Traffic site screening method and system for marketing and storage medium

Abstract

The invention relates to the technical field of data processing, and discloses a traffic site screening method and system for marketing and a storage medium. The method comprises the following steps: collecting multi-dimensional data through a monitoring system, and calculating a quality evaluation index; extracting user features to establish a portrait, and calculating a matching degree; predicting an effect in combination with operation data; screening sites based on prediction evaluation; and monitoring and analyzing the screening result, and dynamically adjusting to obtain an optimization scheme. According to the application, multi-dimensional evaluation of shopping mall quality, accurate matching of user portraits and dynamic optimization of screening results are realized, so that scientificity and success rate of merchant site selection decision are improved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a method, system and storage medium for screening traffic sites for marketing. Background Art

[0002] With the rapid development of the commercial real estate market, shopping malls, as important retail and consumption places, their site selection decisions have a crucial impact on the business results of merchants. Traditional methods for screening traffic sites in shopping malls mainly rely on manual experience judgment and simple passenger flow statistics data. When choosing a store location, merchants often only consider surface factors such as basic passenger flow and rent level, lacking in-depth analysis of customer behavior characteristics and shopping mall quality. In the prior art, there are already some data collection systems that can record the passenger flow data of shopping malls, and there are also some shopping mall management software that can provide simple passenger flow analysis functions. However, most of these systems operate independently and fail to form a complete data analysis chain.

[0003] However, there are obvious deficiencies in the application of these prior arts. First, the data collection dimension is single, and most systems only focus on passenger flow while ignoring the quality of passenger flow. Second, there is a lack of accurate construction of user portraits, and it is impossible to effectively evaluate the matching degree between the customer groups in shopping malls and the needs of target merchants. Third, there is no objective site quality evaluation system, resulting in overly subjective site selection decisions. Finally, the screening results lack a dynamic monitoring and optimization mechanism and cannot cope with market and passenger flow changes. These problems lead to insufficient scientificity and accuracy in merchants' site selection decisions and increase business risks. Summary of the Invention

[0004] This application provides a method, system and storage medium for screening traffic sites for marketing, which realizes multi-dimensional evaluation of shopping mall quality, accurate matching of user portraits and dynamic optimization of screening results, thereby improving the scientificity and success rate of merchants' site selection decisions.

[0005] First aspect, the present application provides a method for screening traffic sites for marketing. The method for screening traffic sites for marketing includes: collecting multi-dimensional data on the total number of visitors, number of independent visitors, stay duration, and bounce rate of a shopping mall traffic site through deploying a site monitoring system, and storing them in a labeled manner according to source channels, devices, and regions to obtain an initial evaluation matrix; calculating a basic passenger flow score, a user stickiness coefficient, a user engagement coefficient, and a site attribute coefficient based on the initial evaluation matrix, and obtaining a site quality evaluation database through weighted calculation; extracting the device type distribution, access time distribution, regional distribution, and consumption preference characteristics of users according to the quality evaluation database, establishing a site user portrait, and calculating the matching score of the target merchant user portrait; obtaining a predicted score of the site operation effect based on the matching score, combined with characteristics such as passenger flow, conversion rate, and consumption amount in historical operation data, as well as the site quality score, business hours, and promotion activity form; conducting a quality access evaluation and a user matching evaluation on the site based on the predicted score of the operation effect, setting a grading standard according to the expected effect, and screening in combination with the merchant site selection magnitude threshold to obtain a site screening combination result; monitoring and analyzing the quality evaluation indicators, user portrait matching degree, and actual operation effect of the site according to the site screening combination result, and making dynamic adjustments through early warning triggering and effect comparison to obtain a shopping mall site optimization plan.

[0006] Second aspect, the present application provides a system for screening traffic sites for marketing. The system for screening traffic sites for marketing includes:

[0007] A collection module, configured to collect multi-dimensional data on the total number of visitors, number of independent visitors, stay duration, and bounce rate of a shopping mall traffic site through deploying a site monitoring system, and store them in a labeled manner according to source channels, devices, and regions to obtain an initial evaluation matrix;

[0008] A calculation module, configured to calculate a basic passenger flow score, a user stickiness coefficient, a user engagement coefficient, and a site attribute coefficient based on the initial evaluation matrix, and obtain a site quality evaluation database through weighted calculation;

[0009] An extraction module, configured to extract the device type distribution, access time distribution, regional distribution, and consumption preference characteristics of users according to the quality evaluation database, establish a site user portrait, and calculate the matching score of the target merchant user portrait;

[0010] A prediction module, configured to obtain a predicted score of the site operation effect based on the matching score, combined with characteristics such as passenger flow, conversion rate, and consumption amount in historical operation data, as well as the site quality score, business hours, and promotion activity form;

[0011] An evaluation module, configured to perform quality access evaluation and user matching degree evaluation on a venue based on the predicted score of the business effect, set a grading standard according to the expected effect, and perform screening in combination with the merchant site selection magnitude threshold to obtain a venue screening combination result;

[0012] A monitoring module, configured to monitor and analyze the quality evaluation index, user portrait matching degree and actual business effect of the venue according to the venue screening combination result, and perform dynamic adjustment through early warning triggering and effect comparison to obtain a shopping mall venue optimization plan.

[0013] A third aspect of the present invention provides a computer-readable storage medium, in which instructions are stored, and when the instructions run on a computer, the computer is enabled to execute the above-mentioned traffic venue screening method for marketing.

[0014] In the technical solution provided by this application, multi-dimensional data collection is carried out through the deployment of a site monitoring system, achieving comprehensive monitoring of key indicators such as the total number of mall visitors, the number of independent visitors, the stay duration, and the bounce rate, breaking through the limitation of traditional methods that only focus on single passenger flow. With the help of tagged storage technology, attributes such as source channels, devices, and regions are structured and organized to form an initial evaluation matrix rich in information, laying a solid data foundation for subsequent analysis. Based on this evaluation matrix, the method introduces multi-dimensional evaluation indicators such as basic passenger flow scores, user stickiness coefficients, user engagement coefficients, and site attribute coefficients, and generates a site quality evaluation database through weighted calculation, realizing the scientific quantification of mall quality. Further, the device type distribution, access time distribution, regional distribution, and consumption preference characteristics of users are extracted from the quality evaluation database to establish a three-dimensional site user portrait, and the matching score with the target merchant user portrait is calculated through an intelligent matching algorithm, effectively solving the problem that it is difficult for traditional methods to accurately evaluate the customer group matching. On this basis, combined with the characteristics of passenger flow, conversion rate, consumption amount, etc. in historical business data, as well as factors such as site quality score, business hours, and promotion activity forms, a prediction model is applied to generate a predicted score of the site operation effect, greatly improving the scientificity and accuracy of site selection decisions. Based on the predicted score, the method designs a dual screening mechanism of quality access evaluation and user matching evaluation, and introduces the expected effect grading standard and merchant site selection magnitude threshold as constraint conditions to obtain the optimal site screening combination result. Finally, by establishing a monitoring and analysis closed-loop, the quality evaluation indicators, user portrait matching degree, and actual operation effect of the selected site are continuously monitored, and dynamic adjustment is carried out based on the warning trigger mechanism and effect comparison analysis to generate a mall site optimization plan, effectively coping with the challenges brought by changes in the market environment and passenger flow characteristics. The entire solution makes full use of artificial intelligence algorithms to deeply mine data and identify patterns. Especially in the aspects of user portrait construction, matching degree calculation, and operation effect prediction, advanced algorithms such as clustering analysis, similarity calculation, and regression prediction are introduced. These algorithm features play a key role in improving data processing accuracy, discovering potential association patterns, and achieving accurate prediction, significantly enhancing the efficiency and accuracy of mall traffic site screening. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0016] Figure 1 FIG. is a schematic diagram of an embodiment of a method for screening traffic sites for marketing in an embodiment of this application;

[0017] Figure 2 This is a schematic diagram of an embodiment of a traffic venue screening system for marketing in an embodiment of the present application. Detailed implementation manners

[0018] The embodiments of the present application provide a traffic venue screening method, system and storage medium for marketing. Terms such as "first", "second", "third", "fourth", etc. (if any) in the specification, claims and above-mentioned drawings of the present application are used to distinguish similar objects, and do not have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so that the embodiments described here can be implemented in an order other than those illustrated or described here. In addition, the terms "include" or "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units does not have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0019] For ease of understanding, the specific process of the embodiments of the present application will be described below. Please refer to Figure 1 An embodiment of the traffic venue screening method for marketing in the embodiments of the present application includes:

[0020] Step S101: Collect multi-dimensional data on the total number of visitors, the number of independent visitors, the stay duration, and the bounce rate of the mall traffic venue through the deployed venue monitoring system, and store them in a labeled manner according to the source channel, device, and region to obtain an initial evaluation matrix;

[0021] Step S102: Calculate the basic passenger flow score, user stickiness coefficient, user participation coefficient, and venue attribute coefficient according to the initial evaluation matrix, and obtain a venue quality evaluation database through weighted calculation;

[0022] Step S103: Extract the device type distribution, access time distribution, regional distribution, and consumption preference characteristics of users according to the quality evaluation database, establish a venue user portrait, and calculate the matching score of the target merchant user portrait;

[0023] Step S104: According to the matching score, combine features such as the passenger flow, conversion rate, and consumption amount in the historical operation data, as well as the venue quality score, business hours, and promotion activity forms, to obtain the predicted score of the venue operation effect;

[0024] Step S105: Based on the predicted score of the operation effect, conduct a quality access evaluation and user matching evaluation on the venue, set a grading standard according to the expected effect, and combine the merchant site selection magnitude threshold for screening to obtain the venue screening combination result;

[0025] Step S106: According to the venue screening combination result, monitor and analyze the quality evaluation indicators, user portrait matching degree, and actual operation effect of the venue, and perform dynamic adjustment through early warning triggering and effect comparison to obtain the shopping mall venue optimization plan.

[0026] It can be understood that the execution subject of this application can be a traffic venue screening system for marketing, or a terminal or a server. Specifically, it is not limited here. This application example is described with the server as the execution subject.

[0027] Specifically, data is collected by deploying a venue monitoring system. The venue monitoring system includes devices such as passenger flow counters, Wi-Fi probes, and facial recognition cameras. These devices are distributed at the shopping mall entrance, main passageways, and key areas on each floor. The passenger flow counter records the number of people entering and leaving. The Wi-Fi probe captures the mobile phone MAC address to identify independent visitors. The facial recognition camera tracks the movement path of customers in the shopping mall. Through these technical means, the system can accurately collect key indicators such as the total number of visitors, the number of independent visitors, the stay duration, and the bounce rate. For example, for a shopping mall with an area of 50,000 square meters, the average daily passenger flow is about 25,000 person-times, the number of independent visitors is about 18,000 people, the average stay duration is 95 minutes, and the bounce rate (the proportion of those who leave after staying briefly in a single area) is 30%. The system stores the data in a tagged manner according to the source channel (such as subway drainage, surrounding communities, online activity promotion), the type of access device (such as smartphones, wayfinding query devices), and the user's region (based on IP address resolution or membership registration information) to form an initial evaluation matrix.

