Urban business center identification method and system based on big data

Abstract

The invention discloses an urban business center identification method and system based on big data, and relates to the technical field of urban management, and the method comprises the steps: training a gradient lifting decision tree through multi-source data based on differential privacy and a homomorphic encryption algorithm, carrying out the gradient aggregation, carrying out the collaborative modeling based on business features, and generating a joint feature weight matrix; based on the joint feature weight matrix, obtaining an optimal spatial-temporal feature weight through a quantum annealing algorithm in combination with classical simulated annealing adjustment, monitoring the stability of the optimal spatial-temporal feature weight through a dynamic adjustment mechanism, and performing dynamic adjustment; and in combination with the optimal spatial-temporal feature weight and the road network real-time state data, performing spatial-temporal kernel density calculation on the whole-city grid through an adaptive kernel function, and identifying the city business center according to a calculation result. According to the invention, the global optimization and dynamic adjustment of the feature weight are realized by combining the temporal-spatial feature weight optimization of classical simulated annealing based on the quantum annealing algorithm.

Description

Technical Field

[0001] The present invention relates to the field of urban management technology, and in particular to a method and system for identifying urban commercial centers based on big data. Background Art

[0002] With the acceleration of urbanization and the increasing complexity of commercial activities, the identification and planning of urban commercial centers has become a crucial issue in urban management and development. Traditional commercial center identification methods primarily rely on manual surveys and statistical data analysis, such as commercial outlet density and foot traffic statistics. However, these methods suffer from long data collection cycles, high costs, and limited coverage. In recent years, with the rapid development of big data technologies, commercial center identification methods based on multi-source data have gradually become a research hotspot. For example, the use of heterogeneous multi-source data such as mobile phone signaling data, UnionPay transaction data, and traffic flow data, combined with machine learning algorithms (such as gradient boosting decision trees and random forests) to extract and model commercial features, has been applied in multiple cities. Despite this, existing technologies still have certain limitations, particularly in terms of data privacy protection, model generalization, and real-time performance, and urgently need further optimization and improvement.

[0003] The main shortcomings of existing technologies are: First, data privacy protection has not been effectively addressed. Existing methods often directly use multi-source data for modeling, lacking encryption and protection measures for sensitive data, and posing the risk of data leakage. Second, existing methods lack real-time and dynamic adaptability. The distribution and characteristics of commercial centers change over time, and existing methods mostly rely on static analysis of historical data, which is difficult to reflect real-time dynamic changes. Furthermore, existing methods also have shortcomings in feature weight optimization and model generalization, making them difficult to adapt to the diverse needs of different cities and regions. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a city commercial center identification method based on big data to solve the problems of insufficient data privacy protection, poor real-time and dynamic adaptability, and insufficient feature weight optimization in the existing technology.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions: In the first aspect, the present invention provides a method for identifying urban commercial centers based on big data, which includes collecting and preprocessing multi-source data, and extracting commercial features based on the preprocessed multi-source data; the multi-source data includes mobile phone signaling data, UnionPay transaction data, traffic flow data and commercial outlets data; based on differential privacy and homomorphic encryption algorithms, a gradient boosting decision tree is trained through multi-source data, and gradient aggregation is performed, and collaborative modeling is performed based on commercial features to generate a joint feature weight matrix; based on the joint feature weight matrix, the optimal spatiotemporal feature weights are obtained through quantum annealing algorithm combined with classical simulated annealing tuning, and the stability of the optimal spatiotemporal feature weights is monitored and dynamically adjusted through a dynamic adjustment mechanism; combined with the optimal spatiotemporal feature weights and real-time status data of the road network, the spatiotemporal kernel density of the entire city grid is calculated through an adaptive kernel function, and the urban commercial center is identified based on the calculation results.

[0007] As a preferred solution of the urban commercial center identification method based on big data described in the present invention, the preprocessing includes data cleaning, format conversion, deduplication, normalization and outlier processing.

[0008] As a preferred solution of the method for identifying urban commercial centers based on big data of the present invention, the commercial features are extracted based on the pre-processed multi-source data, and the specific steps are as follows: Based on the commercial network data, the density characteristics of commercial network are extracted by kernel density estimation method; Based on mobile phone signaling data, the characteristics of pedestrian flow are extracted through the node centrality analysis method in graph theory; Based on traffic flow data, the number of transportation hubs in the urban area is counted to extract the connectivity characteristics of transportation hubs; Based on UnionPay transaction data, the transaction growth rate characteristics are extracted by calculating the changes in transaction amounts within a specific time period.

