Advertisement putting method, system and equipment based on retail cabinet and medium

Abstract

The invention relates to the technical field of advertisement putting, in particular to an advertisement putting method, system and device based on a retail cabinet and a medium. Behavior data such as user purchase records, residence time and commodity attention are collected through a sensor and a transaction system of a retail cabinet, user consumption ability and shopping preference characteristics are analyzed and extracted by using a deep learning algorithm, and a user portrait model is constructed; the system intelligently matches the user portrait with an advertisement content library, selects the most suitable advertisement content for delivery, and continuously collects effect data such as user watching duration, interaction times and purchase behaviors; based on the feedback data, the system automatically adjusts advertising strategies such as advertisement display time and target user groups. According to the invention, accurate matching between the advertisement content and the user demand is realized, a closed-loop optimization mechanism of advertisement putting is established, and the putting effect is improved.

Description

Technical Field

[0001] The present application relates to the technical field of advertising placement, and in particular to a method, system, device, and medium for advertising placement based on retail counters. Background Art

[0002] With the development of the Internet of Things and big data technologies, the retail industry is gradually transforming towards intelligent solutions. As a new type of retail terminal, smart retail lockers, with their convenience and intelligent features, are widely used in scenarios such as unmanned retail and community supermarkets, and are gradually becoming a key channel connecting products and consumers. Current smart retail locker systems typically feature displays for advertising, using a pre-set ad rotation mechanism and a fixed schedule. The system automatically adjusts the order and frequency of ads based on inventory status and sales data, providing basic advertising functionality.

[0003] However, this fixed-mode advertising delivery mechanism is unable to identify and analyze users' personalized characteristics, resulting in a mismatch between advertising content and actual user needs; this situation needs further improvement. Summary of the Invention

[0004] To address the problem that existing fixed-mode advertising delivery mechanisms are unable to identify and analyze users' personalized characteristics, resulting in a mismatch between advertising content and actual user needs, this application provides a retail counter-based advertising delivery method, system, device, and medium, using the following technical solutions: In a first aspect, the present application provides a method for placing advertisements in a retail cabinet, comprising the following steps: Collect user purchase history, length of stay, and interest preference information to obtain user behavior data; Analyze the user behavior data using a deep learning algorithm to obtain user feature data; Building a target user profile based on the user feature data; According to the target user profile, the advertising resources in the advertising content library are matched to determine the advertising content to be delivered; Based on the advertisement content to be delivered, executing an advertisement delivery operation to obtain advertisement delivery effect data; Dynamically adjust the advertising delivery strategy based on the advertising delivery effect data.

[0005] By adopting the technical scheme, the application first collects various behavior data of the user in the shopping process through the sensor and transaction system built in the retail cabinet, including the purchase record, the residence time in different commodity areas, and the attention degree to specific commodity categories; then, the deep learning algorithm is used to analyze the behavior data, and the characteristic information of the user, such as the consumption ability and the shopping preference, is extracted; based on the characteristic information, the system constructs the user portrait model, and intelligently matches it with the resources in the advertisement content library, and selects the advertisement content most suitable for the current user characteristics; in the advertisement delivery process, the system continuously collects the effect data including the user watching time, the interaction times and the final purchase behavior, and automatically adjusts the delivery strategy according to the feedback data, such as adjusting the advertisement display time, updating the target user group, etc.; through the behavior data analysis and the deep learning algorithm, the accurate matching of the advertisement content and the user demand is realized; through the real-time effect monitoring and the strategy adjustment, the closed-loop optimization mechanism of the advertisement delivery is established, and the pertinence and the conversion effect of the advertisement delivery are significantly improved.

[0006] Optionally, the purchase history, the residence time and the interest preference information of the user are collected to obtain the user behavior data, specifically including the following steps: extracting commodity category data and consumption amount data from the purchase history; determining the residence time data of the user in each commodity area according to the residence time information of the user and the relative positions of the commodity areas in the retail cabinet layout information; determining the interest preference type and degree of the user according to the purchase history and the residence time information; associating the commodity category data, the consumption amount data, the residence time data and the interest preference data to obtain the user behavior data.

