristic indicator data; by coupling a hybrid multi-head attention mechanism and a long short-term memory network with an efficient routing strategy, a urban residential community housing occupancy rate prediction model is constructed to achieve accurate prediction of urban residential community housing occupancy rate at the urban community spatial scale and monthly time scale.

[0014] In an exemplary embodiment, Figure 1 As shown, a method for predicting housing occupancy rate in urban residential areas is provided, comprising the following steps: S1: Determine the characteristic indicator data of the urban residential area to be predicted during the set time period based on its spatial boundary information, topographic data, near-surface PM2.5 concentration, POIs, urban road network data, NPP-VIIRS data, and housing prices. The characteristic indicator data include social and psychological indicators, social and environmental indicators, socioeconomic indicators, and resident welfare indicators. Social and psychological indicators include pre-occupancy rates; social and environmental indicators include slope and air quality index; socioeconomic indicators include average housing prices, road network density, and nighttime light intensity; and resident welfare indicators include public transportation station coverage, school coverage, and medical service facility coverage.

[0015] As an optional implementation, S1 specifically includes: S11: Based on the terrain data, the slope of the entire city is determined using the surface fitting method, and the zoning statistics are performed based on the spatial range boundary information to obtain the slope of the urban residential area to be predicted in the set time period, and the slope is converted into a raster data format.

[0016] S12: Based on the PM2.5 concentration of near-surface pollutants, the Kriging interpolation method is used to determine the air quality index of the entire city. The air quality index of the entire city is divided into zones and statistics according to the boundary information of the spatial range. The air quality index of the residential area to be predicted in the set time period is obtained and converted into a raster data format.

[0017] S13: The mean of the housing prices of the urban communities is used as the average housing price of the urban residential community to be predicted in the set time period, and is converted into a vector data format.

[0018] S14: Based on the urban road network data, the road center lines of the urban road network are extracted, and the kernel density distribution of the road center lines is calculated to obtain the road network density of the entire city. The road network density of the entire city is divided into zones and statistics according to the spatial range boundary information to obtain the road network density of the urban residential area to be predicted in the set time period, and converted into a raster data format.

[0019] S15: Calculate the nighttime light intensity of the entire city based on the NPP-VIIRS data, perform zoning statistics on the nighttime light intensity of the entire city based on the spatial range boundary information, and obtain the nighttime light intensity of the residential area to be predicted in the set time period.

[0020] S16: Filter POIs by setting filtering conditions to obtain public facility points, calculate the kernel density distribution of each public facility point, obtain the public transportation site coverage rate, school coverage rate and medical service facility coverage rate of the entire city, perform zoning statistics on the public transportation site coverage rate, school coverage rate and medical service facility coverage rate of the entire city according to the spatial range boundary information, obtain the public transportation site coverage rate, school coverage rate and medical service facility coverage rate of the urban residential area to be predicted, and convert them into raster data format; public facility points include public transportation sites, schools and medical service facilities.

[0021] S2: Preprocessing the characteristic index data of the urban residential area to be predicted in a set time period to obtain processed characteristic index data.

[0022] As an optional implementation, S2 specifically includes: S21: Time alignment and spatial alignment are performed on the characteristic indicator data of the urban residential quarters to be predicted in the set time period to obtain aligned characteristic indicator data; the time alignment uses months as the time scale, and the spatial alignment uses urban residential quarters as statistical units, and the corresponding characteristic indicator data are aggregated to the corresponding urban residential quarters statistical units in accordance with regional statistics.

[0023] S22: performing normalization processing on the aligned feature index data to obtain processed feature index data.

[0024] S3: Based on the processed characteristic indicator data, the housing occupancy rate prediction model of the urban residential area is used to determine the housing occupancy rate of the urban residential area to be predicted; wherein, the housing occupancy rate prediction model of the urban residential area is obtained by training the initial model using the training data set; the training data set includes the characteristic indicator data of the urban residential area used for training in the historical set period and the corresponding true value of the housing occupancy rate; the true value of the housing occupancy rate is calculated using the urban residential area housing occupancy rate estimation model; the urban residential area housing occupancy rate estimation model is constructed based on the basic information of the urban residential area and mobile phone signaling data; the basic information of the urban residential area includes spatial range boundary information, the number of households in the urban residential area and the average number of people per household; the initial model includes a feature embedding coding module, a position coding layer, a hybrid multi-head attention mechanism module coupled with an efficient routing strategy, a residual connection module, a layer normalization module, a long short-term memory network and a second fully connected layer connected in sequence; the feature embedding coding module includes a feature embedding coding layer and a first fully connected layer.