[0009] As a preferred solution of the urban commercial center identification method based on big data described in the present invention, wherein: based on differential privacy and homomorphic encryption algorithm, gradient boosting decision tree is trained by multi-source data and gradient aggregation is performed. The specific steps are as follows: Initialize the gradient boosting decision tree based on the preprocessed multi-source data; The extracted business features are encrypted using a homomorphic encryption algorithm to generate an encrypted feature vector; Use the encrypted feature vector to train the initialized gradient boosting decision tree and obtain the gradient decision update value ; Use homomorphic encryption to update gradient decisions Encrypt, generate the gradient decision update value in the encrypted state, and upload it to the central server through a secure channel; The central server decides the updated value based on the gradient in the encrypted state Aggregate the gradient and inject Gaussian noise through differential privacy to generate noisy aggregate gradient ; Aggregate gradient with noise via key pair in homomorphic encryption algorithm Decrypt and generate the global parameters of the gradient boosting decision tree; Based on the global parameters of the gradient boosting decision tree, the business features are collaboratively modeled to generate a joint feature weight matrix ; The global parameters of the gradient boosting decision tree include the node splitting characteristics, splitting threshold and leaf node weight of each tree in the gradient boosting decision tree.

[0010] As a preferred solution of the big data-based urban commercial center identification method of the present invention, the optimal spatiotemporal feature weights are obtained by combining the quantum annealing algorithm with classical simulated annealing tuning based on the joint feature weight matrix. The specific steps are as follows: The joint feature weight matrix is ​​converted into Each element in corresponds to the spin variable of the quantum bit, and a Hamiltonian of the quadratic unconstrained binary optimization form is constructed, and the joint feature weight matrix Mapping to the objective function of the quantum annealing algorithm ; Each feature weight in the joint feature weight matrix corresponds to a quantum bit; Based on the objective function, set the initial parameters of the quantum annealing algorithm; The initial parameters of the quantum annealing algorithm include annealing time, number of quantum bits and initial temperature; Based on the objective function and initial parameters of the quantum annealing algorithm, the annealing process is performed to gradually optimize the objective function and generate the feature weights after annealing optimization. ; The optimal feature weight combination is converted into a classical feature weight vector through quantum measurement and binary decoding; Based on the quantum annealing results and the joint feature weight matrix, the objective function of classical simulated annealing is defined ; Perform simulated annealing based on the objective function of classic simulated annealing to generate feature weights after simulated annealing optimization ; The feature weights after annealing optimization and the feature weights after simulated annealing optimization are combined to generate the optimal spatiotemporal feature weights. .

[0011] As a preferred solution of the method for identifying urban commercial centers based on big data of the present invention, the stability of the optimal spatiotemporal feature weights is monitored and dynamically adjusted through a dynamic adjustment mechanism. The specific steps are as follows: Calculate the standard deviation of the optimal spatiotemporal feature weights ; The standard deviation of the optimal spatiotemporal feature weight It is defined as the weight stability index; Define a stability threshold based on the standard deviation distribution of feature weights in historical data ; when > When , the optimal spatiotemporal feature weight is considered unstable, and the annealing time and the number of quantum bits are increased; when ≤ When , the optimal spatiotemporal feature weights are considered stable.

[0012] As a preferred solution of the big data-based urban commercial center identification method of the present invention, the method combines the optimal spatiotemporal feature weights and the real-time status data of the road network, calculates the spatiotemporal kernel density of the entire city grid through an adaptive kernel function, and identifies the urban commercial center based on the calculation results. The specific steps are as follows: The real-time status data of the road network includes traffic flow, congestion index and road closure information; Divide the urban area into M×N grids, treat each grid as an independent area and assign a unique identifier; Using Gaussian kernel function as a framework, combined with dynamic bandwidth adjustment of real-time road network status data to form an adaptive kernel function ; Combine the adaptive kernel function with the time decay function to calculate the spatiotemporal kernel density of each region in the grid ; Based on the distribution of spatiotemporal kernel density in historical data, a high-density threshold K1 and a low-density threshold K2 are defined; when When ≥K1, the current area is considered to be the main commercial center; When K2≤ When <K1, the current area is considered to be a secondary commercial center; when When <K2, the current area is considered to be a normal area.

[0013] In the second aspect, the present invention provides a city commercial center identification system based on big data, including a data acquisition module, a collaborative modeling module, a spatiotemporal feature weight generation module, and a business center identification module; the data acquisition module is used to collect multi-source data and perform preprocessing, and extract business features based on the preprocessed multi-source data; the collaborative modeling module is used to train a gradient boosting decision tree through multi-source data based on differential privacy and homomorphic encryption algorithms, and perform gradient aggregation, and at the same time perform collaborative modeling based on business features to generate a joint feature weight matrix; the spatiotemporal feature weight generation module is used to obtain the optimal spatiotemporal feature weight based on the joint feature weight matrix through a quantum annealing algorithm combined with classical simulated annealing tuning, and monitor the stability of the optimal spatiotemporal feature weight through a dynamic adjustment mechanism, and perform dynamic adjustment; the business center identification module is used to combine the optimal spatiotemporal feature weight and real-time status data of the road network, perform spatiotemporal kernel density calculation on the entire city grid through an adaptive kernel function, and identify the city commercial center according to the calculation results.