[0007] By adopting the technical solution, the application firstly extracts the commodity category and specific consumption amount information from the historical transaction records of the user, establishes the basic consumption portrait of the user; through the sensors equipped in the retail cabinet, records the moving track of the user in the shopping process, and combines the pre-set retail cabinet commodity area layout information to calculate the specific residence time of the user in each commodity area; the system further analyzes the association between the purchase behavior and the residence behavior of the user, quantifies the interest degree of the user to different categories of commodities, such as a user frequently stays in the cosmetics area but has a low purchase frequency, which may indicate that the user has a demand for such commodities but is sensitive to the price; finally, the system integrates and associates the extracted commodity category, consumption amount, area residence time and interest preference and the like to form a complete user behavior data set; the comprehensive capture of the user behavior characteristics is realized; by combining the physical layout information of the retail cabinet, the analysis accuracy of the behavior data is improved; the system can record the explicit purchase behavior and implicit interest performance of the user at the same time, and provides a data basis for subsequent personalized advertisement delivery.

[0008] Optionally, according to the user behavior data, a deep learning algorithm is used for analysis to obtain user feature data, specifically including the following steps: The commodity category data is encoded to obtain commodity category features; The consumption amount data is normalized to obtain consumption ability features; The residence time data is divided according to time periods to obtain time features; The interest preference data is quantitatively calculated to obtain preference features; The commodity category features, consumption ability features, time features and preference features are combined to obtain user feature data.

[0009] By adopting the technical solution, the application firstly encodes the commodity category data, converts the text category into a numerical vector representation, so that the relationship between different categories can be measured by vector distance; then, the consumption amount data of the user is normalized, and different orders of magnitude of the consumption data are unified to the same scale space; for the residence time data, the system is divided according to the pre-set time interval, converts the continuous time data into discrete time features, and at the same time, the interest preference data of the user is quantitatively calculated, and the qualitative interest performance is converted into a calculable numerical feature; finally, the system combines and integrates the processed features to form a unified user feature vector; through the standardized expression of multi-dimensional features, the comparability and calculability of the feature data are improved.

[0010] Optionally, according to the target user portrait, the advertisement resources in the advertisement content library are matched to determine the advertisement content to be delivered, specifically including the following steps: According to the target user portrait, a target audience group and an expected conversion rate are obtained, and according to the target audience group and the expected conversion rate, an overall advertisement selection parameter is obtained; Retail cabinet display position information is obtained, and according to the display position information and the overall advertisement selection parameter, personalized delivery parameters are set for different positions; According to the personalized delivery parameters and the overall advertisement selection parameter, advertisement content screening is performed; The estimated conversion parameter is obtained by comparing the estimated conversion parameter with the expected conversion rate, and an expected difference value is obtained; If the expected difference value exceeds a preset threshold, an advertisement optimization is triggered, and a parameter adjustment suggestion is obtained according to the advertisement optimization result; According to the parameter adjustment suggestion, the personalized delivery parameters are updated.

[0011] By adopting the above technical solution, the traditional advertisement matching method often only considers a single dimension of matching rule, which cannot adapt to the complex retail scene demand, resulting in unstable advertisement delivery effect; the present application first determines the target audience group characteristics and the expected conversion rate index based on the target user portrait, and generates the overall advertisement selection parameter including audience characteristics, delivery time period, display form, etc.; then, combined with the actual layout and display position information of the retail cabinet, unique delivery parameters are set for each advertisement display position, such as display time length, playback frequency, etc. of different positions; the system uses these parameters to screen the advertisement content library, and preliminarily determines the advertisement content set that meets the conditions; for the screened advertisement content, the system will perform effect estimation, calculate the expected conversion parameter, and compare it with the set target conversion rate; when there is a significant difference between the expected effect and the target, the system will automatically trigger the optimization process, generate parameter adjustment suggestions such as adjusting the display time period, replacing the display position, etc. by analyzing the historical delivery data; finally, the system updates the personalized delivery parameters according to these optimization suggestions, realizes the dynamic adjustment of the advertisement delivery strategy; through hierarchical parameter configuration, the fine control of advertisement delivery is realized; through the effect estimation and dynamic optimization mechanism, the adaptability of advertisement delivery is improved; the system can automatically adjust the strategy according to the actual delivery effect, and ensures the continuous optimization of the advertisement delivery effect.