[0028] Next, multiple key indicators are calculated based on the initial evaluation matrix. The basic passenger flow score is a weighted combined value after standardizing the total number of visitors and the number of independent visitors, reflecting the basic traffic level of the venue. The user stickiness coefficient is calculated by comparing the stay duration with the industry standard value, indicating the customer's willingness to stay at the venue. The user engagement coefficient is converted from the bounce rate data, reflecting the customer's activity level at the venue. The venue attribute coefficient synthesizes data from three dimensions: the credibility of the source channel, the device coverage range, and the regional distribution. These four indicators are weighted and calculated to form a venue quality evaluation database, providing a basis for subsequent analysis.

[0029] Based on the quality assessment database, the system analyzes and extracts user characteristics. The distribution of device types reflects the proportion of consumers using mobile terminals, guide devices and membership cards; the distribution of visit time periods shows the time pattern of peak and off-season periods; the regional distribution shows the geographical breadth of customer sources; and the consumption preference determines which type of goods or services customers prefer by analyzing purchase records. These characteristics together constitute the user portrait of the venue. At the same time, the target merchants provide standard user portraits based on the characteristics of their own customer groups. The system quantifies the matching score by calculating the Euclidean distance between the two portraits in each dimension to determine the degree of fit between the venue customer group and the merchant's target audience. The matching score is combined with historical operating data to predict the effect. Historical operating data includes indicators such as passenger flow, conversion rate (the proportion of actual consumers to total passenger flow), and average customer price. The system also considers factors such as venue quality score, business hour characteristics, and promotion activity forms (such as discount type, intensity and scope) to establish a prediction model and derive a venue operation effect prediction score, which is used to evaluate the potential performance of merchants after entering the venue.

[0030] Based on the predicted scores, the system performs quality access assessment and user matching assessment. The quality access assessment sets the minimum entry threshold to screen out venues with qualified basic conditions; the user matching assessment ensures that the customer group is highly consistent with the needs of the merchants; the expected effect classification divides the venues into three levels: high, medium and low; the delivery level threshold sets the site selection range according to the merchant budget. After cross-validation of these standards, a combination of venue screening results is formed.

[0031] The system continuously monitors the performance of the venue and optimizes it dynamically. The monitoring content includes the fluctuation of quality evaluation indicators, changes in user portraits, and the difference between actual operating results and expectations. When the indicators exceed the warning threshold, the system automatically triggers the warning mechanism. Through effect comparison and dynamic adjustment, the optimization plan is continuously improved. The whole process forms a closed loop to ensure that the shopping mall traffic venue screening results always meet the needs of merchants and market changes. For example, a brand clothing store is looking for a new store location, and the system analyzes the data of 50 shopping malls in the city. Through the initial evaluation matrix analysis, 30 shopping malls with qualified basic customer flow were selected. Combined with the brand's target customer group characteristics (18-35 years old, middle and high income, preference for fashion products) and the user portraits of each shopping mall, the matching degree is calculated, and the scope is further narrowed to 15 shopping malls. After analyzing the historical operating data, the system predicts the 5 most promising venues and continuously monitors the performance indicators of these shopping malls to ensure the timeliness and accuracy of the recommendation results.

[0032] In the embodiments of the present application, by deploying a site monitoring system for multi-dimensional data collection, comprehensive monitoring of key indicators such as the total number of mall visitors, the number of unique visitors, the stay duration, and the bounce rate is achieved, breaking through the limitation of traditional methods that only focus on single passenger flow. With the help of tagged storage technology, attributes such as source channels, devices, and regions are structured and organized to form an initial evaluation matrix rich in information, laying a solid data foundation for subsequent analysis. Based on this evaluation matrix, the method introduces multi-dimensional evaluation indicators such as the basic passenger flow score, user stickiness coefficient, user engagement coefficient, and site attribute coefficient, and generates a site quality evaluation database through weighted calculation to achieve scientific quantification of the mall quality. Further, the device type distribution, access time period distribution, regional distribution, and consumption preference characteristics of users are extracted from the quality evaluation database to establish a three-dimensional site user portrait, and the matching score with the target merchant user portrait is calculated through an intelligent matching algorithm, effectively solving the problem that it is difficult for traditional methods to accurately evaluate the customer group matching. On this basis, combined with the characteristics of passenger flow, conversion rate, consumption amount, etc. in historical operation data, as well as factors such as site quality score, business hours, and promotion activity forms, a prediction model is applied to generate a prediction score for the site operation effect, greatly improving the scientificity and accuracy of site selection decisions. Based on the prediction score, the method designs a dual screening mechanism of quality access evaluation and user matching evaluation, and introduces the expected effect grading standard and merchant site selection magnitude threshold as constraint conditions to obtain the optimal site screening combination result. Finally, by establishing a monitoring and analysis closed-loop, continuous monitoring of the quality evaluation indicators, user portrait matching degree, and actual operation effect of the selected site is carried out, and dynamic adjustment is made based on the early warning trigger mechanism and effect comparison analysis to generate a mall site optimization plan, effectively coping with the challenges brought by changes in the market environment and passenger flow characteristics. The entire solution makes full use of artificial intelligence algorithms to deeply mine data and identify patterns. Especially in the aspects of user portrait construction, matching degree calculation, and operation effect prediction, advanced algorithms such as clustering analysis, similarity calculation, and regression prediction are introduced. These algorithm features play a key role in improving data processing accuracy, discovering potential association patterns, and achieving accurate prediction, significantly enhancing the efficiency and accuracy of mall traffic site screening.

[0033] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0034] (1) Collect the passenger flow records of each mall traffic site within a specified time window through the monitoring system to obtain the total number of visitors to the mall traffic site;

[0035] (2) De-duplicate the user identification information in the total number of visitors to the mall traffic site to obtain the number of unique visitors to the mall traffic site;

[0036] (3) By recording the entry time and exit time of users in the mall traffic area, calculating the time difference for each visit, and obtaining the residence duration in the mall traffic area;

[0037] (4) Calculating the ratio of the number of users who only visit a single area in the mall traffic area to the total number of visiting users to obtain the bounce rate of the mall traffic area;

[0038] (5) Extracting the source channel identifier based on the information at the time of user access and classifying it according to online drainage, surrounding communities, business cooperation, and brand activities to obtain the source channel data of the mall traffic area;

[0039] (6) By parsing the device identifier information of users, classifying and counting according to mobile terminals, mall guidance devices, and membership cards to obtain the access device data of the mall traffic area;

[0040] (7) Performing geographical location parsing based on user location information and conducting distribution statistics according to provinces, cities, and regions to obtain the user geographical data of the mall traffic area;

[0041] (8) Associating the source channel data, access device data, and user geographical data of the mall traffic area in multiple dimensions, establishing a label index system, and obtaining the initial evaluation matrix.

[0042] Specifically, by means of a monitoring system, passenger flow records of the mall traffic area within a specified time window are collected to obtain the total number of visitor data. Here, the monitoring system refers to the collection of hardware devices such as passenger flow counting devices, wireless probes, and cameras deployed at each entrance, main passage, and area of the mall. The specified time window is usually the daily operating period, such as from 10:00 am to 10:00 pm, and can also be set to other time dimensions such as weeks or months according to needs. The passenger flow record refers to the original data such as the time point when customers enter the mall and their movement trajectories. The total number of visitors refers to the total number of person-times entering the mall within the time window, including repeated entries. The user identification information in the total number of visitors to the mall traffic area is de-duplicated to obtain the number of unique visitors. The user identification information includes, but is not limited to, the mobile phone MAC address captured by the Wi-Fi probe, the Bluetooth device ID, the membership card number, and the temporary ID generated by the face recognition system, etc. The de-duplication process uses algorithms such as hash table mapping to merge multiple visits of the same user into one record, thereby calculating the number of different customers who actually visited within the specified time window, that is, the number of unique visitors. By recording the entry time and exit time of users in the mall traffic area, the time difference of each visit is calculated to obtain the stay duration. The entry time refers to the time stamp when the customer is first captured by the monitoring system, and the exit time refers to the time stamp when the customer is last captured by the system. The time difference is calculated by subtracting the time stamps, and the unit is usually minutes. For customers who enter and exit the mall multiple times, each entry and exit will be calculated separately and finally averaged. The stay duration reflects the customer's willingness to stay in the mall and the shopping experience, and is an important indicator for evaluating the attractiveness of the venue. The ratio of the number of users who only visit a single area in the mall traffic area to the total number of visiting users is calculated to obtain the bounce rate. A single area refers to the functional partition divided within the mall, such as the first floor lobby, elevator entrance, a certain counter, etc. Only visiting a single area means that after entering the mall, the customer only stays briefly in one area and leaves without further exploring other areas. The bounce rate is calculated by dividing the number of users visiting a single area by the total number of visiting users, which reflects the degree of response of customers to the overall attractiveness of the mall. The higher the ratio, the weaker the mall's ability to attract and retain customers.

[0043] Extract the source channel identifier based on the information source channel when the user visits, and classify it according to online drainage, surrounding communities, business cooperation, and brand activities to obtain the source channel data. The source channel identifier can be obtained from various channels: the source page of scanning the mall QR code, the distribution channel of using coupons, the channel mark during membership registration, etc. Online drainage refers to the channels that guide customers to the store through online platforms, such as social media promotion, search engine advertising, etc.; the surrounding community refers to the natural passenger flow in the residential and office areas around the mall; business cooperation refers to the passenger flow guided by joint marketing with other enterprises; brand activities refer to the passenger flow attracted by promotional activities, exhibitions, performances, etc. held by the mall. The source channel data is of great significance for understanding the customer acquisition channels and evaluating the marketing effect.

[0044] By parsing the device identification information of users, classifying and counting according to mobile terminals, mall wayfinding devices, and membership cards, access device data is obtained. The device identification information includes attributes such as device type, brand, and model. Mobile terminals refer to the smart phones used by customers, which can be captured through Wi-Fi probes or Bluetooth signals; mall wayfinding devices refer to facilities such as interactive screens and inquiry machines in the mall; membership cards include physical membership cards and electronic membership cards. The access device data reflects the ways and technical preferences of customers' interactions with the mall, and has guiding significance for optimizing service facilities and enhancing the user experience. By performing geographical location parsing on the user location information and conducting distribution statistics according to provinces, cities, and regions, user geographical data is obtained. The location information can be obtained from sources such as IP addresses, mobile phone GPS signals, and membership registration information. Geographical location parsing uses technologies such as GeoIP to map the location information into the administrative division system. The user geographical data can show the geographical source distribution of the mall customer group, and has important value for understanding the radiation range and influence of the mall.

[0045] Perform multi-dimensional association on the source channel data, access device data, and user geographical data of the mall traffic site, establish a tag index system, and obtain an initial evaluation matrix. Multi-dimensional association means associating the above various types of data according to the user ID to form a complete user access portrait. The tag index system is a structured data organization method that classifies and indexes user attributes and behavior characteristics in the form of tags. The initial evaluation matrix is a multi-dimensional data table, where the rows represent different users or user groups, the columns represent various index characteristics, and the values in the matrix represent the performance or attributes of specific users on specific characteristics.