[0014] In a third aspect, the present invention provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the urban commercial center identification method based on big data as described in the first aspect of the present invention.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the urban commercial center identification method based on big data as described in the first aspect of the present invention.

[0016] The beneficial effects of the present invention are as follows: through collaborative modeling and gradient aggregation based on differential privacy and homomorphic encryption algorithms, the security of sensitive data during transmission and calculation is ensured, while supporting collaborative modeling of multi-party data; through spatiotemporal feature weight optimization based on quantum annealing algorithm combined with classical simulated annealing, global optimization and dynamic adjustment of feature weights are achieved; combined with data acquisition and preprocessing, dynamic adjustment mechanism and spatiotemporal kernel density calculation, the present invention realizes the accurate identification of urban commercial centers, providing a scientific basis for urban planning and management. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0018] Figure 1 This is a flowchart of the urban commercial center identification method based on big data in Example 1.

[0019] Figure 2 This is a schematic diagram of the city commercial center identification system based on big data in Example 1. DETAILED DESCRIPTION

[0020] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0021] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0022] Secondly, the term "one embodiment" or "embodiment" herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in various places throughout this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive of other embodiments.

[0023] Example 1, with reference to Figure 1 and Figure 2 , which is the first embodiment of the present invention, provides a method for identifying urban commercial centers based on big data, comprising the following steps: S1. Collect and preprocess multi-source data, and extract business features based on the preprocessed multi-source data.

[0024] Multi-source data includes mobile phone signaling data, UnionPay transaction data, traffic flow data, and commercial outlet data; It should be noted that mobile phone signaling data automatically records users' geographic location and communication activities through mobile operators' base stations. UnionPay transaction data is provided by banks or payment processors and includes information such as the time, location, and amount of consumer transactions using bank cards or mobile payments at various commercial locations. Traffic flow data is typically collected by a city's traffic management department or intelligent transportation system, including information such as vehicle traffic volume, speed, and congestion conditions. Commercial outlet data may be derived from business registration information, map service providers, or field surveys, and covers information such as the location, size, and business type of shops, restaurants, and other commercial facilities.

[0025] Preprocessing includes data cleaning, format conversion, deduplication, normalization and outlier processing.

[0026] It should be noted that the collected multi-source data, including mobile phone signaling data, UnionPay transaction data, traffic flow data, and commercial outlet data, is first cleaned to remove incomplete, erroneous, or irrelevant data records. Format conversion is then performed to ensure that all data sources adhere to a unified format standard for subsequent processing. Deduplication is then performed to eliminate duplicate data entries to ensure the uniqueness of the dataset. Normalization is used to map data values ​​from different sources to the same scale to prevent certain features from having an undue impact on the results due to large differences in magnitude. Finally, outlier processing is used to identify and correct data points that significantly deviate from the normal range to ensure the accuracy and reliability of the dataset.

[0027] Based on the commercial network data, the density characteristics of commercial network are extracted by kernel density estimation method; It should be noted that the location coordinates of all commercial outlets in the study area are first determined, and then the density of commercial outlets around each location point is calculated using the kernel density estimation method. By selecting appropriate bandwidth parameters, the distribution of commercial outlets in the entire urban area can be smoothly estimated, and finally the commercial outlet density characteristics are generated. Based on mobile phone signaling data, the characteristics of pedestrian flow are extracted through the node centrality analysis method in graph theory; It should be noted that the city is first divided into multiple grid cells. A network model is constructed based on user movement records between different grid cells from mobile phone signaling data. Each grid cell is considered a node, and edges represent user movement between grid cells. Node centrality metrics (such as betweenness centrality or closeness centrality) are then applied to assess the importance of each node to extract pedestrian flow characteristics.

[0028] Based on traffic flow data, the number of transportation hubs in the urban area is counted to extract the connectivity characteristics of transportation hubs; It should be noted that the locations of various transportation hubs in the city, including but not limited to bus stops, subway stations, etc., are identified and counted, and then the connection relationship between these transportation hubs and the service scope they can cover are analyzed, and the connectivity characteristics of the transportation hubs are calculated based on this.

[0029] Based on UnionPay transaction data, the transaction growth rate characteristics are extracted by calculating the changes in transaction amounts within a specific time period.