[0012] Optionally, the target audience group and the expected conversion rate are obtained, and according to the target audience group and the expected conversion rate, the overall advertisement selection parameter is obtained, specifically including the following steps: The target stay time is extracted from the expected conversion rate, and the target stay time is taken as the main evaluation index; According to the target stay time and the target audience group, the crowd matching parameter, the commodity association parameter and the position optimization parameter are obtained; According to the target dwell time, the crowd matching parameter, the commodity correlation parameter and the position optimization parameter, the overall advertisement selection parameter is obtained.

[0013] By adopting the technical solution, the target dwell time threshold is extracted from the expected conversion rate index, and the user's stay time before the advertisement content is taken as the core index for evaluating the attraction of the advertisement. Further, the matching parameter reflecting the attributes of the target audience group is generated, such as the quantified index of the age range and the consumption ability dimension. Meanwhile, the system analyzes the shopping behavior characteristics of the target audience, and generates the commodity correlation parameter for ensuring that the advertisement content matches the potential shopping demand of the user. Considering the physical layout characteristics of the retail cabinet, the system also generates the position optimization parameter for adjusting the advertisement placement strategy of different display positions. Finally, the system integrates the target dwell time and the subdivided parameters to form a complete advertisement selection parameter system. The parameter configuration can be flexibly adjusted according to different scene characteristics to ensure the actual effect of the advertisement placement.

[0014] Optionally, the position optimization parameter includes the dynamic line parameter and the visual focus parameter, the crowd matching parameter includes the consumption ability parameter and the shopping habit parameter, and the obtaining of the crowd matching parameter, the commodity correlation parameter and the position optimization parameter according to the target dwell time and the target audience group includes the following steps. According to the target dwell time and the target audience group, the dynamic line parameter, the visual focus parameter, the consumption ability parameter, the shopping habit parameter and the commodity correlation parameter are obtained. According to the consumption ability parameter and the shopping habit parameter, the crowd matching parameter is obtained.

[0015] By adopting the technical solution, the traditional parameter configuration method often ignores the natural behavior pattern of the user in the retail space, resulting in the lack of scientific basis for the selection of the advertisement position. The application first analyzes the typical moving path of the user in front of the retail cabinet based on the target dwell time and the target audience group characteristics, and generates the dynamic line parameter reflecting the moving characteristics of the user. Meanwhile, the visual focus parameter of each position is determined by analyzing the visual attention distribution law of the user, which is used to evaluate the visual attraction of different positions. In terms of user characteristics, the system generates the consumption ability parameter reflecting the purchasing ability of the target audience and the shopping habit parameter characterizing the shopping rule of the target audience according to the historical behavior of the target audience. In addition, the system also analyzes the correlation between commodities to generate the commodity correlation parameter for guiding the position allocation of the advertisement content. Finally, the system integrates the consumption ability parameter and the shopping habit parameter to form a complete crowd matching parameter system. By introducing the dynamic line and the visual focus analysis, the scientificity of the selection of the advertisement position is improved.