[0046] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0047] (1) Normalize the total visitor data and unique visitor data in the initial evaluation matrix respectively, and perform a linear combination according to the ratio of 6:4 to obtain the basic passenger flow score;

[0048] (2) Convert the stay duration data in the initial evaluation matrix into minutes, and calculate through the ratio relationship with the industry average stay duration to obtain the user stickiness coefficient;

[0049] (3) Perform reverse conversion on the bounce rate data in the initial evaluation matrix, and calculate the deviation degree from the industry benchmark bounce rate to obtain the user engagement coefficient;

[0050] (4) According to the source channel data in the initial evaluation matrix, perform weighted credibility scoring on each channel to obtain the source channel credibility;

[0051] (5) According to the access device data in the initial evaluation matrix, count the number and distribution ratio of covered device types to obtain the device coverage range;

[0052] (6) According to the user geographical data in the initial evaluation matrix, calculate the concentration degree and coverage breadth of geographical distribution to obtain the geographical distribution;

[0053] (7) Perform a linear combination of the source channel credibility, device coverage range, and geographical distribution to obtain the site attribute coefficient;

[0054] (8) Perform a weighted summation operation on the basic passenger flow score, user stickiness coefficient, user participation coefficient, and site attribute coefficient to obtain the site quality evaluation database.

[0055] Specifically, normalizing the total visitor volume data and independent visitor data in the initial evaluation matrix is to eliminate the differences between data of different magnitudes. The normalization process uses the maximum-minimum normalization method, that is, mapping the original data to the interval [0,1]. Specifically, for the total visitor volume data, take the maximum and minimum values in the mall historical data or industry standards as reference points, and calculate the relative position of the current total visitor volume within this range; the same method is used to process the independent visitor data. After both are normalized, a linear combination is performed according to the weight ratio of 6:4 to form the basic passenger flow score. The 6:4 ratio here reflects that the number of independent visitors is more important than the total number of visitors in the mall evaluation, because the number of independent visitors can better reflect the actual covered customer group size of the mall.

[0056] Converting the stay duration data in the initial evaluation matrix to minutes is to unify the measurement standard for easy calculation and comparison. The stay duration in the original data may be recorded in seconds or hours and needs to be uniformly converted to minutes. Calculate the user stickiness coefficient through the ratio relationship with the industry average stay duration. The industry average stay duration is a benchmark value determined according to factors such as mall type, scale, and positioning. The user stickiness coefficient reflects the performance of customers' willingness to stay in this mall compared with the industry standard. The higher the coefficient, the stronger the mall's attractiveness and customer retention ability.

[0057] Performing reverse conversion on the bounce rate data in the initial evaluation matrix is because the bounce rate is a negative indicator, that is, the higher the bounce rate, the lower the site attractiveness. The purpose of reverse conversion is to convert it into a positive indicator for easy weighted calculation together with other indicators. The specific calculation formula is:

[0058]

[0059] Among them, PEC represents the Participation Engagement Coefficient, BR represents the Bounce Rate of the current mall, BR base represents the industry benchmark bounce rate, BR max represents the theoretical maximum bounce rate (usually set at 100%), and λ represents the adjustment coefficient used to control the sensitivity of the conversion. When the mall bounce rate is equal to the industry benchmark value, the PEC value is 1; when the mall bounce rate is higher than the industry benchmark value, the PEC value is less than 1; when the mall bounce rate is lower than the industry benchmark value, the PEC value is greater than 1. This calculation method ensures that the user participation engagement coefficient can accurately reflect the relative performance of the mall in retaining customers.

[0060] Based on the source channel data in the initial evaluation matrix, a weighted credibility score is given to each channel. The values of different source channels are different. For example, customers attracted through brand activities usually have higher brand loyalty and consumption willingness, while the natural customer flow in the surrounding community is more stable. The weighted credibility score is the weight coefficient set according to factors such as the stability, conversion rate, and customer value of each channel. By multiplying the proportion of customer flow of each channel by its weight coefficient and summing them up, a source channel credibility index is obtained, which reflects the overall quality and stability of the mall's customer flow source.

[0061] Based on the access device data in the initial evaluation matrix, the number and distribution ratio of covered device types are counted to obtain the device coverage. The number of device types reflects the diversity of contact points between the mall and customers, and the distribution ratio reflects the balance degree of various contact points. The device coverage is an indicator to measure the mall's omnichannel service ability. When calculating, both the number of supported device types and the balance degree of the use of each device are considered, and usually an information entropy or similar diversity index is used for quantification.

[0062] Based on the user geographical data in the initial evaluation matrix, the concentration degree and coverage breadth of the geographical distribution are calculated to obtain the geographical distribution index. The concentration degree reflects the aggregation degree of the customer source, usually calculated by the Gini coefficient or similar inequality measurement index; the coverage breadth reflects the geographical scope of the mall's influence, usually calculated by counting the number of administrative regions covered by effective customer sources and combining distance factors. The geographical distribution index comprehensively considers these two aspects, paying attention not only to the deep penetration of the mall in the core business district but also to its radiation ability in a wider area.

[0063] Linearly combine the credibility of the source channel, the device coverage, and the geographical distribution to obtain the site attribute coefficient. Linear combination means assigning weights to the three indicators and then calculating the weighted average. The weight configuration is usually determined according to the strategic focus and business objectives of the mall. For example, high-end malls that pay more attention to the quality of customer sources will assign a higher weight to the credibility of the source channel, while regional malls that pursue broad market coverage may place more emphasis on the geographical distribution indicator. The site attribute coefficient reflects the comprehensive performance of the mall in terms of the customer source structure.

[0064] Perform a weighted summation operation on the basic passenger flow score, the user stickiness coefficient, the user engagement coefficient, and the site attribute coefficient to obtain the site quality assessment database. Weighted summation is a commonly used aggregation method in multi-criteria decision-making, which can synthesize the performance of each dimension to form an overall evaluation. The weight configuration is usually determined based on expert experience and historical data analysis, and can also be derived through scientific methods such as the analytic hierarchy process. The site quality assessment database not only contains the final comprehensive score, but also retains the original data and intermediate calculation results of each dimension, facilitating subsequent analysis and application.

[0065] Taking the evaluation of a shopping center as an example, the daily average passenger flow of the shopping center is 15,000 person-times, and the number of independent visitors is 9,000 people. Divide these two data by the maximum values of 20,000 person-times and 12,000 people of similar shopping malls in this area respectively, and the normalized values obtained are 0.75 and 0.75 respectively. Combining them in a ratio of 6:4, the basic passenger flow score is 0.75×0.6 + 0.75×0.4 = 0.75. The average stay time of customers in this shopping center is 105 minutes, and the ratio to the industry average of 90 minutes is 1.17, that is, the user stickiness coefficient is 1.17. The bounce rate of the shopping center is 25%, the industry benchmark value is 35%, and the theoretical maximum value is 100%. Taking the adjustment coefficient λ as 0.8 and substituting it into the formula for calculation, the user participation coefficient is 1 + (0.35 - 0.25) / (1 - 0.35)×0.8 = 1.12. Among the source channels, brand activities account for 30%, the surrounding community accounts for 40%, online drainage accounts for 20%, and business cooperation accounts for 10%. The weight coefficients of each channel are 1.2, 1.0, 0.9, and 1.1 respectively. The weighted average gives the credibility of the source channel as 1.05. This shopping center supports three types of devices: mobile terminals, mall wayfinding devices, and membership cards. The usage ratio distribution is relatively balanced, and the calculated device coverage is 0.95. The geographical distribution of customer sources is concentrated in three main administrative regions, and the coverage is moderate. The calculated geographical distribution index is 0.85. Combining these three indicators in a linear combination according to the weight ratio of 3:2:5, the site attribute coefficient is 0.945. Finally, the indicators of the four dimensions are weighted and summed according to their respective importance. Assuming the weights are 0.35, 0.25, 0.25, and 0.15 respectively, the comprehensive score is 0.75×0.35 + 1.17×0.25 + 1.12×0.25 + 0.945×0.15 = 0.99625. This score and related detailed data are stored in the site quality evaluation database.

[0066] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0067] (1) Extract the device type data of the users of the mall traffic site from the quality evaluation database, and statistically calculate the distribution ratios of mobile devices, wayfinding terminals, and membership cards to obtain the device type distribution;

[0068] (2) Extract the user access time data from the quality evaluation database, divide the time interval according to 24 hours, and statistically calculate the access frequency of each time interval to obtain the access time period distribution;

[0069] (3) Extract the user location area data from the quality evaluation database, calculate the proportion of the number of access users in provincial and municipal regions to obtain the geographical distribution;

[0070] (4) Extract user consumption data from the quality assessment database, perform weighted calculations on consumption categories, amount levels, and purchase frequencies to obtain consumption preferences.

[0071] (5) Combine the device type distribution, access time period distribution, geographical distribution, and consumption preferences in multiple dimensions, and establish weight coefficients to obtain the portrait of the venue users.

[0072] (6) Construct the characteristics of the target merchant customer group through demographic characteristics, consumption ability levels, and category preference dimensions to obtain the portrait of the target merchant users.

[0073] (7) Calculate the eigenvectors of the corresponding dimensions of the venue user portrait and the target merchant user portrait, calculate the dimension difference value through the Euclidean distance formula, and obtain the matching score.

[0074] Specifically, establishing user portraits and calculating the matching degree are key steps in the shopping mall traffic venue screening method. Extracting the device type data of the shopping mall traffic venue users from the quality assessment database refers to classifying and counting the terminal devices used by users during the access data analysis process. The device type data includes three categories: mobile devices (smartphones, tablets, etc.), wayfinding terminals (self-service navigation devices, interactive screens, etc. in the shopping mall), and membership cards (physical cards or electronic membership cards). To statistically calculate the distribution ratios of these devices, it is necessary to first count the number of usage times of each type of device, and then divide by the total number of access times to obtain the percentage value. For example, the data captured through Wi-Fi probes and Bluetooth signals shows that among 10,000 visits to a certain shopping mall, mobile devices accounted for 8,500 times, wayfinding terminals accounted for 1,000 times, and membership cards accounted for 500 times, and the calculated distribution ratios are 85%, 10%, and 5% respectively. These values constitute the device type distribution, reflecting the main ways of interaction between users and the shopping mall and their technological preferences. Extracting the user access time data from the quality assessment database is to analyze the time pattern of customers' visits to the shopping mall. Divide the time interval into 24 hours, usually in 1-hour units, that is, 0 - 1 o'clock, 1 - 2 o'clock... 23 - 24 o'clock, a total of 24 intervals. Counting the access frequency for each time interval means calculating how many customers enter the shopping mall during each time period. This data can be obtained through passenger flow counting devices, camera recognition systems, or Wi-Fi probes at the entrance. Divide the number of access times in each time period by the total number of access times to obtain the access ratio for each time period, thus forming the access time period distribution. This distribution can intuitively display the peak and off-peak periods of the shopping mall's passenger flow, providing a reference for merchant location selection and business hours arrangement.