[0030] It should be noted that we first collect UnionPay transaction amount data from various commercial outlets over a period of time (such as a month or a quarter), then compare the total transaction amounts in adjacent time periods to calculate the transaction growth rate characteristics.

[0031] S2. Based on differential privacy and homomorphic encryption algorithms, the gradient boosting decision tree is trained through multi-source data and gradient aggregation is performed. At the same time, collaborative modeling is performed based on business features to generate a joint feature weight matrix.

[0032] Initialize the gradient boosting decision tree based on the preprocessed multi-source data; It should be noted that, first, a preprocessed dataset containing features such as commercial point density, foot traffic, transportation hub connectivity, and transaction growth rate is used as input. Next, an initial prediction value is selected, typically the average of the training sample labels, to construct the first base learner. These features are then used to train a series of regression decision trees based on the residual between the actual label and the current prediction. Each tree attempts to minimize the residual of the previous tree, improving accuracy by gradually reducing the residual. At each iteration, a weight is assigned to each new tree to correct previous errors, until a predetermined number of trees are reached and no significant improvement is achieved, ultimately forming an initialized gradient boosting decision tree.

[0033] The extracted business features are encrypted using a homomorphic encryption algorithm to generate an encrypted feature vector; It should be noted that, first, the pre-processed features of commercial outlet density, foot traffic, transportation hub connectivity, and transaction growth rate are prepared. These features are then encrypted using a homomorphic encryption algorithm. During the encryption process, each feature value is converted into an encrypted feature vector based on the selected homomorphic encryption scheme. This ensures that even when processing encrypted data, necessary computations can be performed without revealing the specific values.

[0034] Use the encrypted feature vector to train the initialized gradient boosting decision tree and obtain the gradient decision update value ; It should be noted that, first, the encrypted feature vectors generated by the homomorphic encryption algorithm, including the encrypted commercial network density features, pedestrian flow features, transportation hub connectivity features, and transaction growth rate features, are used as input data. Then, based on the initialized gradient boosting decision tree, these encrypted feature vectors are processed by a secure computing method supported by homomorphic encryption, so that the necessary addition and multiplication operations can be performed without decryption to evaluate the error at each node and determine the optimal split point. Based on this process, the residual between the predicted result and the actual label in each round of iteration is calculated, and the gradient boosting decision tree parameters are adjusted accordingly to obtain the gradient decision update value. For example, when processing data from a specific area, the encrypted commercial network density feature "0x1A2B3C4D" is securely calculated to determine its impact on the gradient boosting decision tree error, and the weights in the gradient boosting decision tree are updated accordingly, ultimately obtaining the gradient decision update value in the encrypted state. ,The privacy protection of data and the effectiveness of gradient boosting decision tree training are ensured throughout the process.

[0035] Use homomorphic encryption to update gradient decisions Encrypt, generate the gradient decision update value in the encrypted state, and upload it to the central server through a secure channel; It should be noted that, first, the gradient decision update value obtained from the gradient boosting decision tree training , applying a homomorphic encryption algorithm to convert it into an encrypted gradient decision update value, ensuring data privacy and security even during transmission. This encryption step renders the original gradient decision update value in ciphertext form, enabling certain types of computational operations without decryption. The encrypted gradient decision update value is then transmitted to a central server via a pre-established secure channel. This secure channel ensures confidentiality and integrity during data transmission, preventing unauthorized access and data leakage.

[0036] The central server decides the updated value based on the gradient in the encrypted state Aggregate the gradient and inject Gaussian noise through differential privacy to generate noisy aggregate gradient , the expression is: ; in, Represents the gradient decision update value The number of data holders, It is Gradient decision update value in encrypted state , is the gradient decision update value in the encrypted state The index variable of It should be noted that after receiving all encrypted gradient decision update values, the central server calculates the average aggregation result of these encrypted values. In order to ensure the data privacy of all parties involved, the aggregate gradient is perturbed by injecting Gaussian noise during the aggregation process to produce the final noisy aggregate gradient. This process allows for the effective training and parameter update of the gradient boosting decision tree while protecting the sensitive information of each data source. Throughout the entire process, the consistency of names such as encrypted gradient decision update values ​​is maintained to ensure the accuracy and security of the processing process. For example, when processing commercial feature data from different sources, the central server can securely aggregate these encrypted update values ​​and protect the original data from being leaked by adding noise, while providing support for subsequent analysis.

[0037] Aggregate gradient with noise via key pair in homomorphic encryption algorithm Decrypt and generate the global parameters of the gradient boosting decision tree; It should be noted that, first, the noisy aggregate gradient generated by the central server is encrypted using the preset homomorphic encryption algorithm key. Decryption is performed to restore the encrypted values ​​to their original form. This process ensures that even though Gaussian noise is added during data transmission and aggregation to protect privacy, the effective information used to adjust the gradient boosting decision tree can still be accurately recovered. The decrypted aggregated gradient retains the noise-treated properties, protecting the data privacy of all parties involved while providing information that can be used to optimize the gradient boosting decision tree.