[0016] Optionally, the personalized delivery parameter includes an advertisement display time period parameter and an advertisement display form parameter, the retail cabinet display position information is acquired, and personalized delivery parameters are set for different positions according to the display position information and the overall advertisement selection parameter, specifically including the following steps: Acquire retail cabinet passenger flow distribution information and shopping dynamic line information; According to the position optimization parameter, the passenger flow distribution information and the shopping dynamic line information, the display strategy parameter is obtained for the display position; According to the crowd matching parameter, the commodity association parameter, the passenger flow distribution information and the shopping dynamic line information, the advertisement display time period parameter and the advertisement display form parameter are obtained for the display position.

[0017] By adopting the above technical solution, the real-time passenger flow data of each area of the retail cabinet is first collected, and a complete space behavior model is constructed in combination with the typical shopping path of the user. Based on these basic data, the system combines the previously obtained position optimization parameter to formulate a unique display strategy parameter for each display position, including display priority, update frequency and other elements; at the same time, the system comprehensively analyzes the crowd matching parameter and the commodity association parameter, and combines the passenger flow distribution law and the shopping dynamic line characteristics to determine the best advertisement display time period for each display position, such as adjusting the display content according to the passenger characteristics of different time periods; in addition, the system will also set the corresponding advertisement display form parameter according to the visual characteristics of the position and the surrounding environment, such as adjusting the display size, playing time and the like; by combining the position characteristics and the passenger flow law, the scene adjustment of the advertisement display strategy is realized; through the fine configuration of the time period parameter and the form parameter, the utilization efficiency of the advertisement resources is improved; the system can dynamically optimize the display strategy according to the real-time passenger flow change, and ensure the maximization of the advertisement delivery effect.

[0018] In a second aspect, the application provides a retail cabinet-based advertisement delivery system, comprising: A data acquisition module is configured to acquire purchase history, stay duration and interest preference information of a user, and obtain user behavior data; A feature analysis module is configured to analyze the user behavior data by using a deep learning algorithm, and obtain user feature data; A user portrait module is configured to construct a target user portrait based on the user feature data; An advertisement matching module is configured to match advertisement resources in an advertisement content library according to the target user portrait, and determine advertisement content to be delivered; A delivery execution module is configured to execute an advertisement delivery operation based on the advertisement content to be delivered, and obtain advertisement delivery effect data; A strategy adjustment module is configured to dynamically adjust an advertisement delivery strategy according to the advertisement delivery effect data.

[0019] In a third aspect, the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the steps of the retail cabinet-based advertisement delivery method when executing the computer program.

[0020] In a fourth aspect, the present application provides a computer-readable storage medium, which stores a computer program, wherein the computer program is executed by a processor to implement the steps of the retail cabinet-based advertisement delivery method.

[0021] In summary, the present application includes at least one of the following beneficial technical effects: The present application collects user behavior data such as purchase records, dwell time, and product attention through the sensors and transaction system of the retail cabinet, uses deep learning algorithms to analyze and extract user consumption ability and shopping preference features, and constructs a user portrait model; the system intelligently matches the user portrait with the advertisement content library, selects the most suitable advertisement content for delivery, and continuously collects effect data such as user viewing time, interaction frequency, and purchase behavior; based on these feedback data, the system automatically adjusts the advertisement display time and target user group delivery strategy, realizes accurate matching of advertisement content and user demand, establishes a closed-loop optimization mechanism for advertisement delivery, and improves delivery effect; The present application determines target audience characteristics and conversion rate indicators based on user portraits, generates advertisement selection parameters including audience characteristics, delivery time period, and display form, and sets unique delivery parameters including display time and playback frequency for each display position in combination with the layout of the retail cabinet; the system uses these parameters to filter advertisement content and perform effect estimation, compares expected conversion parameters with targets, and automatically triggers an optimization process when there is a significant difference, analyzes historical data to generate adjustment suggestions such as changing display time and position, and updates delivery parameters accordingly; through this hierarchical parameter configuration and dynamic optimization mechanism, the present application realizes fine-grained control and continuous optimization of advertisement delivery; The present application analyzes typical movement paths of users based on target dwell time and audience characteristics, generates movement line parameters, determines visual focus parameters for each position by analyzing visual attention distribution rules, generates consumption ability parameters and shopping habit parameters based on historical behaviors of target audiences, and analyzes product correlation to generate product correlation parameters for advertisement position allocation; finally, the consumption ability and shopping habit parameters are weighted and integrated to form a crowd matching parameter system; through this analysis method combining movement lines and visual focus, the present application improves the scientificity and accuracy of advertisement position selection. BRIEF DESCRIPTION OF DRAWINGS