[0075] Extracting data on the user's location from the quality assessment database refers to analyzing the geographical distribution of customer sources. Data on the user's location can be obtained through various means, such as IP address resolution, mobile base station positioning, membership registration information, etc. Calculating the proportion of the number of visiting users in provincial and municipal regions means counting the number of customers from each province and city, dividing by the total number of customers, and obtaining a percentage value. This calculation focuses on both the in-depth penetration within the core business district and the coverage ability of the wide-area market. Geographical distribution data can reflect the geographical radiation range and influence of the mall, and is crucial for evaluating the accuracy and attractiveness of the mall's positioning.

[0076] Extracting the user's consumption data from the quality assessment database, and performing weighted calculations on the consumption categories, amount levels, and purchase frequencies to obtain consumption preferences. This process involves complex multi-dimensional data analysis and can be expressed by the following formula:

[0077]

[0078] Among them, CP represents Consumption Preference, T represents the total number of time periods, and ω t represents the weight coefficient for time period t. C represents the total number of consumption categories, and ν c represents the importance weight for category c. CC t,c represents the number of consumption times of category c within time period t, and TC t represents the total number of consumption times within time period t. V represents the total number of amount levels, and ρ v represents the weight coefficient for amount level v. CV t,v represents the number of consumption times of amount level v within time period t. F represents the number of purchase frequency groups, and σ f represents the weight coefficient for frequency group f. CF t,f represents the number of consumers in frequency group f within time period t. This formula comprehensively considers the three dimensions of consumption categories, amount levels, and purchase frequencies, and obtains the consumption preference characteristics of users through weighted calculation. For example, for a certain shopping center, analyze the consumption proportions of customers in different categories such as clothing, catering, and entertainment, as well as the distribution of high, medium, and low consumption amount levels, and then combine the characteristics of different frequency groups such as visiting once a month, 2 - 3 times, and more than 4 times, to form the consumption preference portrait of this mall through weighted calculation.

[0079] The device type distribution, access time distribution, geographical distribution, and consumption preferences are combined in multiple dimensions, and a weight coefficient is established to obtain the site user profile. The multi-dimensional combination means integrating the feature data in the above four aspects into a unified data structure to form a complete user behavior profile. The establishment of the weight coefficient is usually based on the importance and discriminability of the features, and methods such as expert experience method, historical data analysis, or machine learning algorithms can be used to determine it. The site user profile is a digital description of the characteristics of the customer group in this shopping mall, including information in multiple dimensions such as device usage habits, access time patterns, geographical origin distribution, and consumption behavior preferences.

[0080] The target merchant customer group characteristics are constructed through demographic characteristics, consumption ability levels, and category preference dimensions to obtain the target merchant user profile. Demographic characteristics include basic attributes such as age, gender, education level, and occupation; the consumption ability level reflects the economic strength and price sensitivity of the target customer group; the category preference describes the interest tendency of the target customer group in different product or service categories. The target merchant user profile is a digital description of the ideal customer group of the merchant, usually provided by the merchant according to its own business characteristics and market positioning, or can be generated by analyzing the customer data of the merchant in other stores.

[0081] Feature vector calculations are performed on the corresponding dimensions of the site user profile and the target merchant user profile, and the dimension difference value is calculated through the Euclidean distance formula to obtain the matching score. Feature vector calculation means converting the feature of each dimension of the two profiles into a comparable numerical vector. The Euclidean distance is the straight-line distance between two points in an n-dimensional space, and the calculation formula is the square root of the sum of the squares of the differences between the two points in each dimension. In the calculation of the user profile matching degree, the smaller the Euclidean distance, the closer the two profiles are, and the higher the matching degree. Usually, the calculated distance value is converted into a matching score between 0 and 1 through normalization or function transformation, and the value closer to 1 indicates a higher matching degree.

[0082] For example, a well-known coffee brand plans to open a new store in a shopping mall and needs to evaluate whether the customer group in the mall meets the characteristics of its target customers. First, by analyzing the device type data in the mall quality assessment database, it is found that the proportion of customers using mobile devices in the mall is 88%, the proportion of wayfinding terminals is 7%, and the proportion of membership cards is 5%, indicating that the customer group has a high level of mobile Internet activity. Then, by analyzing the distribution of access time periods, it is identified that 12:00 - 14:00 and 17:30 - 20:00 on weekdays, as well as 14:00 - 18:00 on weekends, are the peak passenger flow periods, which conform to the work and rest patterns of office workers and family consumers. Next, by analyzing the geographical distribution, it is determined that the main customer sources come from high-end residential areas and office building clusters within a 5-kilometer radius of the surrounding area. Again, by analyzing the consumption data and applying the weighted calculation formula, it is determined that the proportion of customers' consumption in the food and beverage category in this mall is relatively high (35%), and the middle and high-end food and beverage consumption amount level (per capita 100 - 300 yuan) dominates (60%), and the purchase frequency is mainly 2 - 3 times per week (45%). Combining these characteristic data in multiple dimensions forms the user portrait of this mall. At the same time, the coffee brand provides its own target customer portrait: mainly white-collar workers aged 25 - 45, with medium to high consumption ability, preferring a quality life, and the main consumption times are during the lunch break and after work on weekdays, as well as during the weekend leisure time. By calculating the characteristic vectors and Euclidean distances of the two portraits, the matching score is obtained as 0.82 (with a full score of 1), indicating that the customer group characteristics of this mall highly match the target customers of the coffee brand and are suitable for opening a new store.

[0083] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0084] (1) From the historical operation data of the mall traffic site, count the number of passenger flows according to the daily time granularity to obtain the passenger flow data;

[0085] (2) Statistically analyze the consumption behaviors of users in the passenger flow data, and obtain the conversion rate data through the ratio operation of the number of consuming users to the passenger flow;

[0086] (3) Statistically analyze the consumption amounts of users in the conversion rate data, and obtain the per customer transaction data through the ratio operation of the consumption amount to the number of consuming users;

[0087] (4) Group the passenger flow, conversion rate, and per customer transaction at different time periods in the mall traffic site to obtain the business time period data;

[0088] (5) Extract the promotion activity types, discount intensities, and activity scopes from the mall traffic site to obtain the promotion activity form data;

[0089] (6) Combine the matching score with the passenger flow data, conversion rate data, average customer spending data, business hour data, and promotion activity form data for feature combination to obtain the business feature data;

[0090] (7) Normalize the business feature data and perform weighted calculation in combination with the site quality score to obtain the predicted score of the site business effect.

[0091] Specifically, in the shopping mall traffic site screening method, the prediction of business effect is a key link, which directly affects the accuracy of the final screening result. Counting the number of passenger flows according to the daily time granularity from the historical business data of the shopping mall traffic site means extracting the passenger flow records of different time periods (usually in hours) every day from the original data collected by passenger flow counting devices, camera recognition systems, Wi-Fi probes, etc. at each entrance, and conducting summary statistics. The number of passenger flows refers to all the person-times entering the shopping mall within a specific time period, including the situation of repeated entry. These passenger flow data are organized according to the date and time period, forming a two-dimensional table form, with the horizontal axis being the date and the vertical axis being the time period (such as 10:00 - 11:00, 11:00 - 12:00, etc.), and the values in the table being the passenger flow person-times of the corresponding time unit. Statistically analyze the consumption behavior of users in the passenger flow data, and obtain the conversion rate data through the ratio operation of the number of consuming users to the passenger flow. The statistical analysis of consumption behavior refers to extracting the number of users who actually made purchases from the shopping mall sales system, cashier system, or membership management system. The number of consuming users refers to the number of customers who completed purchase behaviors in any store within the shopping mall within a specific time period, and is obtained through the integration of the sales data of each store in the shopping mall. The conversion rate calculation formula is the number of consuming users divided by the passenger flow in the same time period, indicating what proportion of the customers entering the shopping mall actually made purchases, and is an important indicator to measure the marketing effect and attractiveness of the shopping mall. The conversion rate data is also organized according to the date and time period, having the same structural form as the passenger flow data.

[0092] Statistics are carried out according to the consumption amount of users in the conversion rate data. By calculating the ratio of the consumption amount to the number of consuming users, the per customer transaction amount data is obtained. The consumption amount statistics refer to the total amount of all sales transactions extracted from the mall sales system within a specific time period. The formula for the per customer transaction amount is the total consumption amount divided by the number of consuming users, which represents the average amount spent by each actual consuming customer and reflects the consumption ability of the mall's customer group and the positioning of commodity value. The per customer transaction amount data is also organized by date and time period, forming a data table with the same structure as the passenger flow and conversion rate. Data grouping is performed on the passenger flow, conversion rate, and per customer transaction amount in different time periods in the mall traffic area to obtain business hour data. Data grouping means classifying and integrating time periods with similar characteristics according to the time distribution rules of these three types of indicators. Common grouping methods include clustering analysis (such as the K-means algorithm) or experience-based division (such as the morning market from 9:00 to 12:00, the lunch market from 12:00 to 14:00, the afternoon tea from 14:00 to 17:00, the dinner market from 17:00 to 21:00, etc.). The business hour data not only includes the time range of each group, but also includes the average passenger flow, average conversion rate, and average per customer transaction amount of each group, as well as statistical characteristics such as the coefficient of variation of these indicators, comprehensively describing the differences in the business performance of the mall at different times.

[0093] The promotion activity type, discount intensity, and activity scope are extracted from the mall traffic area to obtain promotion activity form data. The promotion activity type includes different categories such as holiday promotions, member-exclusive activities, brand joint activities, seasonal discounts, etc.; the discount intensity refers to the degree of price preference during the activity, usually expressed as a discount ratio or the amount of full reduction; the activity scope refers to the coverage of shops participating in the promotion and the coverage of commodity categories. The promotion activity form data is obtained by collecting the mall's historical marketing planning documents, the usage records of discount codes in the sales system, and the participation of merchants, and is transformed into a quantifiable feature vector through structured processing, such as the one-hot encoding of the activity type, the numerical representation of the discount intensity, and the coverage rate of the activity scope.

[0094] The matching score is combined with the passenger flow data, conversion rate data, per customer transaction amount data, business hour data, and promotion activity form data to obtain business feature data. The matching score refers to the degree of matching between the mall user portrait calculated in the previous step and the target merchant user portrait. Feature combination means integrating these data in different dimensions into a unified data structure to form a complete business feature description. The combination methods include direct splicing, weighted average, principal component analysis, etc. The purpose is to combine the user matching degree with the actual business performance to form a more comprehensive evaluation basis. The business feature data is a multi-dimensional vector or matrix, and each dimension represents an evaluation index, jointly describing the business potential and performance of the mall.

[0095] Normalize the business feature data, perform weighted calculation in combination with the site quality score, and obtain the predicted score of the site business effect. Normalization refers to converting indicators with different dimensions and ranges to the same scale (usually the 0-1 interval), and common methods include maximum-minimum normalization, Z-score normalization, etc. The site quality score is a comprehensive score calculated based on the basic passenger flow score, user stickiness coefficient, user participation coefficient, and site attribute coefficient in the previous steps. Weighted calculation means assigning different weights to each indicator and then calculating the weighted average, and the weight configuration is usually determined based on industry experience or historical data analysis. The predicted score of the site business effect is a value between 0 and 1, and the higher the value, the greater the business potential of the mall.