[0038] Based on the global parameters of the gradient boosting decision tree, the business features are collaboratively modeled to generate a joint feature weight matrix ; It should be noted that, first, the decrypted global parameters of the gradient boosting decision tree, including information such as node splitting characteristics, splitting thresholds, and leaf node weights for each tree, serve as the basis for gradient boosting decision tree optimization. Next, the contribution of each business feature in the gradient boosting decision tree is adjusted based on these global parameters. Through collaborative modeling, the interactions between commercial network density, pedestrian flow, transportation hub connectivity, and transaction growth rate are comprehensively considered. During this process, the influence of each feature on the final prediction result is calculated, and a joint feature weight matrix W is generated based on this, reflecting the relative importance of all business features.

[0039] Data holders include operators, UnionPay, and transportation management departments in the city; The global parameters of the gradient boosting decision tree include the node splitting features, splitting thresholds, and leaf node weights of each tree in the gradient boosting decision tree.

[0040] S3. Based on the joint feature weight matrix, the optimal spatiotemporal feature weights are obtained through the quantum annealing algorithm combined with classical simulated annealing tuning. The stability of the optimal spatiotemporal feature weights is monitored and dynamically adjusted through the dynamic adjustment mechanism.

[0041] The joint feature weight matrix is ​​converted into Each element in corresponds to the spin variable of the quantum bit, and a Hamiltonian of the quadratic unconstrained binary optimization form is constructed, and the joint feature weight matrix Mapping to the objective function of the quantum annealing algorithm , the expression is: ; in, is the joint feature weight matrix The weight of the feature, It is The spin variable of the quantum bit corresponding to the feature, is the total number of features in the joint feature weight matrix, Indicates the Features and The correlation between the features, It is The spin variable of the quantum bit corresponding to the feature, The weight representing the correlation between feature weight and features; It should be noted that, first, the joint feature weight matrix is ​​identified The weight of each feature in the algorithm is calculated and associated with the corresponding qubit spin variable, while considering the correlation between features and the interaction between the qubit spin variables corresponding to these features. Based on this, an objective function is constructed that comprehensively reflects the influence of all feature weights and their interactions.

[0042] Each feature weight in the joint feature weight matrix corresponds to a quantum bit; Based on the objective function, set the initial parameters of the quantum annealing algorithm; The initial parameters of the quantum annealing algorithm include annealing time, number of qubits, and initial temperature; Based on the objective function and initial parameters of the quantum annealing algorithm, the annealing process is performed to gradually optimize the objective function and generate the feature weights after annealing optimization. , the expression is: ; The optimal feature weight combination is converted into a classical feature weight vector through quantum measurement and binary decoding; It should be noted that, first, after the quantum annealing algorithm completes optimization, the resulting quantum state is measured to obtain the specific values ​​of each qubit's spin variable. These qubit spin variable values ​​exist as quantum states, representing the optimal feature weight combination. Next, binary decoding is used to convert these quantum states into classical binary values, which are then mapped to the corresponding specific weight values ​​in the original feature space. In this way, the optimal feature weight combination originally represented in the quantum computing framework is converted into a classical feature weight vector that can be used in a classical computing environment.

[0043] Based on the quantum annealing results and the joint feature weight matrix, the objective function of classical simulated annealing is defined , the expression is: ; in, Indicates the The feature weights after annealing optimization, is the first in the joint feature weight matrix The weight of the feature, Indicates the joint feature matrix Features and The joint weight of the features, is the weight coefficient of the feature weight, the feature weight after annealing optimization, and the deviation of the joint feature weight matrix; It should be noted that the optimal feature weight combination obtained from the quantum annealing process is first compared with the corresponding weight value in the joint feature weight matrix. For each feature, the difference between its feature weight after quantum annealing optimization and the original weight in the joint feature weight matrix is ​​calculated, and these differences are squared and summed to form the error metric of the first part. At the same time, considering the interaction between features, for each pair of features, the difference between their joint weight in the joint feature weight matrix and the product of the feature weight after quantum annealing optimization is calculated. These differences are also squared and summed as the error metric of the second part. By introducing a weight coefficient to balance the influence of the two parts of the error, the objective function of the classical simulated annealing is constructed. .

[0044] Perform simulated annealing based on the objective function of classic simulated annealing to generate feature weights after simulated annealing optimization , the expression is: ; The feature weights after annealing optimization and the feature weights after simulated annealing optimization are combined to generate the optimal spatiotemporal feature weights. , the expression is: ; in, It is the fusion coefficient of the feature weight after annealing optimization and the feature weight after simulated annealing optimization.