[0022] Figure 1 is a flowchart of a retail cabinet-based advertisement delivery method according to an embodiment of the present application; Figure 2 is a flowchart of step S100 in an embodiment of a retail cabinet-based advertisement delivery method of the present application; Figure 3 is a flowchart of step S200 in an embodiment of a retail cabinet-based advertisement delivery method of the present application; Figure 4 is a flowchart of step S400 in an embodiment of a retail cabinet-based advertisement delivery method of the present application; Figure 5 is a flowchart of step S410 in an embodiment of a retail cabinet-based advertisement delivery method of the present application; Figure 6 is a flowchart of step S412 in an embodiment of a retail cabinet-based advertisement delivery method of the present application; Figure 7 is a flowchart of step S420 in an embodiment of a retail cabinet-based advertisement delivery method of the present application; Figure 8 is a block diagram of a retail cabinet-based advertisement delivery system according to an embodiment of the present application; Figure 9 is an internal structure diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0023] The terminology used in the following embodiments of the present application is for the purpose of describing particular embodiments only and is not intended to be limiting of the present application. As used in the specification and the appended claims, the singular forms "a," "an" and "the" are intended to include both singular and plural forms, unless the context clearly indicates otherwise. It will be further understood that the terms "and / or," as used in the specification and in the claims, is used to mean "one or the other or both" of the items so conjoined.

[0024] Hereinafter, the terms "first" and "second" are used only for the purpose of description and should not be construed as suggesting or implying relative importance or implicitly indicating the number of the technical features indicated. Thus, the features defined with "first" and "second" can explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, the meaning of "a plurality of" is two or more, unless otherwise specified.

[0025] The embodiments of the present application will be described in further detail below with reference to the accompanying drawings.

[0026] In a first aspect, the present application provides a retail cabinet-based advertisement delivery method, referring to Figure 1 , comprising the following steps: S100, collect the purchase history, stay time and interest preference information of the user to obtain user behavior data.

[0027] In this embodiment, the user behavior data refers to various interaction information generated by the user in front of the retail cabinet, including purchase records, area stay time, product attention frequency, product detail viewing times and touch screen times. These data are collected by the sensor network carried on the front end of the retail cabinet, including infrared sensors, cameras, touch screens, and stored and processed by the back-end server.

[0028] Specifically, the system first establishes a user behavior database and designs a standardized data table containing user ID, behavior type, behavior time and behavior location. When the user is in front of the retail cabinet, the system records various behavior data in real time. For example, user A stays in the cosmetics area for 3 minutes, views the details of 4 products during this period, and finally purchases 1 product. All this information will be stored in a structured manner.

[0029] S200, according to the user behavior data, using deep learning algorithm for analysis, to obtain user feature data.

[0030] In this embodiment, the user feature data is obtained by processing and converting the original behavior data. The system pre-establishes a feature mapping table to map different types of behavior data to a standardized feature space.

[0031] S300, based on the user feature data, construct the target user portrait.

[0032] In this embodiment, the target user portrait is designed in a hierarchical structure, including a basic attribute layer, a behavior feature layer and a consumption feature layer. The basic attribute layer records the user's consumption ability level, product preference type and active period; the behavior feature layer records the user's browsing habits, stay rules and interaction methods; the consumption feature layer records the user's purchase frequency, single price distribution and promotion sensitivity.