[0096] For example, a shopping mall plans to introduce a children's education brand and needs to evaluate whether the business environment of the mall is suitable for this brand. First, extract the daily passenger flow data from the mall's historical business data and find that the average daily passenger flow from Monday to Friday is 15,000 person-times, and the average daily passenger flow on weekends is 25,000 person-times. The data also shows that 10:00-12:00 and 15:00-18:00 are two peak passenger flow periods. Then, extract the consumption data from the sales system and calculate that the average conversion rate from Monday to Friday is 35%, and the average conversion rate on weekends is 45%, indicating that about one-third to half of the visitors will consume in the mall. Then, calculate the average customer unit price based on the consumption amount and the number of consuming users and find that the average customer unit price from Monday to Friday is 220 yuan, and the average customer unit price on weekends is 320 yuan, reflecting the increase in the proportion of family consumption on weekends. Group and analyze the passenger flow, conversion rate, and customer unit price by time period and find that 10:00-12:00 and 15:00-18:00 on weekends are the golden periods for family consumption, and the sales of children-related products during this period are significantly higher than other periods. Extract data from the promotion activity records and identify that the conversion rates of promotions related to children, such as Children's Day and the start of school season, have increased significantly, and the parent-child activities held by the mall every month can bring a stable flow of target customers. Combine the user portrait matching score (0.85, indicating a high match between the mall's customer group and the target users of the education brand) calculated in the previous steps with the above business data to form a complete business feature description. Finally, normalize these feature data and combine them with the site quality score of the mall (0.92, based on the calculations in the previous steps), and obtain a predicted score of the site business effect of 0.88 through weighted calculation, indicating that the mall has high business potential for the children's education brand and is suitable as a new store location.

[0097] In a specific embodiment, the process of executing step S105 may specifically include the following steps:

[0098] (1) Calculate the comparison between the predicted score of business operation effect and the industry average level, and conduct quality access evaluation by setting the minimum access score to obtain a list of access sites;

[0099] (2) Calculate the user overlap based on the matching scores of each mall traffic site in the access site list to obtain the user matching evaluation result;

[0100] (3) Analyze the historical business performance of each mall traffic site in the access site list, set three levels of effect criteria: high, medium, and low, to obtain the expected effect classification data;

[0101] (4) Count the daily average passenger flow of the mall traffic sites in the access site list, and set the upper and lower limits of the site selection magnitude according to the merchant's investment budget to obtain the merchant site selection magnitude threshold;

[0102] (5) Cross-validate the expected effect classification data with the merchant site selection magnitude threshold, and screen out the eligible mall traffic site combinations to obtain the preliminary selection site list;

[0103] (6) Analyze the overlap of the covered populations of each mall traffic site in the preliminary selection site list, and obtain the site screening combination result through deduplication.

[0104] Specifically, calculating the comparison between the predicted score of business operation effect and the industry average level is the first step of the screening process. The predicted score of business operation effect is a value between 0 and 1 obtained through multi-dimensional feature combination and weighted calculation in the previous steps, representing the business potential of the mall. The industry average level is a benchmark value calculated from industry association data, market research reports, or historical accumulated data. The comparison calculation uses a simple difference method or ratio method to calculate the difference between the current mall and the industry benchmark. The minimum access score is a set threshold value, usually 10%-20% higher than the industry average level, to ensure that the selected malls have a competitive advantage. Quality access evaluation is to screen out the malls with a predicted score of business operation effect higher than the minimum access score to form a list of access sites. This list contains basic information such as the mall name, location, area, etc., as well as evaluation data such as the predicted score of business operation effect and the industry comparison result.

[0105] Calculate the user overlap based on the matching scores of each mall traffic site in the access site list to obtain the user matching evaluation result. The matching score is the similarity between the mall user portrait and the target merchant user portrait calculated in the previous steps. The calculation of user overlap involves more complex data analysis and can be expressed by the following formula:

[0106]

[0107] Among them, UOI represents the User Overlap Index, κ represents the direct matching degree weight coefficient (taking values between 0 and 1), MS represents the Match Score, S represents the total number of merchant categories in the mall, η s represents the correlation coefficient of category s, TS s represents the transaction share of category s, TS total represents the total transaction share. This formula comprehensively considers the direct user profile matching degree and the indirect associated merchant transaction data, reflecting a more comprehensive user matching relationship. For example, when evaluating the location selection of a catering brand, not only the direct matching degree between the mall customers and the target customer group is considered, but also the transaction situations of other associated business forms such as other catering brands and entertainment venues in the mall are considered, because the customers of these business forms often have a high customer group overlap with the target catering brand. The user matching degree evaluation result is a comprehensive indicator, reflecting the overall fit degree between the mall customer group and the merchant's target customers. By analyzing the historical business performance of each mall traffic site in the list of eligible sites, three levels of effect standards, high, medium, and low, are set to obtain the expected effect classification data. The historical business performance includes multiple indicators such as the passenger flow growth rate, the shop rent growth rate, the sales growth rate, and the brand replacement rate. These data are extracted from the mall management system, lease contract archives, and sales reports, usually collecting data for the most recent 12 - 36 months. The analysis methods include trend analysis, seasonal analysis, and volatility analysis, etc. The setting of the three levels of effect standards, high, medium, and low, usually adopts the quantile method or the natural breakpoint method to determine the demarcation points according to the data distribution characteristics. The expected effect classification data contains the scores of each indicator of each mall and its comprehensive classification result, providing a stratification basis for subsequent screening.

[0108] Statistics are made on the daily average passenger flow of the mall traffic sites in the list of eligible sites. According to the merchant's investment budget, the upper and lower limits of the location selection level are set to obtain the merchant's location selection level threshold. The daily average passenger flow refers to the average number of passenger flow per day within a complete business cycle (such as one year) of the mall, which is an important indicator of the mall scale. The calculation method is to divide the total number of passenger flow during a specific period by the number of business days. The merchant's investment budget includes the sum of various costs such as rent budget, decoration budget, equipment budget, and initial operating funds. The determination of the upper and lower limits of the location selection level is based on the investment return model, and the considerations include the industry average single - customer consumption amount, conversion rate, gross profit margin, cost structure, and target payback period, etc. The merchant's location selection level threshold is a passenger flow interval, ensuring that the malls within this interval can not only meet the merchant's sales target but also will not affect the investment return due to excessive costs.

[0109] Cross-validate the expected effect grading data with the merchant site selection magnitude threshold to screen out eligible mall traffic site combinations and obtain a preliminary site list. Cross-validation is a two-dimensional screening method where the horizontal axis is the expected effect grading (high, medium, low), and the vertical axis is whether the passenger flow is within the threshold range (yes / no). Only malls that simultaneously meet the conditions of "high or medium expected effect" and "passenger flow within the threshold range" are retained. This cross-validation ensures that the screening results have both good business potential and are in line with the merchant's investment scale. The preliminary site list contains mall information that meets the dual conditions, and the number is significantly less than the access site list, being closer to the final recommended result.

[0110] Conduct an overlap analysis of the covered populations of each mall traffic site in the preliminary site list, and obtain the site screening combination result through deduplication. The covered population refers to the target customer group of the mall, usually determined based on the mall's geographical location, traffic accessibility, and radiation range. The overlap analysis uses Geographic Information System (GIS) technology to calculate the intersection area of the radiation ranges of different malls and the population quantity within the intersection area. The deduplication process uses the solution method of the Set Covering Problem, with the goal of selecting the minimum number of malls to cover the largest range of target populations. Common algorithms include the greedy algorithm, genetic algorithm, or integer programming, etc. The site screening combination result is the final site selection plan recommended to the merchant, containing all relevant information of the recommended malls and an analysis of the combination advantages.

[0111] For example, a chain clothing brand plans to open a new store in a region. First, it collects data on 30 candidate shopping malls. Through the analysis of the predicted operation effect scores, it is found that the predicted scores of these shopping malls are distributed between 0.65 and 0.92, and the industry average level is 0.75. The minimum access score is set at 0.80. After screening, 18 shopping malls are retained and entered into the access site list. Then, the user overlap index of these 18 shopping malls is calculated. Among them, the weight coefficient κ of the direct matching degree is taken as 0.6, and the correlation coefficients η of each merchant category are set as follows: 0.8 for the same clothing brand, 0.6 for the accessory brand, 0.4 for the catering brand, and 0.2 for other brands. By analyzing the transaction shares of different categories of merchants in each shopping mall, it is calculated that the user overlap index of the 18 shopping malls is distributed between 0.55 and 0.88. According to the merchant requirements, the user overlap index is set to be greater than 0.70. After screening again, 12 shopping malls are retained. Then, the historical operation data of these 12 shopping malls in the past 24 months are analyzed, including indicators such as the passenger flow growth rate (average 17.5%, standard deviation 6.8%), rent growth rate (average 8.2%, standard deviation 2.5%), and sales growth rate (average 15.3%, standard deviation 5.4%). The comprehensive scoring method is used to calculate the overall performance, and according to the natural breakpoint method, the shopping malls are divided into three categories: 4 high-effect (score > 85), 6 medium-effect (score 70 - 85), and 2 low-effect (score < 70). At the same time, according to the single-store investment budget of 4.5 million yuan and the target payback period of 18 months of the clothing brand, combined with parameters such as the average single-customer transaction amount of 350 yuan, conversion rate of 15%, and gross profit margin of 60% in the industry, it is calculated that the appropriate daily passenger flow of the shopping mall should be between 6,000 and 12,000 person-times. The expected effect grading and the passenger flow threshold are cross-validated. Finally, 3 high-effect + within-threshold and 4 medium-effect + within-threshold, a total of 7 shopping malls, are retained and entered into the preliminary selection site list. Finally, by analyzing the geographical locations and radiation ranges of these 7 shopping malls, it is found that there is obvious overlap in the customer groups of some shopping malls. The overlap matrix between each pair is calculated through the GIS system, and the greedy algorithm is used for combinatorial optimization. Finally, 3 shopping malls with the widest coverage and the smallest overlap with each other are selected as the final recommendation, forming the site screening combination result.

[0112] In a specific embodiment, the process of executing step S106 may specifically include the following steps:

[0113] (1) Real-time monitor the quality evaluation indicators of the traffic sites of each shopping mall in the site screening combination result, and obtain the quality fluctuation data through deviation calculation with the historical benchmark value;

[0114] (2) Regularly update and compare the user group characteristics according to the user portrait matching degree of the traffic sites of the shopping malls in the site screening combination result, and obtain the user change data;

[0115] (3) Collect the operation data of the shopping mall traffic sites in the site screening combination results, and obtain the actual operation effect data by calculating the difference between the actual conversion rate and the expected conversion rate;

[0116] (4) Comprehensively analyze the quality fluctuation data, user change data and actual operation effect data, and make a warning judgment according to the set threshold to obtain the site warning information;

[0117] (5) According to the site warning information, classify the shopping mall traffic sites with different warning levels, and obtain the effect comparison result through comparative analysis with historical performance;

[0118] (6) Dynamically score the effect comparison results according to the weight index, re-rank and replace the shopping mall traffic sites to obtain the shopping mall site optimization plan.

