Commodity type and operation time analysis method for virtual business district

Abstract

The invention provides a commodity type and operation time analysis method for a virtual business district, and the method comprises the steps: obtaining the offline sounding data of the virtual business district, determining the influence degree of the offline sounding of the virtual business district, and forming an offline sounding influence matrix of the virtual business district; the method comprises the following steps: acquiring digital marketing data of a tourist attraction county, determining a pulling effect of digital marketing on commercial development of the tourist attraction county, and generating a digital marketing influence coefficient; the method comprises the following steps: obtaining shop data of main shops and non-main shops in a business district, performing category analysis on commodity categories, and defining complementarity of the main shops and the non-main shops in each dimension, thereby determining a complementarity score matrix, obtaining a network structure of complementarity scores, and obtaining a business form complementarity network model of the business district; and according to the business district business state complementarity network model, combining the association degree matrix of the virtual business district and the entity business district, the virtual business district offline sounding influence matrix and the digital marketing influence coefficient to generate a business gene map of the county business district.

Description

Technical Field

[0001] The present invention relates to the field of information technology, and in particular to a method for analyzing commodity categories and business hours in a virtual business district. Background Art

[0002] There are many technical difficulties and contradictions in the analysis of commodity types and operating hours in virtual business districts. First, there is a complex correlation between the commodity categories, operating hours, and business structure data of virtual business districts and physical business districts. How to accurately quantify the correlation and construct a correlation matrix based on the correlation is a technical challenge that needs to be solved urgently. Secondly, in the process of virtual business districts exploring offline, it is necessary to consider the influence indicators of passenger flow, sales, number of offline stores, and geographical distribution. How to achieve offline exploration of the influence matrix of virtual business districts through influence indicators is another technical difficulty. Furthermore, the digital marketing data of tourist scenic counties, extracting features from online promotion frequency and online transaction amount to determine the intensity of the driving effect of digital marketing on business development, and obtaining the digital marketing influence coefficient is a challenging technical problem. In addition, how to construct a complementary scoring matrix between main stores and non-main stores and obtain a complementary network model of business district formats is a key step in realizing the genetic map of county business districts. Finally, when constructing the commercial genetic map of county business districts, it is necessary to comprehensively consider metaverse embedding and scene simulation to realize the prediction of business district space planning and design, which is a technical problem that needs to be explored urgently. Summary of the invention

[0003] The present invention provides a method for analyzing commodity categories and business hours in a virtual business district, which mainly includes:

[0004] Obtain data on virtual business districts and physical business districts, and determine the correlation matrix between virtual business districts and physical business districts;

[0005] Obtain offline exploration data of virtual business districts, determine the impact of offline exploration of virtual business districts, and form an offline exploration influence matrix of virtual business districts;

[0006] Obtain digital marketing data of tourist scenic areas and counties, determine the driving effect of digital marketing on the commercial development of tourist scenic areas and counties, and generate digital marketing influence coefficients;

[0007] Obtain the store data of anchor stores and non-anchor stores in the business district, and define the complementarity of anchor stores and non-anchor stores in various dimensions by performing category analysis on product categories, thereby determining the complementarity scoring matrix, obtaining the network structure of the complementarity scoring, and obtaining the complementary network model of the business district formats;

[0008] Based on the complementary network model of business district formats, combined with the correlation matrix between virtual business districts and physical business districts, the offline influence matrix of virtual business districts, and the digital marketing influence coefficient, a commercial gene map of county business districts is generated.

[0009] Furthermore, the acquiring of virtual business district and physical business district data and determining a correlation matrix between the virtual business district and the physical business district includes:

[0010] Obtain commodity category, business hours and business structure data from the virtual business district transaction database and the physical business district operation database; perform encoding conversion on merchant location information to obtain spatial distribution data of virtual and physical business districts; conduct transaction volume and passenger flow statistics on the spatial distribution data of virtual and physical business districts to obtain business rule matching data of virtual and physical business districts; extract commodity category association strength and user location trajectory to obtain corresponding data of virtual and physical business district structures; construct commodity category association table, business period overlap table and location proximity table to obtain three-dimensional association data of virtual and physical business districts; perform calculations on the three-dimensional association data to obtain feature vectors, calculate the distance values ​​between feature vectors, and obtain a comprehensive association matrix of virtual and physical business districts.

[0011] Furthermore, the acquisition of offline exploration data of the virtual business district, determining the influence degree of offline exploration of the virtual business district, and forming an offline exploration influence matrix of the virtual business district include:

[0012] Receive store location coordinate information, customer flow data and sales records collected by radio frequency identification equipment; divide the business district into regions and calculate the density of the number of stores within the regional unit; collect regional customer flow data, perform spatial density mapping and curve fitting to obtain customer flow distribution data; extract the average daily transaction amount, perform spatial distribution mapping and normalization processing to obtain sales distribution data; construct a three-dimensional feature matrix, extract feature vectors and assign weight values ​​to obtain the offline exploration influence matrix of the business district.

[0013] Furthermore, the acquisition of digital marketing data of tourist scenic area counties, determination of the driving effect of digital marketing on the commercial development of tourist scenic area counties, and generation of digital marketing influence coefficients include:

[0014] Obtain advertising records, social communication data and search engine clicks in the marketing database, and classify and annotate the marketing data; divide the promotion frequency data into time clusters, and normalize the marketing reach data; obtain the product exposure, order records, and transaction amounts within the preset time sliding window, align timestamps and group aggregation to obtain transaction activity data; extract corresponding records, perform curve fitting on marketing input and sales output data, and obtain marketing pull relationship data; perform dimensionality reduction on the marketing pull relationship data, extract the principal components and assign weight values, and perform normalization to obtain the digital marketing influence coefficient.

[0015] Furthermore, the store data of the main stores and non-main stores in the business district are obtained, and the complementarity of the main stores and non-main stores in various dimensions is defined by performing category analysis on the commodity categories, thereby determining the complementary scoring matrix, obtaining the network structure of the complementary scoring, and obtaining the complementary network model of the business district, including:

[0016] The data of anchor stores and non-anchor stores are obtained from the business district database, including commodity category codes, sales records and business area data; the main feature dimensions are selected to normalize the commodity category structure data; the commodities are divided into a preset number of category clusters, and the complementary relationship between the category clusters is quantified; the shopping records of anchor stores and non-anchor stores are extracted to obtain the customer flow linkage relationship value between stores; an undirected weighted graph network is constructed, the network nodes are divided into a preset number of communities, the connection strength between nodes is determined, and a complementary scoring matrix in a standard format is obtained.