[0025] The following describes a specific embodiment of the process of constructing the urban residential area housing occupancy rate prediction model used in the urban residential area housing occupancy rate prediction method of the present application.

[0026] In this embodiment, if Figure 2 As shown, the specific steps are as follows: S10: Build an urban residential area housing occupancy rate estimation model based on basic information of urban residential areas and mobile phone signaling data, and use the actual collected housing occupancy rates to correct the model. The basic information of urban residential areas includes the spatial boundary information of urban residential areas, the number of households in urban residential areas, and the average number of people per household in the city.

[0027] In this embodiment, if Figure 3 As shown, S10 specifically includes: S101: Calculate the actual resident population of the urban residential area based on the spatial boundary information of the urban residential area and mobile phone signaling data.

[0028] In practical applications, the actual permanent population of urban residential areas is determined based on the spatial boundary information of urban residential areas and mobile phone signaling data. First, the obtained spatial boundary information of the urban residential area is spatially aligned with the urban residential area location points. The mobile phone signaling data is cleaned to effectively deal with invalid data, ping-pong data, and drift data, thereby obtaining high-quality trajectory point data. The trajectory point data is spatiotemporally clustered and valid travel information is extracted to obtain stop point information and trip origin and destination information (trip origin and destination information). The optimal time period for permanent population statistics is determined, and population data is counted between 8 pm and 7 am the following morning to calculate the actual permanent population. The actual population distribution is determined at a 150m×150m grid scale. The actual permanent population of urban residential areas is calculated using methods such as spatial intersection statistics.

[0029] S102: Construct an initial estimation model for the housing occupancy rate of urban residential areas based on the number of households in the urban residential areas, the average number of people per household in the city, and the actual number of permanent residents in the urban residential areas.

[0030] In practical applications, based on the number of households, average household population, and actual permanent population of urban residential communities, the initial estimation model for urban residential community housing occupancy rates is used, combined with the spatial boundary information of urban residential community r, to determine the estimated value of the housing occupancy rate of urban residential community r. The initial estimation model for urban residential community housing occupancy rates is: .

[0031] Where: is the estimated housing occupancy rate of urban residential area r; is the actual resident population of urban residential area r; is the average household population, is the number of households in urban residential area r.

[0032] S103: Sampling existing urban residential communities, obtaining the actual housing occupancy rates of the sampled urban residential communities through actual sampling, and calculating the estimated housing occupancy rates based on the initial estimation model of the housing occupancy rates of the urban residential communities.

[0033] S104: Fitting the estimated housing occupancy rate with the actual housing occupancy rate, revising the estimation model of the housing occupancy rate of the urban residential area, and constructing an urban residential area housing occupancy rate estimation model based on mobile phone signaling data for calculating the actual housing occupancy rate of the urban residential area.

[0034] The estimated housing occupancy rates of urban residential areas were revised based on actual sampling data to calculate the actual housing occupancy rates of urban residential areas. Urban residential areas were sampled and information was collected through field surveys at night to obtain the actual housing occupancy rates of the sampled urban residential areas. The calculated estimated housing occupancy rates of urban residential areas were revised to construct a housing occupancy rate estimation model for urban residential areas based on mobile phone signaling data: .

[0035] .

[0036] Where, is the correction factor, the actual housing occupancy rate of urban residential areas The maximum value of is 1, and all values greater than 1 are set to 1. The correlation test and linear regression significance test were conducted with the occupancy rate of urban residential areas collected from the night field survey. If the test failed, the adjustment Recalculate the actual housing occupancy rate of urban residential areas , until the verification is successful. The actual housing occupancy rate of the community is estimated to be 1.79.

[0037] This example will conduct two rounds of precision testing on the data obtained by comparing it with existing research in a certain city and field sampling at night using the Pearson correlation coefficient method and linear regression method. In the published vacancy rate information of a certain city, its vacancy rate is 24.52%. The vacancy rate calculated by this example is about 22.78%, which is only about 1.74% different. The Pearson correlation coefficient between the housing occupancy rate measured in this example and the housing occupancy rate obtained by field sampling at night is 0.934, which passed the two-tailed test with a significance level of 0.01, as shown in Table 1; the linear regression of the two also passed the significance test, R 2 The housing vacancy rate is 0.872. Therefore, the accuracy of the housing vacancy rate estimation is relatively high.