[0119] Specifically, after the initial screening of the shopping mall traffic site screening method, a dynamic optimization mechanism needs to be established to ensure the continuous effectiveness of the recommendation results. Real-time monitoring of the quality evaluation indicators of each shopping mall traffic site in the site screening combination results means continuously collecting the key indicator data of the screened shopping malls and conducting dynamic analysis. The quality evaluation indicators include core indicators calculated in the previous steps such as the basic passenger flow score, user stickiness coefficient, user participation coefficient, and site attribute coefficient. The frequency of real-time monitoring is usually daily or weekly to ensure that changes in indicators can be captured in a timely manner. Calculating the deviation from the historical benchmark value means comparing the current monitored value with the average performance of the shopping mall over a past period (usually 3 - 6 months) or the performance during a specific period (such as the same period last year). The deviation calculation methods include the simple difference method (current value minus the benchmark value), the percentage difference method (difference divided by the benchmark value), or the standard deviation multiple method (difference divided by the historical standard deviation). The quality fluctuation data is a structured data set that contains information such as the current value, benchmark value, deviation value, and change trend of each indicator, reflecting the dynamic changes in the quality status of the shopping mall. According to the user portrait matching degree of the shopping mall traffic sites in the site screening combination results, regularly update and compare the user group characteristics to obtain the user change data. The user portrait matching degree refers to the similarity between the shopping mall user portrait and the target merchant user portrait. The user group characteristics include multiple dimensions such as device type distribution, access time period distribution, geographical distribution, and consumption preferences. Regularly updating and comparing means re-collecting user data, updating the user portrait monthly or quarterly, and conducting comparative analysis with the previous portrait. The comparison methods include feature vector difference calculation, distribution similarity measurement (such as KL divergence, JS divergence), or time series comparison of specific indicators. The user change data records the changes in each characteristic dimension, including the change amplitude, change direction, and change rate, reflecting the evolution trend of the shopping mall customer group structure.

[0120] Collect the operation data of the mall traffic sites in the combined results of site screening. By calculating the difference between the actual conversion rate and the expected conversion rate, obtain the actual operation effect data. The operation data includes business indicators such as passenger flow, transaction volume, and sales amount, which are collected through the mall sales system, cash register system, or third-party monitoring tools. The conversion rate refers to the proportion of the actual number of customers who make purchases to the total passenger flow, and is the core indicator to measure the operation effect of the mall. The expected conversion rate is the target value predicted based on historical data and market environment, usually determined during the screening stage. The difference calculation uses simple subtraction or percentage deviation method to quantify the gap between the actual performance and the expectation. The actual operation effect data contains information such as the actual value, expected value, difference value, and achievement rate of each operation indicator, comprehensively reflecting the operation performance of the mall.

[0121] Conduct a comprehensive analysis of the quality fluctuation data, user change data, and actual operation effect data, and make a warning judgment according to the set threshold to obtain the site warning information. Comprehensive analysis means integrating the three types of data to find the correlation and influence relationship. The analysis methods include correlation analysis, factor analysis, or regression analysis, etc. Setting the threshold means setting the warning trigger point for each indicator, usually determined based on industry experience or historical data statistics. Common threshold setting methods include the fixed value method (such as a conversion rate decrease exceeding 10% triggers a warning), the percentile method (such as an indicator dropping below the 25th percentile of history triggers a warning), or the multiple condition method (such as triggering a warning only when multiple conditions are met). The warning judgment determines the warning status and warning level according to whether the indicator exceeds the threshold. The site warning information contains content such as the warning site, warning indicator, warning level, trigger time, and specific performance, providing a decision-making basis for subsequent adjustments.

[0122] According to the site warning information, classify the mall traffic sites with different warning levels, and through comparative analysis with historical performance, obtain the effect comparison result. Classification processing adopts different analysis strategies according to the warning level and type. Common classifications include red warning (seriously deviating from the expectation, immediate adjustment required), yellow warning (significantly deviating from the expectation, close attention required), and blue warning (slightly deviating from the expectation, optimization recommended). The historical performance comparative analysis means comparing the current performance of the warning site with its past performance in multiple time periods, and making a horizontal comparison in combination with the industry average level and the performance of leading sites. The effect comparison result is a comprehensive analysis report, containing detailed data comparison, problem cause analysis, and improvement potential evaluation, etc.

[0123] The effect comparison results are dynamically scored according to the weighted indicators, and the mall traffic venues are re-ranked and replaced to obtain the mall venue optimization plan. The weighted indicators are the scoring basis determined according to the strategic focus of the merchants and the market environment, and usually include multiple dimensions such as quality stability, customer group matching, business performance, and development potential. Dynamic scoring refers to scoring each mall on each weighted indicator based on the latest data, and then calculating the weighted total score. Re-ranking refers to adjusting the priority ranking of the original malls based on the dynamic scoring results. Replacement processing refers to considering replacing the warning venue with an alternative venue when the warning venue score is significantly lower than the alternative venue. The mall venue optimization plan is the final adjustment suggestion, which includes complete content such as retained venues, replacement venues, adjustment strategies and expected effects.

[0124] For example, a mobile phone brand selected three shopping malls in a region as locations for new stores, and began to implement dynamic monitoring and optimization after six months of operation. First, the quality evaluation indicators of the three shopping malls were monitored in real time, and the original data such as passenger flow data, stay time, and bounce rate were collected weekly to calculate quality indicators such as basic passenger flow score and user stickiness coefficient. These indicators were compared with the benchmark values ​​of the first three months of the store opening, and it was found that the user stickiness coefficient of shopping mall A dropped from 1.15 to 0.92, a decrease of 20%, exceeding the warning threshold of 15%; while the changes in various indicators of shopping malls B and C were within a reasonable range. Then, the user portraits of each shopping mall were updated monthly, and the changes in the current portraits and the initial portraits were compared by the feature vector difference calculation method. The user portrait of shopping mall A changed significantly, especially the distribution of visit time periods changed from mainly afternoon to mainly evening, and the consumption preference changed from mainly mid-to-high-end electronic products to mainly fast-moving consumer goods; the customer group structure of shopping mall B was basically stable; the geographical distribution of users in shopping mall C expanded, and the radiation range increased by about 20%. At the same time, sales data of the three shopping malls were collected to calculate the difference between the actual conversion rate and the expected rate. The actual conversion rate of Mall A is 8%, far below the expected 15%; Mall B is 14%, close to the expected 15%; Mall C is 18%, exceeding the expected 15%. After comprehensive analysis of the three types of data and comparison with the preset warning threshold, it is determined that Mall A has triggered a red warning (multiple core indicators are seriously deviated from expectations), Mall B is in normal condition, and Mall C performs better than expected. Based on the warning information, an in-depth analysis of Mall A is conducted, and compared with its historical performance, it is found that the problem began with the adjustment of the mall's business format 4 months ago. A large number of catering brands have moved in and changed the customer flow structure. At the same time, the investigation of Mall D among the alternative malls found that its recent performance is excellent, and its user portrait is highly matched with the target customer group of mobile phone brands. Finally, according to the weighted indicators of "customer group matching 40%, business performance 30%, development potential 20%, and location complementarity 10%", the dynamic scoring is carried out. Mall A scores 62, which is significantly lower than Mall D's 86. Therefore, the optimization plan recommends replacing Mall A with Mall D, and retaining the well-performing Malls B and C to form a new venue combination.

[0125] In a specific embodiment, the process of performing the step of comprehensively analyzing quality fluctuation data, user change data, and actual business effect data may specifically include the following steps:

[0126] (1) Group and statistically analyze the quality fluctuation data according to the daily fluctuation range. By calculating the number of consecutive fluctuation days and the fluctuation direction, obtain the quality fluctuation trend data;

[0127] (2) Decompose the user change data according to the change dimensions. By calculating the change range and change rate of each dimension, obtain the user change trend data;

[0128] (3) Conduct time series analysis on the actual business effect data. By calculating the cumulative value and change slope of the effect deviation, obtain the effect change trend data;

[0129] (4) Compare the quality fluctuation trend data, user change trend data, and effect change trend data in the same dimension. Through correlation analysis, obtain the trend correlation data;

[0130] (5) Quantitatively calculate the abnormal degree of each index in the trend correlation data. By comparing with the preset warning threshold, obtain the warning level data;

[0131] (6) Associate and mark the warning level data with the abnormal indexes. Through summary analysis, obtain the site warning information.

[0132] Specifically, the warning mechanism in the shopping mall traffic site screening method requires a refined data processing process to ensure the accuracy and timeliness of monitoring. Grouping and statistically analyzing the quality fluctuation data according to the daily fluctuation range is the basic step for warning judgment. The daily fluctuation range refers to the difference between each quality index on the current day and the reference value, usually expressed in percentage form. The grouping statistics adopt the numerical interval division method, dividing the fluctuation range into multiple levels, such as slight fluctuation (0 - 5%), moderate fluctuation (5 - 10%), significant fluctuation (10 - 20%), and severe fluctuation (>20%). Calculating the number of consecutive fluctuation days refers to counting the number of days when a specific index continuously changes in the same direction (rising or falling). The fluctuation direction records the trend of the index change. For example, if the passenger flow continuously decreases for 7 days and the fluctuation range exceeds 10% each day, it is recorded as a "passenger flow - significant - 7 days" fluctuation record. The quality fluctuation trend data is a structured information set, including the fluctuation level distribution, duration, and direction information of each index, which can intuitively reflect the dynamic change trend of the shopping mall quality status.

[0133] Decomposing user change data according to change dimensions is a key step in deeply understanding the changes in the customer group structure. Change dimensions include aspects such as device type distribution, access time period distribution, geographical distribution, and consumption preferences. The decomposition process is to split the changes in the overall user portrait into specific changes in each dimension. By calculating the change amplitude in each dimension, that is, the difference between the current value and the benchmark value, mathematical tools such as Euclidean distance or Mahalanobis distance are usually used for quantification. The change rate is the change amplitude divided by the time interval, reflecting the speed of change. For example, the consumption preference of a certain shopping mall changed from mainly high-end electronic products to mainly mid-range clothing within 3 months, with a change amplitude of 40%, then the change rate is 13.3% / month, and this rate is significantly higher than the normal market seasonal change (usually about 5% / month). User change trend data records the specific situations of changes in each dimension, including information such as the main change direction, key change indicators, and change rate rankings, providing a panoramic view of customer group changes for subsequent analysis.