[0017] Furthermore, the business gene map of the county business district is generated based on the complementary network model of the business district formats, combining the correlation matrix between the virtual business district and the physical business district, the offline exploration influence matrix of the virtual business district, and the digital marketing influence coefficient, including:

[0018] Obtain the commodity category association data in the complementary business network and construct an association matrix; generate category association strength data and perform normalization to obtain normalized association strength values; perform weighted superposition operations to obtain feature dimension data; calculate the marketing communication intensity values ​​of business nodes and generate a node relationship diagram; perform community division, perform business mapping on community units, and obtain a business gene map.

[0019] Furthermore, generating a node relationship graph includes:

[0020] The marketing communication intensity value of the business node is calculated based on the characteristic dimension data; the force-directed layouter is used to generate a node relationship graph, and the node size is positively correlated with the marketing communication intensity value.

[0021] The technical solution provided by the embodiment of the present invention may have the following beneficial effects:

[0022] The present invention discloses a method for analyzing the types of goods and operating hours of a virtual business district. Aiming at the problem of the coordinated development of virtual and physical business districts in the planning of county business districts, a correlation matrix is ​​constructed by analyzing the characteristics of the commodity categories, operating hours, business structure, etc. of virtual and physical business districts, and the influence is evaluated by combining offline touch data and digital marketing data. At the same time, the commodity categories of the main stores and non-main stores in the business district are analyzed for complementarity, and a business format complementarity network model is established. Based on the above analysis results, a commercial gene map of the county business district is generated, and it is embedded in the metaverse space for business district scene simulation. Through multi-dimensional data analysis and model construction, the present invention realizes the coordinated planning of virtual and real business districts, effectively improves the scientificity and accuracy of business district planning, and provides digital decision-making support for county commercial development. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 The figure is a flow chart of a method for analyzing commodity categories and operating hours in a virtual business district according to the present invention.

[0024] Figure 2 It is a schematic diagram of a method for analyzing commodity categories and operating hours of a virtual business district according to the present invention.

[0025] Figure 3 It is another schematic diagram of a method for analyzing commodity categories and operating hours of a virtual business district according to the present invention. DETAILED DESCRIPTION

[0026] The following will describe the technical solutions in the embodiments of the present invention in detail in conjunction with the accompanying drawings in the embodiments of the present invention. The described embodiments are only part of the embodiments of the present invention.

[0027] like Figure 1-3 In this embodiment, a method for analyzing commodity types and business hours of a virtual business district may specifically include:

[0028] S101, obtaining commodity categories, business hours, and business structure data of the virtual business district and the physical business district, determining the correlation between the virtual business district and the physical business district, and obtaining a correlation matrix between the virtual business district and the physical business district.

[0029] According to the commodity category coding dictionary, the commodity identifier and transaction time record are obtained from the virtual business district transaction database, the store category code and business hour record are obtained from the physical business district operation database, and the merchant location information is encoded and converted through the geographic coordinate parser to obtain the spatial distribution data of the virtual business district and the physical business district; the time window sliding statistical tool is used to perform transaction volume statistics on the virtual business district spatial distribution data, and the passenger flow statistics on the physical business district spatial distribution data, and the business law matching data of the virtual business district and the physical business district are obtained through the density clustering algorithm; based on the order time series association tool, the commodity category association strength and user location trajectory are extracted from the business law matching data, and the virtual business district and the physical business district structure corresponding data are obtained through the cosine similarity calculator; according to the structural corresponding data, a commodity category association table, a business period overlap table, and a location proximity table are constructed to obtain the three-dimensional association data of the virtual business district and the physical business district; a non-negative matrix decomposition operation is performed on the three-dimensional association data to obtain a feature vector, and the distance value between the feature vectors is calculated by the Euclidean distance calculator to obtain the comprehensive association matrix of the virtual business district and the physical business district.

[0030] Specifically, we obtain product identifiers, transaction time records, and merchant location information from the virtual business district transaction database based on the product category coding dictionary, obtain store category codes, business hours records, and store geographic coordinates from the physical business district operation database, and convert the merchant addresses into geocodes through the Amap coordinate parser to obtain spatial distribution data of the two types of business districts. We use a 24-hour time window sliding statistics tool to extract the product category proportion, transaction volume per unit time, and consumer order peak period from the virtual business district transaction data, and extract the business format composition ratio, customer flow per unit time, and store business peak period from the physical business district operation data. We obtain the business law matching data of the virtual and physical business districts by density clustering. We use the order time series association tool to extract the product category association strength, user purchase sequence, and store area popularity from the virtual business district, and extract the business format complementarity, consumer location trajectory, and store concentration from the physical business district. We use the cosine similarity calculator to obtain the structural corresponding data of the two types of business districts. Three-dimensional association data is constructed from the commodity category association table, business period overlap table, and location proximity table of the virtual and physical business districts. The non-negative matrix decomposition tool is used to extract features from the association data. The distance value between the feature vectors is calculated based on the Euclidean distance calculator to obtain a comprehensive association matrix of the two types of business districts in three dimensions: category matching, time overlap, and location proximity. From the business district database, each commodity information contains the unique code of the commodity, the geographic coordinates of the store, and the time of transaction. The code adopts a tree classification standard. The commodity is classified into three levels of categories. Each level of category code is 6 digits, and the storage format is a string. The category information is matched according to the commodity code dictionary. In the address code conversion, the Amap API is used for geocoding. The address text is input and the longitude and latitude coordinate values ​​are returned, which are accurate to 6 decimal places. The spatial distance between the two business districts is calculated by the longitude and latitude values. The transaction data in the sliding time window is summarized. The window length is set to 1 hour and the sliding step is 0.5 hours. The number of transactions, transaction amount, and number of active merchants in each time window are recorded to obtain 24-hour time series data. During density clustering, data standardization is used to calculate the density distribution of each business format in the business district. Through order association analysis, the commodity association time threshold is set to 2 hours, and the association coefficient between commodity pairs is calculated, with the association coefficient range from 0 to 1. The location data is gridded, with a grid size of 0.01 degrees in longitude and latitude, and the number of merchants and transaction frequency in each grid are counted. When constructing three-dimensional association data, the values ​​of each dimension are normalized to the range of 0-1, and non-negative matrix decomposition is used to calculate three feature matrices. The feature dimension is set to the square root value of the number of commodity categories. Finally, the Euclidean distance formula is used to calculate the distance value between the feature vectors to obtain the association value between the two business districts. Each cell in the association matrix contains the values ​​of three dimensions: category similarity, time overlap, and location proximity.