[0033] Specifically, the system establishes a user portrait mapping rule library to convert user feature data into structured portrait labels. For example, when the user's monthly consumption amount exceeds 1000 yuan, the consumption ability is marked as "high"; when the user's average stay time in the cosmetics area exceeds 5 minutes, the product interest degree is marked as "strong"; when 90% of the user's purchase behavior occurs during the promotion period, the promotion sensitivity is marked as "high".

[0034] S400, according to the target user portrait, match the advertising resources in the advertising content library to determine the advertising content to be put.

[0035] In this embodiment, the advertisement content library adopts a classification index structure, and an index table is established according to target audiences, display forms and commodity categories of the advertisement content. Each piece of advertisement content contains five basic attributes of advertisement identification, content type, target audience, applicable scenario and expected effect, which are used for matching with the user portrait.

[0036] Specifically, the system uses a rule-based matching algorithm to calculate the matching degree of the user portrait and the advertisement content. First, the price interval of the advertisement content is filtered based on the consumption ability level of the user, then the related commodity advertisement is selected according to the category preference of the user, and finally the display opportunity is determined considering the active period of the user. The advertisement content with the highest matching degree will be selected as the to-be-launched content.

[0037] S500, based on the advertisement content to be launched, performing an advertisement launching operation to obtain advertisement launching effect data.

[0038] In this embodiment, the advertisement launching effect data contains three types of indexes of exposure data, interaction data and conversion data. The exposure data records the actual watching time and frequency of the user watching the advertisement; the interaction data records the behaviors of the user clicking, touching and viewing details; and the conversion data records the purchase behavior of the user.

[0039] S600, dynamically adjusting the advertisement launching strategy according to the advertisement launching effect data.

[0040] In this embodiment, the advertisement launching strategy contains launching period, display position and display form. The system pre-sets effect evaluation indexes, including minimum watching time, expected interaction rate and target conversion rate. When the actual effect data deviates from the expected indexes, the strategy adjustment mechanism is triggered.

[0041] In one embodiment, referring to Figure 2 , in step S100, the purchase history, stay time and interest preference information of the user are collected to obtain user behavior data, specifically including the following steps: S110, extracting commodity category data and consumption amount data from the purchase history.

[0042] In this embodiment, the commodity category data refers to the hierarchical classification information of the commodity, including first-level categories (cosmetics, food, beverages, daily necessities), second-level categories (cosmetics are divided into skin care products, color cosmetics and perfumes), and third-level categories (skin care products are divided into face cream, essence and toner). The consumption amount data refers to the actual payment amount of the user in each purchase behavior, which is used to analyze the consumption ability level of the user.

[0043] Specifically, the system pre-establishes a commodity classification mapping table to correspond the commodity bar code with the standardized category system. When the user completes the purchase, the system automatically extracts the commodity bar code information from the transaction record, converts it into structured category data through the mapping table, and records the payment amount of the transaction at the same time.

[0044] S120, according to the user's stay time length information, combined with the relative position of each commodity area in the retail cabinet layout information, determine the user's residence time data in each commodity area.

[0045] In this embodiment, the retail cabinet layout information includes the spatial coordinates, area range and displayed commodity category of each commodity display area. The system divides the retail cabinet space into multiple detection blocks, each block corresponding to a specific commodity category. The residence time data records the user's stay time length in each block, which is used to analyze the user's attention to different commodity categories.

[0046] Specifically, the system installs an infrared sensor array in front of the retail cabinet and processes the detection space in a grid manner. A region position mapping table is established to correspond each grid coordinate with the commodity display area. When the sensor detects that the user stays in a certain position, the system records the entry time and exit time, and calculates the residence time length of the area.

[0047] S130, according to the purchase history and stay time length information, determine the user's interest preference type and degree.