[0038] Table 1 Pearson correlation coefficient table

[0039] S20: Determine characteristic indicator data using the spatial boundaries, topography, PM2.5 concentration, POI (Point of Interest), urban road network, NPP-VIIRS, and housing prices of the urban residential area to be predicted during a set time period.

[0040] The factors related to the occupancy of urban residential communities were screened, and characteristic indicators related to the occupancy of urban residential communities were constructed. The characteristic indicators include 4 first-level indicators and 9 second-level indicators. The first-level indicators include social psychology, social environment, social economy, and residents' welfare. The first-level social psychology indicator includes 1 second-level indicator, namely the early occupancy rate, which reflects the popularity of the community. The first-level social environment indicator includes 2 second-level indicators, namely the slope and the air quality index. The slope reflects the terrain conditions of the area where the community is located, and the air quality index reflects the air quality in the area where the community is located. The first-level socioeconomic indicator includes 3 second-level indicators, namely the average housing price in residential communities, the road network density, and the night light intensity. The average housing price in residential communities reflects the living cost of residents, the road network density reflects the convenience of citizens' travel, and the night light intensity reflects the economic development intensity of the region. The first-level resident welfare indicator includes 3 second-level indicators, namely the public transportation station coverage rate, the school coverage rate, and the medical service facility coverage rate. The coverage rate of public transportation stations, schools, and medical service facilities reflects the supporting facilities of the living circle.

[0041] S30: Align spatiotemporal data, calculate characteristic indicators, and establish characteristic indicator data for urban residential areas.

[0042] The spatial boundary information of urban residential areas, terrain data (time scale is year, spatial scope is the entire city), near-surface pollutant PM2.5 concentration (time scale is month, spatial scope is the entire city), POI (time scale is year, spatial scope is the entire city), urban road network data (time scale is year, spatial scope is the entire city), urban residential housing prices (time scale is month, spatial scope is the entire city) and other data are used to align the data according to the time scale and spatial scale, and calculate the 9 secondary indicators corresponding to the characteristic index system of urban residential areas. Taking urban residential areas as statistical units, the corresponding indicator data are aggregated to the corresponding urban residential area statistical units in the form of regional statistics, and finally the characteristic indicator data of urban residential areas are formed through normalization.

[0043] In this embodiment, step S30 includes: S301: The slope index uses the terrain data, namely the DEM (Digital Elevation Model) data, to perform slope analysis using the fitting surface method, and uses the degree method to express it. The slope is calculated. The time scale is the year, and the spatial range is the entire city. Zoning statistics are performed based on the boundary information of the spatial range to obtain the slope of the urban residential area to be predicted in the set time period, and the data is converted into a raster data format. The air quality index is calculated using the PM2.5 concentration data of near-surface pollutants, and spatial interpolation is performed on it as a measurement indicator. The Kriging interpolation method is used to perform spatial interpolation on the PM2.5 concentration data of near-surface pollutants to calculate the air quality index. The time scale is the month, and the spatial scale is the urban area. According to the The spatial range boundary information is used to perform zoning statistics on the air quality index of the entire city, and the air quality index of the urban residential area to be predicted in the set time period is obtained and converted into raster data format; the average housing price of the residential area adopts the urban residential area housing price data, and the average transaction price of each house in the urban residential area (urban residential area housing price) is calculated as the average housing price of the residential area, and the average housing price of the residential area is calculated. The time scale is month, the spatial scale is the urban area, and it is converted into vector data format; the road network density is calculated using the urban road network data, the road center line of the urban road network is extracted, the kernel density distribution of the road center line is calculated, and the road network density is calculated. The time scale is The spatial scale is the city area, and the road network density of the entire city is divided into districts and statistics according to the boundary information of the spatial range, and the road network density of the urban residential area to be predicted in the set time period is obtained and converted into raster data format; the night light intensity is calculated using NPP-VIIRS data, and the night light intensity is calculated. The time scale is monthly, and the spatial scale is the city area. The night light intensity of the entire city is divided into districts and statistics according to the boundary information of the spatial range, and the night light intensity of the urban residential area to be predicted in the set time period is obtained; the public transportation station coverage rate, school coverage rate, and medical service facility coverage rate are calculated using POI density Calculations were performed to filter POI density by setting filtering conditions. Qualified public transportation stations, schools, medical service facilities, and other public facilities were selected. The kernel density distribution of the corresponding public facilities was calculated, and the public transportation station coverage rate, school coverage rate, and medical service facility coverage rate were calculated. The time scale was years, and the spatial scale was the city area. Based on the spatial range boundary information, the public transportation station coverage rate, school coverage rate, and medical service facility coverage rate of the entire city were divided into statistical regions. The public transportation station coverage rate, school coverage rate, and medical service facility coverage rate of the residential area to be predicted were obtained and converted into raster data format. This is shown in Table 2.