[0031] Feature extraction is performed on the data of virtual and physical business districts, and the category matching degree, business hour correlation and business format overlap between virtual and physical business districts are calculated for correlation analysis and weight distribution. A correlation matrix is ​​constructed based on the correlation between commodity categories, business hours and business format structure characteristics, and the correlation and weight distribution between commodity categories, business hours and business format structure characteristics are filled in the matrix.

[0032] The virtual business district data is obtained from the commodity transaction database and the physical business district data is obtained from the store operation database. The spatial distribution grid matrix is ​​obtained through the Manhattan distance calculator based on the business district data, and the location mapping tool is used to perform regional matching on the grid matrix; according to the regional matching results, the time window tool is used to extract data on the proportion of virtual business district commodity categories and the proportion of physical business district business formats, and the K-means clustering tool is used to group the data extraction results into intervals; the Jaccard similarity coefficient calculator is used to process the interval grouped data to obtain the intersection of virtual business district commodity categories and the overlap of physical business formats; according to the preset commodity classification standards, the intersection of the commodity categories and the overlap of the business formats are mapped into categories, and the structural matching data is calculated through the vector inner product tool; for the structural matching data, the principal component analysis tool is used to perform feature dimensionality reduction, the dimensionality reduction results are normalized, and the business district correlation matrix is ​​constructed through the array dot product tool.

[0033] Specifically, the commodity category identification, transaction time point, and store distribution data of the virtual business district are obtained from the commodity transaction database. At the same time, the business format code, business time interval, and store location coordinates of the physical business district are obtained from the store operation database. The regional units are divided according to the 500-meter grid. The spatial distribution grid matrix of the two types of business districts is obtained by the Manhattan distance calculator, and the grid units are regionally matched using the location mapping tool. According to the preset unified commodity classification standard, the time window tool with a 1-hour window and a 30-minute step length is used to extract the commodity category proportion, hourly transaction volume, and consumer active period of the virtual business district, and the business format composition ratio, hourly passenger flow, and store business hours of the physical business district are processed. The business district operation data is grouped and classified by intervals using the K-means clustering tool. The Jaccard similarity coefficient calculator is used to extract the commodity category intersection, store density, and regional coverage from the virtual business district, and the business format overlap, store concentration, and business district radiation range are calculated from the physical business district. The two types of business districts are mapped according to the commodity code dictionary, and the vector inner product tool is used to calculate the structural matching data of the two types of business districts. The principal component analysis tool is used to reduce the dimension of the commodity category association table, business hours overlap data, and business format structure similarity data of the virtual business district and the physical business district. The weight coefficient interval is set between 0 and 1, and the three dimensions of category matching, time correlation, and business format overlap are normalized. The array dot product tool is used to construct a business district association matrix that meets the standard format. In the business district data processing, the virtual business district data record contains the unique identifier of the commodity, the transaction time stamp, and the geographical coordinates of the store. The physical business district data contains the store business code, business hours interval, and location longitude and latitude. The grid is divided into square area units with a side length of 500 meters. Each grid is assigned a unique number. The Manhattan distance is used to calculate the distance value between adjacent grids to obtain the grid association matrix. In the time window processing, the window size is set to 1 hour, and it slides backward every 30 minutes. The number of transactions, commodity category distribution, and number of consumers in each time window are counted, and the business status, customer flow changes, and business format proportion of each store in the physical business district are recorded. Commodity classification is coded with six digits, the first two digits represent the first-level category, the middle two digits represent the second-level category, and the last two digits represent the third-level category. K-means clustering is used to divide commodities into 10 clusters according to transaction rules. When calculating structural similarity, the Jaccard coefficient is used to calculate the overlap of commodity categories, with a value range of 0 to 1. The store density is expressed by the number of merchants per unit area, and the regional coverage is the actual coverage area divided by the theoretical maximum coverage area. The principal component analysis retains the eigenvectors with an explained variance ratio of 85%. The weight coefficient is normalized so that the sum of the weights of the three dimensions is 1. In the final association matrix, each cell contains data in three dimensions: category matching value, time overlap value, and business format similarity value.

[0034] S102. Obtain offline exploration data of the virtual business district, the offline exploration data including customer flow, sales, number of offline stores, and geographical location distribution, determine the impact of offline exploration of the virtual business district, and form an offline exploration influence matrix of the virtual business district.

[0035] Receive store location coordinate information, customer flow data and sales records collected by the radio frequency identification device, and the radio frequency identification device collects data at preset time intervals; use the hexagonal honeycomb grid tool to divide the business district into regions according to the store location coordinate information, and obtain regional units with a side length of fifty meters, and use the DBSCAN clusterer to calculate the density of the number of stores in the regional units; use infrared sensors to collect regional customer flow data, and use heat map tools to perform spatial density mapping on the customer flow data, and use spline interpolators to perform curve fitting on the customer flow data to obtain customer flow distribution data; extract the average daily transaction amount according to the preset time granularity based on the store sales records, and use the kernel density calculator to perform spatial distribution mapping on the average daily transaction amount, and obtain sales distribution data through maximum and minimum value normalization processing; use the store density calculation results, the customer flow distribution data and the sales distribution data to construct a three-dimensional feature matrix, and use the principal component analysis tool to extract feature vectors, assign weight values ​​to the feature vectors through matrix dot multiplication operations, and obtain the business district offline exploration influence matrix.