[0134] Performing time series analysis on the actual business performance data is a scientific method for evaluating the trend of the shopping mall's performance. Time series analysis uses statistical tools such as the moving average method, exponential smoothing method, or ARIMA model to identify the trend component, seasonal component, and random component in the business data. By calculating the cumulative value of the effect deviation, that is, the sum of the deviation values at each time point, the performance gap accumulated in the long term can be evaluated. For example, the conversion rate of a certain shopping mall has been lower than expected for 3 consecutive months, and the monthly deviations are -2%, -3%, and -4% respectively, with a cumulative deviation of -9%. This continuous decline and accelerating deterioration trend is more worthy of attention than a single large fluctuation. The change slope is the slope of the trend line after linear regression fitting, intuitively showing the speed and direction of the indicator's change over time. Effect change trend data contains the trend parameters, periodic characteristics, and outlier information of each business indicator, providing a complete insight into the shopping mall's performance evaluation in the time dimension.

[0135] Comparing the quality fluctuation trend data, user change trend data, and effect change trend data in the same dimension is an important means of discovering the root causes of problems. Comparing in the same dimension means comparing the three types of trend data side by side on the same time scale to find the sequence and mutual influence of changes. Through correlation analysis methods, such as statistical tools like Pearson correlation coefficient, Spearman rank correlation, or cross-lagged correlation, the correlation strength between different indicators is calculated. For example, it is found through analysis that the change in the user portrait (decrease in consumption ability) of a certain shopping mall leads the change in the quality indicator (decrease in user stickiness) by 1 - 2 weeks, and the change in the quality indicator leads the change in the business performance (decrease in conversion rate) by 1 - 2 weeks. This time sequence relationship reveals a potential causal chain. Trend correlation data records the correlation coefficients, leading and lagging relationships, and possible causal paths between each indicator, providing a systematic analysis basis for early warning judgment.

[0136] Quantifying the degree of abnormality of each indicator in trend - related data is the core step of early - warning judgment. Multiple statistical methods are used for quantifying the degree of abnormality, such as the Z - score method (the difference between the current value and the mean divided by the standard deviation), anomaly detection algorithms (such as the IQR method, DBSCAN clustering), or rule - based scoring systems. By comparing with a preset early - warning threshold, it is determined whether an indicator triggers an early - warning. The preset early - warning threshold is determined based on industry experience, historical data analysis, or statistical significance principles. Common settings include conditions such as a Z - score exceeding 2 (about 95% confidence level), a deviation exceeding 30%, or a decline for 3 consecutive weeks. Early - warning level data is usually divided into multiple levels, such as level - 1 early - warning (red, immediate handling required), level - 2 early - warning (yellow, close attention required), and level - 3 early - warning (blue, optimization recommended), and each level corresponds to different degrees of abnormality and urgency.

[0137] Associating and marking the early - warning level data with abnormal indicators is the last step in forming complete early - warning information. The association marking is to bind the early - warning level with information such as specific abnormal indicators, abnormal reasons, and scope of influence to form a structured early - warning record. Through summary analysis, all early - warning information is integrated and prioritized to obtain the final site early - warning information. Summary analysis methods include decision - tree analysis, expert - system evaluation, or comprehensive scoring mechanisms. For example, when a shopping mall triggers both user - portrait change early - warning and business - effect early - warning at the same time and the two are highly correlated, the early - warning level will be upgraded and marked as "business decline caused by customer - group structure change". This kind of association analysis provides more valuable decision - making information than individual indicator early - warnings. The site early - warning information is a comprehensive early - warning report, including the early - warning site, early - warning level, key abnormal indicators, possible reasons, scope of influence, recommended measures, and follow - up points, etc., providing specific operation guidance for the dynamic optimization and adjustment of the shopping mall.

[0138] For example, after a shopping mall is selected as the store location for a high-end cosmetics brand, it enters the dynamic monitoring stage. First, quality indicators such as daily passenger flow and dwell time are collected, the deviation from the benchmark value is calculated, and grouped according to the deviation size. After continuous monitoring for 3 months, it is found that the dwell time indicator has been continuously decreasing since the 8th week, dropping from the benchmark value of 125 minutes to 110 minutes, and remaining within the deviation range of about -10% for 12 consecutive days, forming a quality fluctuation trend record of "dwell time - moderate decrease - 12 days". At the same time, the user portrait data is updated monthly, and the changes in each dimension are decomposed. It is found that the change in the age distribution dimension is the most significant. The proportion of people under 35 years old has dropped from the original 60% to 45%, with a change amplitude of 25% in 3 months and an average monthly change rate of about 8.3%, far higher than the normal fluctuation level of 3%, which is recorded as a high-speed change; the consumption preference has also shifted from mainly cosmetics and clothing to mainly catering and parent-child, with a change amplitude of about 30%. The analysis of business data shows that the conversion rate of high-end products in this shopping mall has continuously declined from the benchmark value of 8% to 5.5%, with an accumulated deviation of -8.5%, and the change slope after linear regression is -0.83% / month, indicating that the downward trend is obvious and has not bottomed out yet. By comparing and analyzing the three types of trend data in the same dimension, it is found that the decrease in the young customer group is highly correlated with the decrease in dwell time (correlation coefficient 0.86), and the decrease in the young customer group leads the decrease in dwell time by about 2 weeks, while the decrease in dwell time leads the decline in high-end conversion rate by about 1 week, forming a clear correlation chain. Quantifying the abnormal degree of relevant indicators, the Z-score of the change in the proportion of the young customer group is -2.4, the Z-score of the change in dwell time is -1.8, and the Z-score of the change in high-end conversion rate is -2.2, all exceeding the preset warning threshold of ±1.5. According to the abnormal degree, it is determined that the proportion of the young customer group triggers a first-level warning (red), the dwell time triggers a second-level warning (yellow), and the high-end conversion rate triggers a first-level warning (red). The warning level is associated with specific abnormal indicators, and through comprehensive analysis, the core problem is determined to be "the loss of young high-consumption customer groups leads to the deterioration of core business indicators". The correlation analysis finds that the newly opened comprehensive entertainment center in the vicinity has significantly diverted the young customer group. Finally, a complete site warning information is formed, including detailed data analysis, problem diagnosis, and adjustment suggestions, providing data support for subsequent site optimization decisions.

[0139] The above describes the method for screening traffic sites for marketing in the embodiments of the present application. Next, the system for screening traffic sites for marketing in the embodiments of the present application will be described. Please refer to Figure 2 , an embodiment of the system for screening traffic sites for marketing in the embodiments of the present application includes:

[0140] The collection module is used to collect multi-dimensional data on the total number of visitors, unique visitors, stay duration, and bounce rate of the shopping mall traffic site through the deployed site monitoring system, store them in a labeled manner according to the source channel, device, and region, and obtain the initial evaluation matrix;

[0141] The calculation module is used to calculate the basic passenger flow score, user stickiness coefficient, user engagement coefficient, and site attribute coefficient according to the initial evaluation matrix, and obtain the site quality evaluation database through weighted calculation;

[0142] The extraction module is used to extract the device type distribution, access time distribution, regional distribution, and consumption preference characteristics of users according to the quality evaluation database, establish a site user portrait, and calculate the matching score of the target merchant user portrait;

[0143] The prediction module is used to obtain the predicted score of the site operation effect according to the matching score, combined with the characteristics of passenger flow, conversion rate, consumption amount, etc. in the historical operation data, as well as the site quality score, business hours, and promotion activity form;

[0144] The evaluation module is used to conduct quality access evaluation and user matching evaluation on the site based on the predicted score of the operation effect, set the grading standard according to the expected effect, and screen in combination with the merchant site selection magnitude threshold to obtain the site screening combination result;

[0145] The monitoring module is used to monitor and analyze the quality evaluation indicators, user portrait matching degree, and actual operation effect of the site according to the site screening combination result, and make dynamic adjustments through warning triggering and effect comparison to obtain the shopping mall site optimization plan.

[0146] Through the collaborative cooperation of the above-mentioned various components, multi-dimensional data collection is carried out by deploying a site monitoring system, achieving comprehensive monitoring of key indicators such as the total number of mall visitors, the number of independent visitors, the stay duration, and the bounce rate. It breaks through the limitations of traditional methods that only focus on single passenger flow. With the help of tagged storage technology, attributes such as source channels, devices, and regions are structured and organized to form an initial evaluation matrix rich in information, laying a solid data foundation for subsequent analysis. Based on this evaluation matrix, the method introduces multi-dimensional evaluation indicators such as the basic passenger flow score, user stickiness coefficient, user engagement coefficient, and site attribute coefficient, and generates a site quality evaluation database through weighted calculation, realizing the scientific quantification of mall quality. Further, the device type distribution, access time distribution, regional distribution, and consumption preference characteristics of users are extracted from the quality evaluation database to establish a three-dimensional site user portrait, and the matching score with the target merchant user portrait is calculated through an intelligent matching algorithm, effectively solving the problem that it is difficult to accurately evaluate the customer group matching by traditional methods. On this basis, combined with characteristics such as passenger flow, conversion rate, and consumption amount in historical business data, as well as factors such as site quality score, business hours, and promotion activity forms, a prediction model is applied to generate a predicted score of the site operation effect, greatly improving the scientificity and accuracy of site selection decisions. Based on the predicted score, the method designs a dual screening mechanism of quality access evaluation and user matching evaluation, and introduces the expected effect grading standard and merchant site selection magnitude threshold as constraint conditions to obtain the optimal site screening combination result. Finally, by establishing a monitoring and analysis closed-loop, continuously monitor the quality evaluation indicators, user portrait matching degree, and actual operation effect of the selected site, and make dynamic adjustments based on the warning trigger mechanism and effect comparison analysis to generate a mall site optimization plan, effectively coping with the challenges brought by changes in the market environment and passenger flow characteristics. The entire solution makes full use of artificial intelligence algorithms to deeply mine data and identify patterns. Especially in the links of user portrait construction, matching degree calculation, and operation effect prediction, advanced algorithms such as clustering analysis, similarity calculation, and regression prediction are introduced. These algorithm features play a key role in improving data processing accuracy, discovering potential association patterns, and achieving accurate prediction, significantly enhancing the efficiency and accuracy of mall traffic site screening.

[0147] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the above method is implemented. It can be understood that the computer-readable storage medium in this embodiment can be a volatile readable storage medium or a non-volatile readable storage medium.

[0148] Those of ordinary skill in the art can understand that all or part of the processes of implementing the methods in the above embodiments can be completed by instructing relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above methods. Among them, any reference to a memory, storage, database, or other medium provided in the present invention and used in the embodiments can include non-volatile and / or volatile memories. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM, etc.

[0149] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the above-described systems, systems, and units can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated herein.

[0150] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The foregoing storage medium includes: USB flash drives, mobile hard disks, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical discs, etc., which can store program codes.