[0044] S302: For the indicators obtained by the above calculation, in terms of time alignment, each indicator is uniformly aligned according to the monthly time scale. For the data on the annual scale, the corresponding month uniformly adopts the consistent corresponding annual data; in terms of spatial alignment, the raster data format is resampled, and the resampling size is 50m×50m. The spatial range of the urban residential area is used as the statistical unit, and the corresponding indicator data is aggregated to the courtyard range statistical unit of the corresponding residential area in the form of regional statistics; for the vector data format, it is attached to the corresponding urban residential area through the spatial relationship, and the corresponding indicator data is aggregated to the courtyard range statistical unit of the corresponding residential area; finally, the characteristic indicator data of the urban residential area is formed through the normalization method.

[0045] Table 2 Characteristic index table

[0046] S40: By coupling a hybrid multi-head attention mechanism based on an efficient routing strategy with a long short-term memory network, a housing occupancy rate prediction model for urban residential areas is constructed. The preprocessed feature indicator data is input to determine the predicted value of the housing occupancy rate of the urban residential area to be predicted.

[0047] In this embodiment, step S40 constructs an initial model for predicting the housing occupancy rate of urban residential areas. Figure 4 As shown in the figure, the characteristic indicator data of urban residential areas is used as the input time series dataset, which is decomposed into different time series slices as the input of the prediction model. Feature embedding encoding and position encoding are first performed, and then the global information of each time series slice is captured through a hybrid multi-head attention mechanism based on an efficient routing strategy, and the time series global features of the characteristic indicator dataset are learned; after the hybrid multi-head attention mechanism based on an efficient routing strategy, a long short-term memory network is added, and the gating mechanism of the long short-term memory network is used to learn the time sequence features of the time series slices, capture the long-term dependency of the characteristic indicator sequence data, and learn the time series local features of the characteristic indicator dataset; after the long short-term memory network, a second fully connected layer is added, and the second fully connected layer is used to further learn the nonlinear relationship and high-level features in the time series slices, and map them to the output space through a nonlinear activation function and a linear layer to obtain the prediction results.

[0048] In a specific embodiment, the data processing process of the initial model is as follows: S401: Urban residential area characteristic index data Decompose into different time series slices , slice the time series Input to the encoding layer (feature embedding encoding layer), the encoding layer slices the time series Perform linear affine transformation encoding, map the input features to a high-dimensional feature space through a first fully connected layer, perform feature extraction, and form feature embedding encoding of time series slices .

[0049] .

[0050] in, are input features (time series slices), It is the dimension of the output feature space, Embedding represents the encoding layer, and a linear fully connected layer is used.

[0051] S402: Embedding the feature of the time series Add position information of the time series, that is, position coding information , forming a feature embedding code that integrates position coding information .

[0052] .

[0053] S403: (1) Linear transformation. Embed the feature of the fused position coding information into the code Input to the hybrid multi-head attention mechanism with efficient routing strategy, through the learnable weight matrix 、 、 , respectively calculate the query ( ),key( ) and value ( ).

[0054] .

[0055] (2) Separate. 、 、 Split into h heads: .

[0056] .

[0057] .

[0058] Where i=1,...,h; is the query vector of the i-th head; is the key vector of the i-th head; is the value vector of the i-th head; is the dimension of the i-th head; ; C is the feature dimension; h is the total number of heads.

[0059] (3) Calculate the attention score. Calculate the attention score for each head: .