[0036] Specifically, the location coordinates, customer flow data, and sales records of offline stores are collected every 5 minutes through radio frequency identification devices. The business district is divided into regional units with a side length of 50 meters using a hexagonal honeycomb grid tool. The DBSCAN clusterer is used to calculate the density of the number of stores in the regional unit, and the store density distribution data is obtained according to the distance threshold of 10 meters. According to the customer flow data collected by the infrared sensor, the changes in the flow of people in the area, the length of consumer stay, and the customer movement trajectory are recorded at a spatial accuracy of 10 square meters. The heat map tool with a resolution of 25 meters is used to map the spatial density of the regional customer flow distribution, and the cubic spline interpolator is used to fit the curve of the customer flow data within the 24-hour time window. The store sales data is used to extract the average daily transaction amount, consumption frequency statistics, and commodity structure proportion at a time granularity of 1 hour. The kernel density calculator is used to map the spatial distribution of merchant transaction conditions within the 24-hour interval, and the maximum and minimum value normalization tool is used to obtain regional sales distribution data. A three-dimensional feature matrix was constructed from the store density data, customer flow change curve, and sales distribution data. The principal component analysis tool was used to extract the eigenvectors with a cumulative explained variance of 85%. The eigenvectors were assigned weight values ​​between 0 and 1 through matrix dot multiplication operations to obtain the offline touch-and-go influence matrix of the business district that meets the standard format. In the offline touch-and-go data collection of the business district, radio frequency identification devices were deployed at the entrances and exits of the business district and key node locations. Each device covered a radius of 50 meters, with a sampling interval of 5 minutes, and recorded customer identification numbers, timestamps, and location coordinates. The hexagonal honeycomb grid divides the business district into regional units with a side length of 50 meters. Each unit is assigned a unique number. The DBSCAN clustering parameter sets the density threshold to 10 stores per unit area and a neighborhood radius of 10 meters. Infrared sensors are deployed in front of shops with a spatial resolution of 10 square meters, and the flow of people is recorded once a minute. The heat map tool uses a 25-meter grid size to interpolate the density of human traffic. Colors from blue to red represent density from low to high, and the value range is 0 to 100. Cubic spline interpolation takes one data point per hour in a 24-hour window to form a smooth customer flow change curve. Sales data are counted at a 1-hour time granularity, including the number of transactions, transaction amount, and product category. The bandwidth parameter of kernel density calculation is set to 50 meters. The feature matrix contains three dimensions: the density dimension takes values ​​from 0 to 100 to represent the store concentration, the customer flow dimension takes values ​​from 0 to 1000 to represent the total hourly customer flow, and the sales dimension takes values ​​from 0 to 1 to represent the normalized transaction intensity. The principal component analysis retains the first three eigenvalues, and the corresponding cumulative explained variance ratio is 85%. The weight values ​​of the three dimensions are set to 0.4, 0.35, and 0.25, respectively, and the final influence value is obtained by matrix dot multiplication.

[0037] S103. Obtain digital marketing data of tourist scenic counties, including online promotion frequency and online transaction volume, determine the driving effect of digital marketing on the commercial development of tourist scenic counties, and generate a digital marketing influence coefficient.

[0038] Obtain advertising records, social communication data, and search engine clicks from the marketing database, and perform three-level classification and annotation of marketing data according to channel source, promotion form, and target audience to obtain labeled data; for the labeled data, use the K-means clusterer to divide the promotion frequency data into four time clusters, and normalize the marketing reach data in each time cluster through the maximum and minimum mapping method to obtain normalized marketing data; obtain the product exposure, order record, and transaction amount within the preset time sliding window from the transaction database, align the transaction data with the preset time step, group and aggregate the transaction data according to the product category identifier, and perform grouping and grouping of transactions within the window. The transaction activity data is obtained by mapping the maximum and minimum values ​​of the normalized marketing data and the timestamps of the transaction activity data; the corresponding records are extracted according to the timestamp pairing of the normalized marketing data and the transaction activity data, and the marketing input and sales output data are curve fitted by a cubic polynomial fitting tool. If the goodness of fit is greater than a preset threshold, the marketing pull relationship data is obtained by a Pearson correlation coefficient calculator; for the marketing pull relationship data, a principal component extractor is used to perform dimensionality reduction processing. If the cumulative contribution rate reaches 85%, the first three principal components are extracted, and weight values ​​in the range of zero to one are assigned to the principal components through matrix dot multiplication operations. The weighted results are normalized to obtain the digital marketing influence coefficient.

[0039] Specifically, we obtain the advertising records, social communication data, and search engine clicks of tourist scenic counties from the marketing database, classify and annotate the marketing data at three levels according to the channel source, promotion form, and target audience, set four time clusters using the K-means clusterer to divide the promotion frequency data into time periods, and perform maximum and minimum value normalization on the marketing reach data in each time period. We extract the product exposure, order records, and transaction amounts within a 24-hour sliding window from the transaction database, align the transaction data with timestamps using a 1-hour step, group and aggregate the transaction data according to the product category identifier, and map the transaction values ​​within the window to the maximum and minimum values ​​to obtain the transaction activity in the interval of 0 to 1. We extract the corresponding records from the marketing reach data and transaction activity values ​​according to the timestamp pairing, use the cubic polynomial fitting tool to curve fit the marketing input and sales output data, set the goodness of fit threshold to 0.8, and use the Pearson correlation coefficient calculator to obtain the marketing pull relationship data. The principal component extractor is used to reduce the dimension of the marketing-driven relationship data, and the cumulative contribution rate is set to 85% to extract the first three principal components. The principal components are assigned weight values ​​in the range of 0 to 1 through matrix dot multiplication operations, and the weighted results are normalized to obtain the digital marketing influence coefficient. In the digital marketing data processing of tourist scenic spots, marketing data is divided into three categories according to the source of channels: search engines, social platforms, and vertical websites. The promotion forms include graphic advertisements, short videos, and live broadcasts. The target audience is divided into three dimensions: age, interest preferences, and consumption capacity. K-means clustering divides the 24 hours of the day into four time clusters: 0-6 am, 6-12 am, 12-18 pm, and 18-24 pm. The marketing data of each time cluster is mapped to the 0-1 interval through the maximum and minimum values. The transaction data uses a 24-hour sliding window with a step size of 1 hour. The number of product displays, clicks, and transactions within the window are recorded. The products are marked according to the four categories of catering, accommodation, scenic spot tickets, and specialty products. The transaction activity is normalized by the difference between the current value minus the minimum value and the maximum and minimum values. The marketing reach data and transaction data are matched according to the timestamp accurate to the minute. The corresponding relationship curve between marketing input and sales output is obtained by cubic polynomial fitting. The goodness of fit is determined by the R square value, the threshold is set to 0.8, and the Pearson correlation coefficient in the range of -1 to 1 represents the strength and direction of the correlation. The principal component extraction retains the components with eigenvalues ​​greater than 1, and the cumulative explained variance reaches 85%. The three extracted principal component vectors are assigned weights of 0.5, 0.3, and 0.2, and the marketing influence coefficient in the range of 0 to 1 is obtained by vector dot multiplication.