[0048] In this embodiment, the interest preference type refers to the commodity category that the user is interested in, including the main interest type (the category with the longest residence time or the highest purchase frequency) and the secondary interest type (the category with the second longest residence time or the second highest purchase frequency).

[0049] Specifically, the system establishes an interest degree calculation rule table to calculate the interest degree score in combination with the purchase frequency weight and the residence time length weight.

[0050] S140, associate the commodity category data, consumption amount data, residence time data and interest preference data to obtain user behavior data.

[0051] In this embodiment, the user behavior data is stored in an association data structure, including user identification, behavior time, behavior type, behavior location and behavior data. The system establishes an independent behavior record table for each user, organizes various data in chronological order, and forms a complete behavior track.

[0052] In one embodiment, referring to Figure 3 , in step S200, according to the user behavior data, the deep learning algorithm is used for analysis to obtain user feature data, including the following steps: S210, encode the commodity category data to obtain commodity category features.

[0053] Specifically, the system establishes a commodity code mapping table to convert commodity categories into standard codes.

[0054] S220, normalizing the consumption amount data to obtain consumption ability features.

[0055] In this embodiment, the consumption ability features include single consumption amount features, monthly total consumption amount features, and category consumption proportion features. The system pre-sets an amount grading standard to map consumption amounts in different intervals to standardized values in the range of 0-1, which are used to measure the consumption ability level of the user.

[0056] S230, dividing the residence time data according to time periods to obtain time features.

[0057] In this embodiment, the time features include intraday time period features, weekday features, and holiday features. The intraday time period is divided into four time periods according to the business hours: morning, noon, afternoon, and night. The residence time in each time period is separately counted to form the time activity features of the user.

[0058] S240, quantitatively calculating the interest preference data to obtain preference features.

[0059] In this embodiment, the preference features include category preference intensity, purchase tendency, and price sensitivity. The system calculates the quantitative values of these indicators by analyzing the historical behavior data of the user to describe the consumption preference features of the user.

[0060] Specifically, the system designs preference quantitative calculation rules. The category preference intensity is calculated according to the purchase frequency and residence time of the user in the category; the purchase tendency is calculated by the conversion rate from browsing to purchase; and the price sensitivity is calculated by the purchase proportion of promotional goods.

[0061] S250, combining the commodity category features, consumption ability features, time features, and preference features to obtain user feature data.

[0062] In one embodiment, with reference to Figure 4 In step S400, according to the target user portrait, the system matches the advertising resources in the advertising content library to determine the advertising content to be put, which specifically includes the following steps: S410, obtaining a target audience group and an expected conversion rate according to the target user portrait, and obtaining overall advertising selection parameters according to the target audience group and the expected conversion rate.

[0063] In this embodiment, the overall advertisement selection parameter includes audience characteristic parameters (consumption ability interval, category preference type, active time period), conversion target parameters (display conversion rate, click conversion rate, purchase conversion rate), and delivery control parameters (display frequency, display time length, interaction form).

[0064] Specifically, the system establishes an advertisement parameter mapping table to convert user portrait characteristics into advertisement selection parameters. Through parameter mapping rules, the system can automatically generate a corresponding advertisement selection parameter set according to the consumption ability, category preference, active time period and other characteristics in the user portrait, including target audience targeting, display control and expected effect and other multiple dimensions of parameters.

[0065] S420, obtaining retail cabinet display position information, and setting personalized delivery parameters for different positions according to the display position information and the overall advertisement selection parameter.

[0066] In this embodiment, the display position information includes four dimensions of physical position coordinates, line of sight visibility, surrounding commodity type, and people flow density. The personalized delivery parameters include display time period parameters (play time, frequency), display form parameters (picture size, volume size, interaction mode), and directional optimization parameters (audience screening, scene matching).

[0067] Specifically, the system establishes a position characteristic evaluation table to establish a characteristic scoring system for each display position. The scoring dimensions include position visibility, people flow density, and surrounding correlation. The system configures different display parameters for different positions according to the scoring results to realize accurate delivery of advertisement resources.