[0045] It should be noted that, first, the feature weights after annealing optimization are calculated separately and feature weights after simulated annealing optimization The performance on the validation set is compared, for example, by comparing their accuracy, stability, and error in predicting the city's commercial center. Then, the model is dynamically adjusted based on these performance indicators. Specifically, if the feature weights after annealing optimization perform better on the validation set, then A higher value makes the final optimal spatiotemporal feature weight More inclined to the results of annealing optimization; on the contrary, if the feature weight after simulated annealing optimization performs better, then reduce Value, let It is closer to the result of simulated annealing optimization.

[0046] Calculate the standard deviation of the optimal spatiotemporal feature weights , the expression is: ; in, is the mean of all feature weights in the joint feature weight matrix; The standard deviation of the optimal spatiotemporal feature weights It is defined as the weight stability index; Define a stability threshold based on the standard deviation distribution of feature weights in historical data ; when > When , the optimal spatiotemporal feature weight is considered unstable, and the annealing time is increased and the number of quantum bits is increased; for example, when the standard deviation of the optimal spatiotemporal feature weight is monitored to exceed the preset stability threshold, it indicates that the current optimal spatiotemporal feature weight is unstable. At this time, by increasing the annealing time to 200 microseconds and increasing the number of quantum bits to 1000, the exploration ability of the optimization process can be enhanced, thereby helping to achieve a more stable optimal spatiotemporal feature weight.

[0047] when ≤ When , the optimal spatiotemporal feature weights are considered stable.

[0048] S4. Combining the optimal spatiotemporal feature weights and real-time road network status data, the spatiotemporal kernel density of the entire city grid is calculated using an adaptive kernel function, and the city's commercial centers are identified based on the calculation results.

[0049] Real-time road network status data including traffic flow, congestion index and road closure information; Divide the urban area into M×N grids, treat each grid as an independent area and assign a unique identifier; It should be noted that, first, the size of the grid is determined based on the city's geographic scope and the required accuracy of the analysis, namely M rows and N columns, to ensure that each grid covers an appropriately sized urban area. Next, each divided grid cell is numbered according to its geographic location order or other predetermined rules, ensuring that each grid has a unique identifier. This unique identifier can be generated based on the grid's location coordinates, row and column numbers, or other customized coding schemes, so that each grid area can be accurately located and identified in subsequent data processing and analysis operations. In this way, a discretized representation of urban space is achieved.

[0050] Using Gaussian kernel function as a framework, combined with dynamic bandwidth adjustment of real-time road network status data to form an adaptive kernel function , the expression is: ; in, is the horizontal coordinate of the grid center, is the vertical coordinate of the grid center, is the bandwidth, It is the first The horizontal coordinate of the data point, It is the first The vertical coordinate of the data point, is the index variable for data points in the urban area; It should be noted that, first, the city's geographic area is evenly divided, creating a grid system covering the entire city. Then, according to established rules, each grid cell is assigned a unique identifier. This identifier reflects the grid's location and order, ensuring accurate identification of each grid cell in subsequent data processing and analysis. This approach effectively discretizes urban space and facilitates location-based analysis.

[0051] Combine the adaptive kernel function with the time decay function to calculate the spatiotemporal kernel density of each region in the grid , the expression is: ; in, It is The spatiotemporal kernel density of the region, R is the total number of data points, Time decay coefficient, is the time difference between the current time and the data point time; It should be noted that, first, the entire city is evenly divided according to its geographic boundaries, forming a series of continuous grid cells. Next, a unique identifier is assigned to each grid cell according to a predetermined rule, ensuring that each grid cell is uniquely identified and located in the subsequent analysis. This identifier allows each grid cell to be accurately identified and located in the subsequent analysis.

[0052] It should also be noted that data points refer to discrete observation points used to represent the specific location and value of commercial characteristics (such as commercial outlets, passenger flow, transportation hubs, transaction growth rate, etc.) in spatiotemporal kernel density calculations; Based on the distribution of spatiotemporal kernel density in historical data, a high-density threshold K1 and a low-density threshold K2 are defined; when When ≥K1, the current area is considered to be the main commercial center; When K2≤ When <K1, the current area is considered to be a secondary commercial center; when When <K2, the current area is considered to be a normal area.