[0040] Key features are extracted from the digital marketing data of tourist scenic counties, and the key features are standardized. Based on the standardized key features, an evaluation model for the impact of digital marketing on the commercial development of tourist scenic counties is constructed. The intensity of the pulling effect of digital marketing on commercial development is calculated based on the output results of the model. The pulling effect intensity is standardized and used as the basis for generating the digital marketing influence coefficient.

[0041] The promotion frequency records, marketing reach numbers, and product exposure data in the digital marketing database are obtained, and the principal component extractor is used to extract the principal components of the marketing data to obtain the principal component characteristics; based on the principal component characteristics, the marketing feature data is normalized by the mean variance normalizer to obtain the standardized marketing characteristics; the sliding window tool is used to perform time series statistics on the product sales volume, transaction amount, and customer flow change data in the business development database, and the standardized business development data is obtained by the Z score normalizer; based on the standardized marketing characteristics and the standardized business development data, a random forest regressor is used for cross-validation, and the correlation strength value is obtained by the Pearson correlation coefficient calculator; if the goodness of fit corresponding to the correlation strength value is greater than a preset threshold, the business development pulling effect value is calculated; for the pulling effect value, a maximum and minimum normalization processor is used to perform standardization to obtain a standardized pulling effect value; based on the standardized pulling effect value, a weight coefficient is assigned by the hierarchical analysis method to obtain a digital marketing influence coefficient.

[0042] Specifically, we extract promotion frequency records, marketing reach, and product exposure from the digital marketing database, and use the principal component extractor to retain the first three principal components, whose cumulative explained variance contribution rate exceeds 85%. We use the mean variance normalizer to normalize the marketing feature data, and sort the standardized data in descending order according to the feature importance score. We obtain product sales volume, transaction amount, and customer flow change data from the business development database, and use a sliding window tool with a 24-hour window and a step length of 1 hour to perform time series statistics on business development indicators. We use the Z-score normalizer to normalize the statistical data, and align the business development data with the marketing feature data according to the timestamp index. We use a random forest regressor with 100 trees and a depth of 5 to perform a five-fold cross-validation on the standardized marketing feature data and business development data, and use the Pearson correlation coefficient calculator to obtain the correlation strength value between -1 and 1. We calculate the driving effect of business development for data with a goodness of fit greater than 0.8. The pull effect values ​​between 0 and 100 are standardized using the maximum and minimum normalization processor, and the standardized pull effect values ​​are assigned weight coefficients between 0 and 1 using the hierarchical analysis method. The weighted results are summed to obtain the digital marketing influence coefficient. In the digital marketing feature extraction, the marketing data includes three dimensions: the promotion frequency per hour, the number of users reached, and the product exposure. The principal component extraction shows that the first principal component accounts for 50% to reflect the marketing investment intensity, the second principal component accounts for 25% to indicate the user reach, and the third principal component accounts for 15% to represent the product coverage depth. The cumulative explained variance of the three reaches 90%. The three principal components are standardized using the mean variance method, with the mean set to 0 and the variance set to 1. The business development data uses a 24-hour sliding window, and is counted every hour to record the total sales volume, average transaction amount, and peak customer flow in the window. After data standardization, it is paired with marketing features according to timestamps. The random forest regressor sets 100 decision trees, each with a maximum depth of 5 layers. The data set is cross-validated in five folds, with 80% of the training set and 20% of the validation set in each fold. The Pearson correlation coefficient is used to calculate the correlation strength between marketing investment and business development. The absolute value of the correlation coefficient greater than 0.6 indicates a significant correlation. The pull effect value ranges from 0 to 100, indicating the degree of influence of marketing on business development. The maximum and minimum value normalization process is used to obtain the standardized pull effect value. The three dimensions are assigned weights of 0.5, 0.3, and 0.2 respectively through the hierarchical analysis method, and the weighted results are summed to obtain the marketing influence coefficient between 0 and 1.

[0043] S104. Obtain store data of anchor stores and non-anchor stores in the business district. The store data includes commodity categories. By performing category analysis on commodity categories, the complementarity between anchor stores and non-anchor stores in various dimensions is defined, thereby determining a complementary scoring matrix, obtaining a network structure of complementary scoring, and obtaining a complementary network model of business district formats.

[0044] The data of anchor stores and non-anchor stores whose business areas exceed the area threshold are obtained from the business district database, and the anchor stores and non-anchor stores data include four-level commodity category codes, sales records and business area data; according to the anchor stores and non-anchor stores data, the principal component extractor is used to select the first three main feature dimensions, and the commodity category structure data of the two types of shops are obtained through the maximum and minimum value normalizer; for the commodity category structure data, the K-means clusterer is used to divide the commodities into a preset number of category clusters, and the complementary relationship between the category clusters is quantified through the complement ratio calculator. If the complement ratio exceeds the preset threshold, it is determined that the category combination is complementary; according to the complementarity of the category combination, the association rule miner is used to extract the shopping records of the anchor stores and non-anchor stores within the time window, and the passenger flow linkage relationship value between the stores is obtained through the cosine similarity calculator; an undirected weighted graph network is constructed for the passenger flow linkage relationship value, and the Louvain community discovery algorithm is used to divide the network nodes into a preset number of communities, and the connection strength between the nodes is determined through the relative weight calculator to obtain a complementarity scoring matrix in a standard format.