[0060] in, is the attention score of the i-th head; T is the transpose.

[0061] (4) Routing weight generation. Use the routing network to generate the weight of each head: .

[0062] in, is the attention score of the i-th head; is the mean of the sequence dimension, It is a multi-layer perceptron; is the softmax function.

[0063] (5) Introduce sparsity in the routing network and only select the weights of the top-k heads.

[0064] .

[0065] in, is the weight of the top-k largest weighted heads; The index of the top-k (first k) heads with the largest weights; Generate per-head weights for using the routing network; is a function for selecting the top-k heads with the largest weights; k is the number of heads with the largest weights selected.

[0066] (6) Mix the outputs of the headers. Use the routing weights to combine the outputs of all the headers: .

[0067] in, The output is a weighted combination of the top-k heads.

[0068] (7) Output projection. The output of the mixing head is mapped back to the original dimension through a linear transformation: .

[0069] in, , is the output weight matrix.

[0070] S404: Global Characteristics of Time Series Input and Calculation is performed in the module, where the Add module is a residual connection module. The module is a layer normalization module: .

[0071] Then the first output of the layer normalization module is Sent to the feedforward layer Perform the calculation: .

[0072] The feedforward layer outputs the result and the first output of the layer normalization module Enter again and Calculations are performed in the module: .

[0073] S405: The second output result of the global feature (layer normalization module ) is input into the long short-term memory network to further extract features: .

[0074] in, is the output feature of the long short-term memory network; 、 is the hyperparameter of the long short-term memory network, It is the hidden state at the previous moment.

[0075] S406: To further obtain high-level features in the time series, the output It is sent to the second fully connected layer for further feature extraction.

[0076] .

[0077] in, and are the weights and biases of the fully connected layer, is the ReLU function.

[0078] S407: Finally, the feature matrix output by the second fully connected layer Send it to the output layer for dimensionality reduction activation processing to get the prediction result.

[0079] .

[0080] in, is the housing occupancy rate of urban residential areas; and are the weights and biases of the output layer.

[0081] The model of this embodiment uses a stochastic gradient descent optimizer to calculate the mean square error loss function, sets the learning rate to 0.00085, the momentum to 0.99, and randomly divides the constructed data set into a training data set and a test data set in a ratio of 8:2, and trains for 2500 epochs. In order to verify the effectiveness of the method of this application, it is compared with four different models (Linear Regression Model (LR), Gated Recurrent Unit (GRU), Long Short-Term Memory Network (LSTM), Transformer), and the MAE (Mean Absolute Error, mean absolute error), RMSE (Root Mean Squared Error, root mean square error), Evaluate its accuracy.

[0082] As shown in Table 3, the proposed method outperforms the other four models. Although the Transformer model has achieved good fit during training, there is still room for improvement in performance and efficiency. In contrast, the proposed method shows better performance, with an average It reached 0.8630, MAE was 0.0611, and RMSE was 0.1078, which highlights that the method proposed in this application has smaller errors when fitting data and stronger prediction ability, and this method is more suitable for predicting the occupancy rate of urban communities.

[0083] Table 3 Comparison results of various models

[0084] This example estimates the occupancy rate of some residential communities in a city in 2024. 55 communities were randomly surveyed through field surveys, photos, and visits. The actual housing occupancy rate is as follows: Figure 5 The results show that the actual housing occupancy rate of most communities in 2024 is basically consistent with the housing occupancy rate predicted by the urban residential community housing occupancy rate prediction model. This embodiment calculates the MAE, RMSE, and , where MAE is 0.056 and RMSE is 0.0952, It is 0.7571, which verifies the validity of the results and proves that the prediction accuracy of the model is high.

[0085] The above embodiment provides a method for predicting the housing occupancy rate of urban residential communities. First, a model for estimating the housing occupancy rate of urban residential communities based on mobile phone signaling data is constructed based on the boundary information of the spatial scope of the urban residential communities, mobile phone signaling grid data, the number of households in the urban residential communities, and the average population per household in the city. Second, factors related to the occupancy of urban residential communities are screened to construct a characteristic index system related to the occupancy of urban residential communities. Then, based on the urban residential community occupancy characteristic index system, urban residential community characteristic index data is constructed. Finally, a hybrid multi-head attention mechanism based on an efficient routing strategy and a long short-term memory network are coupled to construct a model for predicting the housing occupancy rate of urban residential communities. The method is verified to be superior to other traditional models based on actual data and has high reliability. The embodiment can effectively predict the housing occupancy rate of urban residential communities, providing a scientific basis for national land space planning and monitoring, land conservation and intensive use, and optimization of urban functional structure.