[0151] As described above, the above embodiments are only used to illustrate the technical solutions of the present application and are not intended to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements for some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for screening traffic venues for marketing, characterized in that: The traffic venue screening method for marketing includes: By deploying a venue monitoring system, we collect multi-dimensional data on the total number of visitors, number of unique visitors, length of stay, and bounce rate at the mall traffic venues, and store them in a tagged manner according to source channels, devices, and regions to obtain an initial evaluation matrix; According to the initial evaluation matrix, the basic passenger flow score, user stickiness coefficient, user participation coefficient and venue attribute coefficient are calculated, and a venue quality evaluation database is obtained through weighted calculation; According to the quality assessment database, extract the user's device type distribution, access time distribution, geographical distribution and consumption preference characteristics, establish a venue user portrait, and calculate the matching score of the target merchant user portrait; According to the matching score, combined with the characteristics of customer flow, conversion rate, consumption amount, etc. in the historical operation data, as well as the venue quality score, business hours and promotion activities, the venue operation effect prediction score is obtained; Based on the predicted scores of business effects, the quality access assessment and user matching degree assessment of the venues are conducted, the grading standards are set according to the expected effects, and the venues are screened in combination with the threshold of the merchant site selection level to obtain the venue screening combination results; According to the site screening combination results, the quality assessment indicators, user portrait matching and actual operating results of the site are monitored and analyzed, and dynamic adjustments are made through early warning triggering and effect comparison to obtain a shopping mall site optimization plan.

2. The method for screening traffic venues for marketing according to claim 1, characterized in that: The site monitoring system is deployed to collect multi-dimensional data on the total number of visitors, number of unique visitors, length of stay and bounce rate of the shopping mall traffic site, and stores them in a labeled manner according to the source channel, device and region to obtain an initial evaluation matrix, including: The monitoring system collects the passenger flow records of each shopping mall traffic venue within the specified time window to obtain the total number of visitors to the shopping mall traffic venue; Deduplication processing is performed on the user identification information in the total number of visitors to the shopping mall traffic venue to obtain the number of independent visitors to the shopping mall traffic venue; By recording the user's entry and exit time at the mall traffic venue, the time difference of each visit is calculated to obtain the length of stay at the mall traffic venue; Calculate the ratio of the number of users who only visit a single area in the shopping mall traffic venue to the total number of visiting users to obtain the shopping mall traffic venue bounce ratio; Extract source channel identifiers based on user visit information, and classify them according to online traffic, surrounding communities, business cooperation, and brand activities to obtain source channel data of shopping mall traffic venues; By analyzing the user's device identification information, classified statistics are made according to mobile terminals, shopping mall guide devices and membership cards to obtain the access device data of the shopping mall traffic venue; Analyze the geographical location based on the user's location information, and make distribution statistics by province, city and region to obtain the user regional data of the shopping mall traffic venue; The source channel data, access device data and user region data of the shopping mall traffic venue are multi-dimensionally associated, a label index system is established, and an initial evaluation matrix is ​​obtained.

3. The method for screening traffic venues for marketing according to claim 1, characterized in that: The method of calculating the basic passenger flow score, user stickiness coefficient, user participation coefficient and venue attribute coefficient according to the initial evaluation matrix, and obtaining a venue quality evaluation database through weighted calculation includes: Normalizing the total visitor data and the independent visitor data in the initial evaluation matrix respectively, and performing linear combination according to a ratio of 6:4 to obtain a basic passenger flow score; The dwell time data in the initial evaluation matrix is ​​converted into minutes, and the user stickiness coefficient is obtained by calculating the ratio with the industry average dwell time; Reversely transform the bounce rate data in the initial evaluation matrix, and obtain the user engagement coefficient by calculating the degree of deviation from the industry benchmark bounce rate; According to the source channel data in the initial evaluation matrix, weighted credibility scores are performed on each channel to obtain source channel credibility; According to the access device data in the initial evaluation matrix, statistics are collected on the number and distribution ratio of covered device types to obtain device coverage; Calculating the concentration and coverage of regional distribution according to the user regional data in the initial evaluation matrix to obtain regional distribution; Linearly combining the source channel credibility, equipment coverage and geographical distribution to obtain a site attribute coefficient; A weighted sum operation is performed on the basic passenger flow score, user stickiness coefficient, user participation coefficient and venue attribute coefficient to obtain a venue quality evaluation database.

4. The method for selecting traffic venues for marketing according to claim 1, characterized in that: The extracting of the user's device type distribution, access time distribution, geographical distribution and consumption preference characteristics according to the quality assessment database, establishing a site user portrait, and calculating the matching score of the target merchant user portrait includes: Extracting device type data of users at shopping mall traffic venues from the quality assessment database, performing statistical calculations on the distribution ratios of mobile devices, guide terminals, and membership cards, and obtaining device type distribution; Extracting user access time data from the quality assessment database, dividing the time interval into 24-hour intervals, and counting the access frequency of each time interval to obtain access time distribution; Extracting the user's region data from the quality assessment database, calculating the proportion of the number of visiting users in the provincial and municipal regions, and obtaining the regional distribution; Extracting user consumption data from the quality assessment database, performing weighted calculation on consumption categories, amount levels and purchase frequencies, and obtaining consumption preferences; The device type distribution, access time distribution, regional distribution and consumption preference are combined in multiple dimensions, and weight coefficients are established to obtain a site user portrait; Build the target merchant customer base characteristics through demographic characteristics, spending power level and category preference dimensions to obtain the target merchant user portrait; The feature vectors of the corresponding dimensions of the venue user portrait and the target merchant user portrait are calculated, and the dimension difference values ​​are calculated using the Euclidean distance formula to obtain a matching score.

5. The method for screening traffic venues for marketing according to claim 1, characterized in that: The matching score is combined with the characteristics of the customer flow, conversion rate, consumption amount, etc. in the historical business data, as well as the venue quality score, business hours and promotion activities to obtain the venue business effect prediction score, including: From the historical operation data of the mall traffic venue, the number of customers is counted according to the daily time granularity to obtain the customer flow data; The consumption behavior of users in the passenger flow data is counted, and the conversion rate data is obtained by calculating the ratio of the number of consuming users to the passenger flow; The consumption amount of the users in the conversion rate data is counted, and the average customer price data is obtained by calculating the ratio of the consumption amount to the number of consuming users; Grouping the customer flow, conversion rate and average customer spending in different time periods in the shopping mall traffic venue to obtain business period data; Extract the promotion activity type, discount strength and activity scope from the shopping mall traffic venue to obtain promotion activity form data; The matching score is combined with customer flow data, conversion rate data, customer unit price data, business hour data and promotion activity form data to obtain business characteristic data; The business characteristic data are normalized and weighted calculated in combination with the site quality score to obtain a predicted score for the site business effect.

6. The method for screening traffic venues for marketing according to claim 1, characterized in that: Based on the predicted scores of the business effects, the quality access assessment and user matching assessment of the venues are performed, the grading standards are set according to the expected effects, and the venues are screened in combination with the threshold of the merchant site selection level to obtain the venue screening combination results, including: Compare and calculate the predicted scores of the operating results with the industry average, conduct quality access assessment by setting a minimum access score, and obtain a list of access sites; According to the matching scores of each shopping mall traffic venue in the access venue list, the user overlap is calculated to obtain a user matching evaluation result; By analyzing the historical operating performance of each shopping mall traffic venue in the list of access venues, three-level effect standards are set to obtain expected effect classification data; The average daily passenger flow of the shopping mall traffic venues in the list of access venues is counted, and the upper and lower limits of the site selection level are set according to the merchant investment budget to obtain the merchant site selection level threshold; Cross-validate the expected effect classification data with the merchant site selection magnitude threshold, screen out eligible shopping mall traffic site combinations, and obtain a preliminary site list; An overlap analysis is performed on the population covered by each shopping mall traffic venue in the preliminary venue list, and a venue screening combination result is obtained by deduplication processing.

7. The method for screening traffic venues for marketing according to claim 1, characterized in that: According to the site screening combination results, the quality assessment indicators, user portrait matching degree and actual operating effect of the site are monitored and analyzed, and dynamic adjustments are made through early warning triggering and effect comparison to obtain a shopping mall site optimization plan, including: The quality assessment index of each shopping mall traffic venue in the venue screening combination result is monitored in real time, and the quality fluctuation data is obtained by calculating the deviation from the historical benchmark value; According to the user portrait matching degree of the shopping mall traffic venue in the venue screening combination result, the user group characteristics are regularly updated and compared to obtain user change data; The operating data of the shopping mall traffic venues in the venue screening combination results are collected, and the actual operating effect data is obtained by calculating the difference between the actual conversion rate and the expected conversion rate; Comprehensively analyze the quality fluctuation data, user change data and actual business effect data, make early warning judgments according to set thresholds, and obtain site early warning information; According to the site warning information, the shopping mall traffic sites with different warning levels are classified and processed, and the effect comparison results are obtained by comparing and analyzing with historical performance; The effect comparison results are dynamically scored according to weight indicators, and the shopping mall traffic venues are reordered and replaced to obtain a shopping mall venue optimization plan.

8. The method for selecting traffic venues for marketing according to claim 7, characterized in that: The quality fluctuation data, user change data and actual operation effect data are comprehensively analyzed, and early warning judgment is performed according to the set threshold to obtain site early warning information, including: The quality fluctuation data are grouped and counted according to the daily fluctuation amplitude, and the quality fluctuation trend data are obtained by calculating the number of consecutive fluctuation days and the fluctuation direction; Decomposing the user change data according to the change dimensions, and obtaining the user change trend data by calculating the change amplitude and change rate of each dimension; Performing time series analysis on the actual operating effect data, and obtaining effect change trend data by calculating the cumulative value and change slope of the effect deviation; Compare the quality fluctuation trend data, user change trend data and effect change trend data in the same dimension, and obtain trend correlation data through correlation analysis; Quantitatively calculate the abnormality of each indicator in the trend-related data, and obtain warning level data by comparing it with a preset warning threshold; The warning level data is associated with abnormal indicators and marked, and site warning information is obtained through summary analysis.

9. A traffic venue screening system for marketing, used to implement the traffic venue screening method for marketing as described in any one of claims 1 to 8, characterized in that: The traffic venue screening system for marketing includes: The collection module is used to collect multi-dimensional data on the total number of visitors, number of unique visitors, length of stay and bounce rate of the mall traffic venues by deploying a venue monitoring system, and store them in a labeled manner according to the source channel, device and region to obtain an initial evaluation matrix; A calculation module, used to calculate the basic passenger flow score, user stickiness coefficient, user participation coefficient and venue attribute coefficient according to the initial evaluation matrix, and obtain a venue quality evaluation database through weighted calculation; An extraction module, for extracting the user's device type distribution, access time distribution, geographical distribution and consumption preference characteristics according to the quality evaluation database, establishing a site user portrait, and calculating a matching score of the target merchant user portrait; A prediction module, for obtaining a predicted score of venue operation effect according to the matching score, combined with characteristics such as customer flow, conversion rate, consumption amount, etc. in historical operation data, as well as venue quality score, business hours and promotion activity form; An evaluation module, for performing a quality access evaluation and a user matching evaluation on the venue based on the predicted scores of the business effects, setting a grading standard according to the expected effects, and screening in combination with a threshold of the merchant site selection level to obtain a venue screening combination result; The monitoring module is used to monitor and analyze the quality evaluation indicators, user portrait matching and actual operating results of the venue according to the venue screening combination results, and to make dynamic adjustments through early warning triggering and effect comparison to obtain a shopping mall venue optimization plan.

10. A computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the processor is enabled to execute the traffic venue screening method for marketing as claimed in any one of claims 1 to 8.

Images