[0045] Specifically, the four-level commodity category codes, sales records, and business area data of anchor stores and non-anchor stores with a business area of ​​more than 300 square meters are obtained from the business district database. The principal component extractor is used to select the first three main feature dimensions, and the business data is mapped to the interval of 0 to 1 through the maximum and minimum value normalizer to obtain the commodity category structure data of the two types of stores. The store commodities are classified and labeled according to the four-level commodity category system, and the K-means clusterer is used to divide the commodities into 8 category clusters. The complementary relationship between different category combinations is quantified by the complement ratio calculator. If the complement ratio exceeds 0.6, the combination is judged to be complementary, and the complementary degree value of the category combination is obtained. The association rule miner with a support threshold of 0.1 is used to extract consumer shopping records of anchor stores and non-anchor stores within a 15-minute time window, and the cosine similarity calculator is used to measure the passenger flow linkage relationship between stores, and the complementary score is mapped to the interval of 0 to 1 through the maximum and minimum value normalizer. An undirected weighted graph network is constructed for complementary scores with values ​​between 0.3 and 1. The Louvain community discovery algorithm is used to divide the network nodes into 3 to 8 communities. The connection strength between nodes is determined by the relative weight calculator, and the complementary score results are rounded to two decimal places to obtain a complementary score matrix in a standard format. In the process of building a complementary network of business districts, a four-level classification system is adopted for commodity categories. The first level includes catering, retail, and services, the second level is subdivided into 30 subcategories, the third level includes 150 subcategories, and the fourth level includes 600 specific categories. Each store is divided according to the business area, and those exceeding 300 square meters are judged as anchor stores. Principal component extraction retains three principal components that explain 85% of the variance, representing business scale, category richness, and customer unit price level. Commodity categories are divided into 8 clusters through K-means clustering, including fresh fruits and vegetables, snacks and beverages, daily necessities, clothing and shoes, beauty and personal care, digital appliances, catering and food, and leisure and entertainment. The complement ratio is calculated by the category difference between different categories. A value above 0.6 indicates complementarity. Within the 15-minute time window, the support threshold of association rule mining is set to 0.1, and the confidence threshold is 0.5. The shopping sequence of consumers in different stores is extracted. Stores with a cosine similarity value of more than 0.3 are judged to have a customer flow linkage relationship. In the network structure, nodes represent stores, and edge weights represent complementary scores. The Louvain algorithm is used to divide the network into 5 communities. The modularity of the division is greater than 0.4, and the connectivity between communities is less than 0.2. The values ​​in the final complementary score matrix retain two decimal places, and the value range is between 0 and 1.

[0046] Obtain the product categories of the main stores and non-main stores in the business district, conduct product category analysis, classify the products according to the preset product categories, analyze the market share of each category and the corresponding sales data of each category, and compare the coverage of the main stores and non-main stores in each product category, identify the strengths of the main stores and non-main stores in different product categories, and use this as the basis for defining the complementarity of the main stores and non-main stores in various dimensions.

[0047] Obtain the product codes, sales records and business categories of the main stores and non-main stores in the business district database, classify and label the product data into eight categories through the K-means clusterer, and obtain the category structure data of catering, clothing, shoes and hats, daily necessities, beauty and personal care, digital appliances, leisure and entertainment, fresh fruits and vegetables, snacks and beverages; calculate the percentage of sales of each category in the total sales of the business district based on the category structure data, use the kernel density calculator to count the number of stores of the main stores and non-main stores within the preset range, and obtain the market distribution data of the number of categories, sales share and number of stores through the Z-score normalizer; according to the market Distribution data, a radar chart numerical calculator is used to calculate the coverage of main stores and non-main stores in each commodity category, and the category difference data of the two types of stores in category structure are obtained through the Euclidean distance calculator; commodity categories with Z scores greater than 2 are extracted from the category difference data as their respective advantage areas, and a cosine similarity calculator is used to quantify the complementary advantage relationship between the two types of stores in different categories to obtain complementary advantage data; if the complementary advantage data is greater than a preset threshold, an array merger is used to construct a complementary definition table of the corresponding relationship between the advantageous categories to obtain the category complementary advantage structure of the main store and the non-main store.

[0048] Specifically, we obtain the product codes, sales records, and business categories of anchor stores and non-anchor stores from the business district database, classify and label the product data according to the four-level product category system of main category, major category, medium category, and minor category, and use the K-means clusterer to divide the products into 8 main categories, including catering, clothing, shoes, and hats, daily necessities, beauty and personal care, digital appliances, leisure and entertainment, fresh fruits and vegetables, and snacks and beverages. We obtain the category structure data through the maximum and minimum value normalizer. According to the sales data of the product category, we calculate the percentage of sales of each category in the total sales of the business district, and use the kernel density calculator to count the number of anchor stores and non-anchor stores within every 100 square meters. We use the Z-score normalizer to normalize the market share data of each category, and obtain market distribution data including the number of categories, sales share, and number of stores. The radar chart numerical calculator is used to calculate the coverage of the main store and the non-main store in each commodity category. The Euclidean distance calculator is used to quantify the difference in the category structure between the two types of stores. The matrix dot product operator is used to obtain the category difference data with two decimal places. From the category difference data, the commodity categories with Z scores greater than 2 are extracted as their respective advantage areas. The cosine similarity calculator is used to quantify the complementary relationship between the two types of stores in different categories. The complementarity threshold is set to 0.6, and the array merger is used to construct a complementary definition table containing the corresponding relationship of the advantage categories. In the complementary analysis of the commodity categories in the business district, the four-level commodity categories include 8 main categories, catering and food account for 25% of the total sales, clothing and shoes account for 20%, daily necessities account for 15%, beauty and personal care account for 12%, digital appliances account for 10%, leisure and entertainment account for 8%, fresh fruits and vegetables account for 6%, and snacks and beverages account for 4%. Each main category has 4 major categories, each of which contains 5 middle categories, and the middle category is further subdivided into 10 small categories to form a hierarchical classification system. The density distribution of anchor stores and non-anchor stores uses 100 square meters as the statistical unit. The spatial distribution heat map is obtained through kernel density calculation. The density of anchor stores in the digital appliance area reaches 0.8, and the density of clothing and shoes area reaches 0.7, while the density of non-anchor stores in the catering and food area is 0.9, and the density of fresh fruit and vegetable area is 0.8. The commodity coverage is represented by eight dimensions of the radar chart, with a value range of 0 to 1. The anchor stores are 0.9 in the digital appliance dimension and 0.8 in the clothing dimension. The non-anchor stores are 0.85 in the catering dimension and 0.75 in the fresh dimension. The category difference is calculated by Euclidean distance. The difference between the two types of stores in the digital category is 0.8, and the difference in the catering category is 0.7. When the Z score corresponding to the difference exceeds 2, it is determined to be a dominant category. The complementarity is calculated by cosine similarity. The category combination with a value greater than 0.6 is determined to be a complementary relationship. The final complementarity definition table contains the category correspondence and complementary strength value.

[0049] S105. Based on the complementary network model of business district formats, combined with the correlation matrix between virtual business districts and physical business districts, the offline influence matrix of virtual business districts, and the digital marketing influence coefficient, a commercial gene map of county business districts is generated.