[0086] In an exemplary embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the above-mentioned urban residential area housing occupancy rate prediction method when executing the computer program.

[0087] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, which, when executed by a processor, implements the above-mentioned method for predicting housing occupancy rates in urban residential areas.

[0088] In an exemplary embodiment, a computer program product is provided, comprising a computer program, which, when executed by a processor, implements the above-mentioned method for predicting housing occupancy rates in urban residential areas.

[0089] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 6 As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, abbreviated as I / O) and a communication interface. The processor, memory and input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, a method for predicting housing occupancy rate in an urban residential area is implemented.

[0090] Those skilled in the art will understand that Figure 6 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0091] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0092] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. In particular, any reference to memory, database, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0093] The databases involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processors involved in the various embodiments provided herein may include, but are not limited to, general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic units, data processing logic units based on quantum computing, and the like.

[0094] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0095] This document uses specific examples to illustrate the principles and implementation methods of this application. The description of the above examples is only intended to help understand the method and core concept of this application. At the same time, for those skilled in the art, based on the concept of this application, there may be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for predicting housing occupancy rate in urban residential areas, characterized in that: include: Determine the characteristic indicator data of the urban residential area to be predicted in the set time period based on the spatial boundary information, terrain data, near-surface pollutant PM2.5 concentration, POI, urban road network data, NPP-VIIRS data, and urban residential housing prices of the urban residential area to be predicted in the set time period; Preprocessing the characteristic index data of the urban residential area to be predicted in a set time period to obtain processed characteristic index data; According to the processed characteristic indicator data, the housing occupancy rate prediction model of the urban residential area is used to determine the housing occupancy rate of the urban residential area to be predicted; wherein, the urban residential area housing occupancy rate prediction model is obtained by training the initial model using a training data set; the training data set includes the characteristic indicator data of the urban residential area used for training in a historical set period and the corresponding true value of the housing occupancy rate; the true value of the housing occupancy rate is calculated using the urban residential area housing occupancy rate estimation model; the urban residential area housing occupancy rate estimation model is constructed based on the basic information of the urban residential area and mobile phone signaling data; the basic information of the urban residential area includes spatial range boundary information, the number of households in the urban residential area and the average number of people per household; the initial model includes a feature embedding coding module, a position coding layer, a hybrid multi-head attention mechanism module coupled with an efficient routing strategy, a residual connection module, a layer normalization module, a long short-term memory network and a second fully connected layer connected in sequence; the feature embedding coding module includes a feature embedding coding layer and a first fully connected layer.

2. The method for predicting housing occupancy rate in urban residential areas according to claim 1, characterized in that: Based on the basic information of urban residential areas and mobile phone signaling data, a housing occupancy rate estimation model for urban residential areas is constructed, which includes: Determine the actual permanent population of urban residential areas based on the spatial boundary information of urban residential areas and mobile phone signaling data; Based on the number of households, average household population, and actual permanent population of urban residential areas, the initial estimation model of housing occupancy rate in urban residential areas was used. , combined with the spatial boundary information of urban residential area r, determine the estimated housing occupancy rate of urban residential area r; where, is the estimated housing occupancy rate of urban residential area r; is the actual permanent population of urban residential area r; is the average household population, is the number of households in urban residential area r; By collecting information through field surveys, we can obtain the measured values of housing occupancy rates in urban residential areas r; The estimated value of housing occupancy rate of urban residential area r is revised by using the measured value of housing occupancy rate of urban residential area r, and the estimation model of housing occupancy rate of urban residential area is constructed. Using the housing occupancy rate estimation model of urban residential areas , determine the true value of the housing occupancy rate of urban residential area r; where, is the actual value of the housing occupancy rate of urban residential area r; is the correction factor; Conduct correlation and significance tests on the measured values of housing occupancy rates in urban residential area r and the true values of housing occupancy rates in urban residential area r; If the test fails, adjust the correction factor and return to "Use the urban residential area housing occupancy rate estimation model to determine the true value of the housing occupancy rate in urban residential area r"; If the test passes, the housing occupancy rate estimation model of urban residential areas is obtained.