[0053] This embodiment also provides a city commercial center identification system based on big data, including: a data acquisition module, a collaborative modeling module, a spatiotemporal feature weight generation module, and a commercial center identification module; The data acquisition module is used to collect and preprocess multi-source data, and extract business features based on the preprocessed multi-source data; The collaborative modeling module is used to train gradient-boosted decision trees using multi-source data based on differential privacy and homomorphic encryption algorithms, perform gradient aggregation, and perform collaborative modeling based on business features to generate a joint feature weight matrix. The spatiotemporal feature weight generation module is used to obtain the optimal spatiotemporal feature weight based on the joint feature weight matrix through the quantum annealing algorithm combined with classical simulated annealing tuning. The module also monitors the stability of the optimal spatiotemporal feature weight through a dynamic adjustment mechanism and performs dynamic adjustments. The commercial center identification module is used to combine the optimal spatiotemporal feature weights and real-time road network status data, calculate the spatiotemporal kernel density of the entire city grid through an adaptive kernel function, and identify the city's commercial centers based on the calculation results.

[0054] This embodiment also provides a computer device, which is suitable for the case of a city commercial center identification method based on big data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute computer-executable instructions to implement the city commercial center identification method based on big data proposed in the above embodiment.

[0055] The computer device may be a terminal, comprising a processor, memory, a communication interface, a display, and an input device connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system and computer programs. The internal memory provides an environment for the operating system and computer programs stored in the non-volatile storage media. The communication interface of the computer device is used to communicate with external terminals via wired or wireless communication. Wireless communication may be achieved via Wi-Fi, a carrier network, NFC (near-field communication), or other technologies. The display of the computer device may be a liquid crystal display or an electronic ink display. The input device may be a touchscreen overlay on the display, buttons, a trackball, or a touchpad on the computer device housing, or an external keyboard, touchpad, or mouse.

[0056] This embodiment also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for identifying urban commercial centers based on big data as proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0057] In summary, the present invention ensures the security of sensitive data during transmission and calculation, while supporting collaborative modeling of multi-party data, through collaborative modeling and gradient aggregation based on differential privacy and homomorphic encryption algorithms; achieves global optimization and dynamic adjustment of feature weights through spatiotemporal feature weight optimization based on quantum annealing algorithm combined with classical simulated annealing; combines data acquisition and preprocessing, dynamic adjustment mechanism and spatiotemporal kernel density calculation, and realizes accurate identification of urban commercial centers, providing a scientific basis for urban planning and management.

[0058] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A method for identifying urban commercial centers based on big data, characterized by: include, Collect and preprocess multi-source data, and extract business features based on the preprocessed multi-source data; the multi-source data includes mobile phone signaling data, UnionPay transaction data, traffic flow data, and commercial network data; Based on differential privacy and homomorphic encryption algorithms, we train gradient boosting decision trees with multi-source data and perform gradient aggregation. We also conduct collaborative modeling based on business features to generate a joint feature weight matrix. Based on the joint feature weight matrix, the optimal spatiotemporal feature weights are obtained through the quantum annealing algorithm combined with classical simulated annealing tuning. The stability of the optimal spatiotemporal feature weights is monitored and dynamically adjusted through a dynamic adjustment mechanism. Combining the optimal spatiotemporal feature weights and real-time road network status data, the spatiotemporal kernel density of the entire city grid is calculated using an adaptive kernel function, and the city's commercial centers are identified based on the calculation results.

2. The method for identifying urban commercial centers based on big data according to claim 1, wherein: The preprocessing includes data cleaning, format conversion, deduplication, normalization and outlier processing.

3. The method for identifying urban commercial centers based on big data according to claim 2, wherein: The specific steps of extracting business features based on pre-processed multi-source data are as follows: Based on the commercial network data, the density characteristics of commercial network are extracted by kernel density estimation method; Based on mobile phone signaling data, the characteristics of pedestrian flow are extracted through the node centrality analysis method in graph theory; Based on traffic flow data, the number of transportation hubs in the urban area is counted to extract the connectivity characteristics of transportation hubs; Based on UnionPay transaction data, the transaction growth rate characteristics are extracted by calculating the changes in transaction amounts within a specific time period.

4. The method for identifying urban commercial centers based on big data according to claim 3, wherein: Based on differential privacy and homomorphic encryption algorithms, the gradient boosting decision tree is trained through multi-source data and gradient aggregation is performed. The specific steps are as follows: Initialize the gradient boosting decision tree based on the preprocessed multi-source data; The extracted business features are encrypted using a homomorphic encryption algorithm to generate an encrypted feature vector; Use the encrypted feature vector to train the initialized gradient boosting decision tree and obtain the gradient decision update value ; Use homomorphic encryption to update gradient decisions Encrypt, generate the gradient decision update value in the encrypted state, and upload it to the central server through a secure channel; The central server decides the updated value based on the gradient in the encrypted state Aggregate the gradient and inject Gaussian noise through differential privacy to generate noisy aggregate gradient ; Aggregate gradient with noise via key pair in homomorphic encryption algorithm Decrypt and generate the global parameters of the gradient boosting decision tree; Based on the global parameters of the gradient boosting decision tree, the business features are collaboratively modeled to generate a joint feature weight matrix ; The global parameters of the gradient boosting decision tree include the node splitting characteristics, splitting threshold and leaf node weight of each tree in the gradient boosting decision tree.