[0050] Obtain commodity category association data in the complementary network of business formats, wherein the commodity category association data includes commodity category information of main stores and non-main stores, and construct an association matrix according to the commodity category association data; generate category association strength data according to the association matrix through a matrix dot multiplication operator, and normalize the category association strength data to obtain a normalized association strength value; for the normalized association strength value, use an offline probing influence matrix to perform a weighted superposition operation, and obtain feature dimension data through a Tucker tensor decomposer, and the cumulative explained variance of the feature dimension data reaches a preset threshold; calculate the marketing communication intensity value of the business format node according to the feature dimension data, and use a force-directed layouter to generate a node relationship graph, in which the node size is positively correlated with the marketing communication intensity value; for the node relationship graph, use the Louva in community discovery algorithm to divide the community, and use an attribute aggregator to perform business format mapping on the community unit, and the business format mapping determines the community coordinate position based on a two-dimensional plane layouter to obtain a business gene map.

[0051] Specifically, the commodity category association data of the main store and non-main store are extracted from the complementary network of the formats, and the association matrix is ​​constructed according to the number of commodity categories. The matrix dot product tool is used to calculate the association matrix and the complementary network data, and the array merger is used to generate the combined expression data containing the category association strength, and the association strength value is normalized in the range of 0 to 1. The weighted superposition of the business district format structure data is carried out according to the offline exploration influence matrix, and the Tucker tensor decomposer is used to retain the feature dimensions with a cumulative explained variance of 85%. The maximum and minimum normalizer is used to map the feature vector to the range of 0 to 1, and the offline influence value is converted into the structural relationship weight through the linear transformer. The marketing communication intensity of the format node is calculated from the digital marketing influence coefficient, and each node is assigned a weight value in the range of 0 to 1 through the weighted average operator. The force-directed layouter is used to generate a relationship diagram with a positive correlation between the node size and the weight. The thickness of the edge represents the degree of association, and the color depth represents the communication intensity. The node relationship graph is divided into 3 to 8 communities using Louvain community discovery algorithm. Similar business combinations are mapped to community units through attribute aggregators. The community structure is located using a two-dimensional plane layout device. The distance between communities indicates the similarity of business formats, and a hierarchical business gene map is obtained. In the process of constructing the business gene map, the commodity categories include 8 categories, forming an 8x8 association matrix. The values ​​of the matrix cells represent the strength of association between categories, and the value range is normalized to the interval of 0-1. Taking the catering and food category as an example, the association strength with fresh fruits and vegetables is 0.8, with snacks and beverages is 0.7, and the association strength with other categories is between 0.3 and 0.5. The offline exploration influence is extracted through Tucker decomposition to extract three main characteristic dimensions. The first dimension explains 40% of the variance, reflecting the spatial distribution characteristics, the second dimension 30% represents the time series characteristics, and the third dimension 15% represents the transaction behavior characteristics. The marketing communication intensity is mapped to the node weight value, with the catering category weight 0.9, the retail category 0.8, and the service category 0.7. The force-directed graph layout is adopted. The node size is set to 20 to 100 pixels in diameter according to the weight. The edge thickness is set to 1 to 5 pixels according to the association strength. The color from light blue to dark blue indicates the transmission strength from 0 to 1. The nodes are divided into 5 communities through the Louvain algorithm. The node spacing within the community is set to

[0052] 50 pixels, the distance between communities is set to 200 pixels, and the distance between communities with high similarity is closer. The final map is displayed on a two-dimensional plane of 1000x1000 pixels, and each community is marked with a background color of different depths to form a hierarchical commercial gene structure.

[0053] S106. Embed the commercial gene map of the county business district into the metaverse space, construct business district scenes, simulate the combination of business formats, commodity categories, and operating hours of the business district, predict the agglomeration effect of the business district under different plans, and form a business district space planning and design plan.

[0054] The business distribution data, category combination data and business period data in the business gene map are obtained, and a three-dimensional grid unit is constructed according to the business distribution data; the node data in the business gene map is converted into a three-dimensional entity through a spatial mapper, and the position is adjusted according to the spatial coordinates of the three-dimensional entity by a layout optimizer; the category combination data is calculated similarity by a spatiotemporal associator, and if the similarity is greater than a preset threshold, the business period data of the corresponding category is matched, and the matching results are spatially allocated by a grid division generator; a Monte Carlo simulator is used to calculate the density of each square meter of area, and the density calculation obtains the crowd density value through iteration, and the spatial distribution of the crowd density value is visualized by a heat map generator; a planning scheme is generated for the spatial distribution visualization result by a multi-objective optimizer, and the planning scheme is optimized and calculated based on the passenger flow maximization goal, the area restriction goal and the proximity constraint goal, and the planning scheme is quantitatively ranked by a weighted sum scorer to obtain a business district space planning plan that meets the agglomeration effect.

[0055] Specifically, the distribution of business formats, commodity category combinations, and business hours in the business district are obtained from the business gene map. A virtual business district scene with a 10m×10m grid unit is constructed using a three-dimensional rectangular coordinate system. The map nodes are converted into three-dimensional entities through a spatial mapper, and the area of ​​a single business format is limited to 300 square meters. The entity is located at a proximity radius of 50 meters using a layout optimizer. Commodity categories are combined and matched according to the business district aggregation data, and the business format similarity threshold is set to 0.6. The time-space associator is used to match the business hours of different categories. The business format combination is spatially allocated through a grid division generator, and a customer flow route map in the business district is generated based on the shortest path algorithm. The Monte Carlo simulator is used to perform 1,000 iterative calculations, and the crowd density per square meter is numerically simulated. The crowd aggregation threshold is set to 0.8. The spatial distribution data is visualized using a heat map generator to obtain passenger flow density distribution data at different times. The simulation data is optimized for spatial planning using a multi-objective optimizer. Three objectives are set: passenger flow maximization, area restriction, and proximity constraint. The planning schemes are quantitatively ranked by a weighted sum scorer, and 10 business district space planning plans that meet the agglomeration effect are limited. In the implementation of the business district space planning, the virtual scene is modeled using a three-dimensional rectangular coordinate system. The spatial range is 1000 meters × 1000 meters × 50 meters. It is divided into 10,000 basic units according to a 10-meter × 10-meter grid. Each grid records three attributes: business type, business period, and passenger flow density. Business entities are laid out according to an area limit of 300 square meters. The distance between adjacent entities is kept within 50 meters, and the height of the entity is set between 10 and 30 meters according to the business type. The commodity category combination is matched based on a similarity threshold of 0.6. The business hours are divided into four periods according to 24 hours: 6-11 am in the morning market, 11-14 pm in the afternoon market, 14-17 pm in the afternoon market, and 17-22 pm in the evening market. The shortest path algorithm is used to generate a walking route network, with the path width set to 3 meters and the distance between nodes not exceeding 20 meters. The Monte Carlo simulation is set to 1,000 iterations, with each iteration simulating a passenger flow of 100-1,000 people. The crowd density threshold is set to 0.8 people per square meter, and the density distribution is displayed through a heat map, with colors from blue to red representing density values ​​from 0 to 1. The passenger flow weight in the optimization target is 0.5, the area constraint weight is 0.3, and the proximity weight is 0.2. The 10 planning schemes generated are sorted from high to low according to the weighted score, with the score range between 0 and 100.