3. The method for predicting housing occupancy rate in urban residential areas according to claim 2, characterized in that: Determine the actual permanent population of urban residential areas based on the spatial boundary information of urban residential areas and mobile phone signaling data, including: Spatially aligning the spatial range boundary information of urban residential areas with the location points of each urban residential area; The mobile phone signaling data of the urban residential areas after spatial alignment is cleaned to obtain trajectory point data; Perform spatiotemporal clustering of trajectory point data and extract travel information to obtain stop point information and travel start-end point information; Calculate the total permanent population of the entire city based on the optimal statistical period, as well as the stop point information and the travel origin and destination information; the optimal statistical period is 8 pm to 7 am the next day; Based on the total permanent population of the entire city, the actual population distribution at the 150m×150m grid scale is determined, and the actual permanent population of urban residential areas is calculated using spatial intersection statistics.

4. The method for predicting housing occupancy rate in urban residential areas according to claim 1, characterized in that: Characteristic indicator data include social-psychological indicators, social-environmental indicators, socio-economic indicators and resident welfare indicators; social-psychological indicators include pre-occupancy rate; social-environmental indicators include slope and air quality index; socio-economic indicators include average house price, road network density and night light intensity; resident welfare indicators include public transportation station coverage, school coverage and medical service facility coverage.

5. The method for predicting housing occupancy rate in urban residential areas according to claim 1, characterized in that: Based on the spatial boundary information, topographic data, PM2.5 concentration of near-surface pollutants, POI, urban road network data, NPP-VIIRS data, and housing prices of the urban residential areas to be predicted during the set time period, the characteristic indicator data of the urban residential areas to be predicted during the set time period are determined, including: Based on the terrain data, the slope of the entire city is determined using the surface fitting method, and the slope of the residential area to be predicted in the set time period is obtained by performing zoning statistics based on the spatial range boundary information, and the slope is converted into a raster data format; Based on the PM2.5 concentration of near-surface pollutants, the Kriging interpolation method is used to determine the air quality index of the entire city. The air quality index of the entire city is divided into different zones according to the spatial range boundary information to obtain the air quality index of the residential area to be predicted in the set time period and convert it into a raster data format; The mean of the housing prices in the urban districts is used as the average housing price of the residential district to be predicted in the set time period and converted into a vector data format; Extracting the road centerlines of the urban road network based on the urban road network data, calculating the kernel density distribution of the road centerlines, and obtaining the road network density of the entire city; performing zoning statistics on the road network density of the entire city based on the spatial range boundary information, and obtaining the road network density of the residential area to be predicted in a set time period, and converting the data into a raster data format; Calculate the nighttime light intensity of the entire city based on the NPP-VIIRS data, perform zoning statistics on the nighttime light intensity of the entire city based on the spatial range boundary information, and obtain the nighttime light intensity of the residential areas to be predicted in the set time period; POIs are filtered by setting filtering conditions to obtain public facility points, and the kernel density distribution of each public facility point is calculated to obtain the public transportation station coverage rate, school coverage rate and medical service facility coverage rate of the entire city. The public transportation station coverage rate, school coverage rate and medical service facility coverage rate of the entire city are divided into statistical regions according to the spatial range boundary information to obtain the public transportation station coverage rate, school coverage rate and medical service facility coverage rate of the urban residential area to be predicted, and converted into raster data format; public facility points include public transportation stations, schools and medical service facilities.

6. The method for predicting housing occupancy rate in urban residential areas according to claim 1, characterized in that: The characteristic index data of the urban residential area to be predicted in the set time period are pre-processed to obtain the processed characteristic index data, specifically including: Perform temporal and spatial alignment on the characteristic indicator data of the urban residential quarters to be predicted in the set time period to obtain aligned characteristic indicator data; temporal alignment uses months as the time scale, and spatial alignment uses urban residential quarters as statistical units; The aligned feature index data are normalized to obtain processed feature index data.