5. The method for identifying urban commercial centers based on big data according to claim 4, characterized in that: Based on the joint feature weight matrix, the optimal spatiotemporal feature weights are obtained by combining the quantum annealing algorithm with the classical simulated annealing tuning. The specific steps are as follows: The joint feature weight matrix is ​​converted into Each element in corresponds to the spin variable of the quantum bit, and a Hamiltonian of the quadratic unconstrained binary optimization form is constructed, and the joint feature weight matrix Mapping to the objective function of the quantum annealing algorithm ; Each feature weight in the joint feature weight matrix corresponds to a quantum bit; Based on the objective function, set the initial parameters of the quantum annealing algorithm; The initial parameters of the quantum annealing algorithm include annealing time, number of quantum bits and initial temperature; Based on the objective function and initial parameters of the quantum annealing algorithm, the annealing process is performed to gradually optimize the objective function and generate the feature weights after annealing optimization. ; The optimal feature weight combination is converted into a classical feature weight vector through quantum measurement and binary decoding; Based on the quantum annealing results and the joint feature weight matrix, the objective function of classical simulated annealing is defined ; Perform simulated annealing based on the objective function of classic simulated annealing to generate feature weights after simulated annealing optimization ; The feature weights after annealing optimization and the feature weights after simulated annealing optimization are combined to generate the optimal spatiotemporal feature weights. .

6. The method for identifying urban commercial centers based on big data according to claim 5, characterized in that: The dynamic adjustment mechanism is used to monitor the stability of the optimal spatiotemporal feature weights and perform dynamic adjustments. The specific steps are as follows: Calculate the standard deviation of the optimal spatiotemporal feature weights ; The standard deviation of the optimal spatiotemporal feature weight It is defined as the weight stability index; Define a stability threshold based on the standard deviation distribution of feature weights in historical data ; when > When , the optimal spatiotemporal feature weight is considered unstable, and the annealing time and the number of quantum bits are increased; when ≤ When , the optimal spatiotemporal feature weights are considered stable.

7. The method for identifying urban commercial centers based on big data according to claim 6, characterized in that: The method combines the optimal spatiotemporal feature weights and the real-time status data of the road network to calculate the spatiotemporal kernel density of the entire city grid through an adaptive kernel function, and identifies the city's commercial center based on the calculation results. The specific steps are as follows: The real-time status data of the road network includes traffic flow, congestion index and road closure information; Divide the urban area into M×N grids, treat each grid as an independent area and assign a unique identifier; Using Gaussian kernel function as a framework, combined with dynamic bandwidth adjustment of real-time road network status data to form an adaptive kernel function ; Combine the adaptive kernel function with the time decay function to calculate the spatiotemporal kernel density of each region in the grid ; Based on the distribution of spatiotemporal kernel density in historical data, a high-density threshold K1 and a low-density threshold K2 are defined; when When ≥K1, the current area is considered to be the main commercial center; When K2≤ When <K1, the current area is considered to be a secondary commercial center; when When <K2, the current area is considered to be a normal area.

8. A city commercial center identification system based on big data, based on the city commercial center identification method based on big data according to any one of claims 1 to 7, characterized in that: Including data acquisition module, collaborative modeling module, spatiotemporal feature weight generation module, and business center identification module; The data acquisition module is used to collect and preprocess multi-source data, and extract business features based on the preprocessed multi-source data; The collaborative modeling module is used to train gradient-boosted decision trees using multi-source data based on differential privacy and homomorphic encryption algorithms, perform gradient aggregation, and perform collaborative modeling based on business features to generate a joint feature weight matrix. The spatiotemporal feature weight generation module is used to obtain the optimal spatiotemporal feature weights based on the joint feature weight matrix through the quantum annealing algorithm combined with classical simulated annealing tuning. The module also monitors the stability of the optimal spatiotemporal feature weights through a dynamic adjustment mechanism and performs dynamic adjustments. The commercial center identification module is used to combine the optimal spatiotemporal feature weights and real-time road network status data, calculate the spatiotemporal kernel density of the entire city grid through an adaptive kernel function, and identify the city's commercial centers based on the calculation results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the urban commercial center identification method based on big data according to any one of claims 1 to 7 are implemented.

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 steps of the urban commercial center identification method based on big data according to any one of claims 1 to 7 are implemented.

Images