[0056] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for analyzing commodity categories and business hours in a virtual business district, characterized in that: The method comprises: Obtain data on virtual business districts and physical business districts, and determine the correlation matrix between virtual business districts and physical business districts; Obtain offline exploration data of virtual business districts, determine the impact of offline exploration of virtual business districts, and form an offline exploration influence matrix of virtual business districts; Obtain digital marketing data of tourist scenic areas and counties, determine the driving effect of digital marketing on the commercial development of tourist scenic areas and counties, and generate digital marketing influence coefficients; Obtain the store data of anchor stores and non-anchor stores in the business district, and define the complementarity of anchor stores and non-anchor stores in various dimensions by performing category analysis on product categories, thereby determining the complementarity scoring matrix, obtaining the network structure of the complementarity scoring, and obtaining the complementary network model of the business district formats; Based on the complementary network model of business district formats, combined with the correlation matrix between virtual business districts and physical business districts, the offline influence matrix of virtual business districts, and the digital marketing influence coefficient, a commercial gene map of county business districts is generated.

2. The method for analyzing commodity categories and business hours of a virtual business district according to claim 1, characterized in that: The step of obtaining the data of the virtual business district and the physical business district and determining the correlation matrix between the virtual business district and the physical business district includes: Obtain commodity categories, business hours, and business structure data from the virtual business district transaction database and the physical business district operation database; Encode and convert the merchant location information to obtain the spatial distribution data of the virtual business district and the physical business district; The transaction volume and customer flow statistics of the spatial distribution data of virtual business districts and physical business districts are carried out to obtain the matching data of the business rules of virtual business districts and physical business districts; Extract the correlation strength of commodity categories and user location trajectories to obtain the corresponding data of the virtual business district and the physical business district structure; Construct commodity category association table, business period overlap table, and location proximity table to obtain three-dimensional association data between virtual and physical business districts; The three-dimensional correlation data is operated to obtain the feature vector, and the distance value between the feature vectors is calculated to obtain the comprehensive correlation matrix between the virtual business district and the physical business district.

3. The method for analyzing commodity categories and business hours of a virtual business district according to claim 1, characterized in that: The step of obtaining offline exploration data of a virtual business district, determining the influence degree of offline exploration of a virtual business district, and forming an offline exploration influence matrix of a virtual business district includes: Receive store location coordinate information, customer flow data and sales records collected by RFID equipment; Divide the business district into regions and calculate the density of stores within the regional units; Collect regional passenger flow data, perform spatial density mapping and curve fitting, and obtain passenger flow distribution data; Extract the average daily transaction amount, perform spatial distribution mapping and normalization processing, and obtain sales distribution data; Construct a three-dimensional feature matrix, extract feature vectors and assign weight values ​​to obtain the offline influence matrix of the business district.

4. The method for analyzing commodity categories and business hours of a virtual business district according to claim 1, characterized in that: The obtaining of digital marketing data of tourist scenic areas and counties, determining the driving effect of digital marketing on the commercial development of tourist scenic areas and counties, and generating digital marketing influence coefficients include: Obtain advertising records, social communication data, and search engine clicks from the marketing database, and classify and annotate the marketing data; Divide promotion frequency data into time clusters and normalize marketing reach data; Obtain product exposure, order records, and transaction amounts within a preset time sliding window, perform timestamp alignment and group aggregation, and obtain transaction activity data; Extract the corresponding records, perform curve fitting on the marketing input and sales output data, and obtain the marketing-driven relationship data; The marketing-driven relationship data is processed for dimensionality reduction, the principal components are extracted and weight values ​​are assigned, and then normalized to obtain the digital marketing influence coefficient.

5. The method for analyzing commodity categories and business hours of a virtual business district according to claim 1, characterized in that: The store data of the main stores and non-main stores in the business district are obtained, and the complementarity of the main stores and non-main stores in various dimensions is defined by performing category analysis on the commodity categories, thereby determining the complementary scoring matrix, obtaining the network structure of the complementary scoring, and obtaining the complementary network model of the business district, including: Obtain data on anchor stores and non-anchor stores from the business district database, including product category codes, sales records, and business area data; Select the main feature dimensions and normalize the product category structure data; Divide the products into a preset number of category clusters and quantify the complementary relationships between the category clusters; Extract shopping records of anchor stores and non-anchor stores to obtain the customer flow linkage relationship values ​​between stores; An undirected weighted graph network is constructed, the network nodes are divided into a preset number of communities, the connection strength between nodes is determined, and the complementarity scoring matrix in a standard format is obtained.

6. The method for analyzing commodity categories and business hours of a virtual business district according to claim 1, characterized in that: The commercial gene map of the county business district is generated based on the complementary network model of the business district format, combined with the correlation matrix of the virtual business district and the physical business district, the offline exploration influence matrix of the virtual business district, and the digital marketing influence coefficient, including: Obtain the commodity category related data in the complementary business network and construct a correlation matrix; Generate category association strength data, and perform normalization processing to obtain normalized association strength values; Perform weighted superposition operations to obtain feature dimension data; Calculate the marketing communication intensity value of the business format node and generate a node relationship diagram; Divide the community, map the business formats of the community units, and obtain a business gene map.

7. The method for analyzing commodity categories and business hours of a virtual business district according to claim 6, characterized in that: The generating of the node relationship graph comprises: Calculate the marketing communication intensity value of the business node based on the characteristic dimension data; A force-directed layouter is used to generate a node relationship graph, and the node size is positively correlated with the marketing communication intensity value.

Images