7. The method for predicting housing occupancy rate in urban residential areas according to claim 1, characterized in that: Based on the processed characteristic index data, the housing occupancy rate prediction model of urban residential quarters is used to determine the housing occupancy rate of the urban residential quarters to be predicted, specifically including: The characteristic index data of the urban residential area to be predicted after processing in the set time period Decompose into different time series slices ; Slicing the time series Input to the feature embedding coding module, the feature embedding coding layer slices the time series Perform linear affine transformation encoding. The first fully connected layer maps the encoded time series slices to a high-dimensional feature space, performs feature extraction, and forms time series slices. Feature embedding coding ; The expression of the feature embedding coding module is: ;in, is a time series slice, is the dimension of the high-dimensional feature space, and Embedding represents the feature embedding coding layer; The feature embedding code ED is input to the position coding layer, and the position coding information PE of the time sequence is added to form a feature embedding code that integrates the position coding information. ; The expression of the position coding layer is: ; Embed the feature of fused position encoding information into encoding The input is fed into the hybrid multi-head attention mechanism module coupled with an efficient routing strategy, which performs linear transformation, splits the heads, calculates the attention scores, generates routing network weights, selects the weights of the top-k heads, outputs the hybrid head, and performs projection mapping to the original dimension to obtain the global features of the time series. ; Embed the feature of fused position encoding information into encoding and global characteristics of time series Input into the residual connection module and the layer normalization module for calculation, and the calculation formula is: ;in, is the first output result of the layer normalization module; Add is the residual connection module, For the layer normalization module: The first output result is sent to the feedforward layer for calculation. The calculation formula is: ;in, is the output result of the feedforward layer, is the feed-forward layer; The output result of the feedforward layer and the first output result are input into the residual connection module and the layer normalization module again for calculation. The calculation formula is: ;in, is the second output result of the layer normalization module; The second output result Input into the long short-term memory network to extract features, the calculation formula is: ;in, is the output feature of the long short-term memory network; Long short-term memory network; 、 is the hyperparameter of the long short-term memory network, is the hidden state at the previous moment; The output features of the long short-term memory network It is sent to the second fully connected layer for feature extraction, and the calculation formula is: ;in, is the feature matrix output by the second fully connected layer; and are the weights and biases of the second fully connected layer; is the activation function; The feature matrix output by the second fully connected layer Input to the output layer, perform dimensionality reduction processing, and obtain the housing occupancy rate of urban residential areas. The calculation formula is: ;in, is the housing occupancy rate of urban residential areas; and are the weights and biases of the output layer.

8. The method for predicting housing occupancy rate in urban residential areas according to claim 7, characterized in that: Embed the feature of fused position encoding information into encoding The input is fed into the hybrid multi-head attention mechanism module coupled with an efficient routing strategy, which performs linear transformation, splits the heads, calculates the attention scores, generates routing network weights, selects the weights of the first k heads, outputs the hybrid head, and performs projection mapping to the original dimension to obtain the global features of the time series. , specifically including: Embed the feature of fused position encoding information into encoding Perform linear transformation through a learnable weight matrix 、 、 , respectively calculate the query ,key Sum , the calculation formula is: ; Perform separate operations and query ,key Sum Split into h heads, the calculation formula is: ;in, is the query vector of the i-th head; is the key vector of the i-th head; is the value vector of the i-th head; i=1, 2, ..., h, is the dimension of the i-th head, , C is the feature dimension, is the number of heads obtained by splitting; Calculate the attention score for each head using the following formula: ;in, is the attention score of the i-th head; T is the transpose; Use the routing network to calculate the feature embedding code that fused the position coding information The routing weight of each header is calculated as follows: ;in, Generate per-head weights for using the routing network; It is the mean of the feature embedding code of the fused position coding information; is a multi-layer perceptron; is the softmax function; Introducing sparsity into the routing network, only selecting the routing weights of the first k heads as the feature embedding code that fused the position coding information Output weights: ;in, is the weight of the top-k largest weighted heads; is the index of the first k heads with the largest weights; is a function that selects the top-k heads with the largest weights; k is the number of heads with the largest weights selected; The output of all heads is weightedly combined using the routing weights as the output of the mixing head: ;in, The output is a weighted combination of the top-k heads. The output of the mixing head is mapped back to the original dimension through linear transformation to obtain the global feature F of the time series: ;in, is the output weight matrix.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for predicting housing occupancy rate in an urban residential area according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for predicting the housing occupancy rate of an urban residential area according to any one of claims 1 to 8 is implemented.

Images