[0068] S430, performing advertisement content screening according to the personalized delivery parameters and the overall advertisement selection parameter.

[0069] In this embodiment, the advertisement content screening adopts a multi-level filtering mechanism and is screened in turn according to the target audience matching degree, display position adaptation degree, and expected effect matching degree. The system pre-labels attribute tags for each advertisement content, including target people, applicable scene and display requirements.

[0070] S440, performing effect estimation on the screening result to obtain estimated conversion parameters; comparing the estimated conversion parameters with the expected conversion rate to obtain an expected difference value.

[0071] In this embodiment, the estimated conversion parameters include estimated display conversion rate, click conversion rate and purchase conversion rate. The system establishes a conversion estimation model based on historical delivery data, which comprehensively considers multiple dimensions of position influence coefficient, time period influence coefficient and audience matching coefficient to predict the expected effect of the advertisement content in a specific scene. The expected difference value is used to measure the deviation degree of the estimated effect from the target effect.

[0072] Specifically, the system constructs a conversion rate estimation rule library, classifies and counts historical delivery data according to position, time period, and audience characteristics, and establishes a basic conversion rate benchmark. On this basis, the estimated conversion rate is obtained through the combination calculation of position coefficient, time period coefficient, and audience coefficient. The system sets an expected difference threshold, and triggers the optimization process when the actual difference exceeds the threshold range.

[0073] S450, if the expected difference value exceeds the preset threshold, triggering the advertisement optimization; according to the advertisement optimization result, obtaining parameter adjustment suggestions.

[0074] In this embodiment, the advertisement optimization scheme includes delivery time period optimization, display position optimization, and display form optimization. The system constructs an optimization strategy library based on historical optimization records, and provides corresponding adjustment schemes for different types of effect bias. The optimization strategy library records the implementation conditions and expected effects of various optimization schemes.

[0075] S460, updating the personalized delivery parameters according to the parameter adjustment suggestions.

[0076] In this embodiment, the parameter update adopts a gradual adjustment strategy, and the system sets a maximum amplitude limit for parameter adjustment to avoid drastic fluctuations affecting the delivery effect.

[0077] In one embodiment, referring to Figure 5 , in step S410, the target audience group and the expected conversion rate are obtained, and the overall advertisement selection parameters are obtained according to the target audience group and the expected conversion rate, specifically including the following steps: S411, extracting the target dwell time from the expected conversion rate, and taking the target dwell time as the main evaluation index.

[0078] In this embodiment, the target dwell time is a key indicator for measuring the attractiveness of the advertisement, reflecting the degree of user attention to the advertisement content. The system establishes a mapping relationship table between conversion rate and dwell time, and converts different levels of expected conversion rate into corresponding target dwell time intervals.

[0079] Specifically, the system constructs a dwell time evaluation system, and divides the dwell time into multiple grade intervals. By analyzing the correlation between dwell time and conversion rate in historical data, a corresponding relationship between the two is established. The system reversely deduces the target dwell time that needs to be achieved according to the set expected conversion rate, as a direct evaluation index of advertisement effect.

[0080] S412, obtaining the population matching parameter, the commodity correlation parameter, and the position optimization parameter according to the target dwell time and the target audience group.

[0081] In this embodiment, the crowd matching parameters include consumption ability parameters and shopping habit parameters, which are used to evaluate the matching degree of the advertisement content and the target audience; the commodity association parameters reflect the association strength of the advertisement commodity and the user interest; the location optimization parameters include the dynamic line parameters and the visual focus parameters, which are used to evaluate the effect of the display location.

[0082] Specifically, the system establishes a parameter generation rule library, and determines the values of various parameters through multi-dimensional analysis. First, the parameter benchmark value is set according to the target residence time, and then adjusted in combination with the target audience characteristics.

[0083] S413、According to the target residence time, the crowd matching parameters, the commodity association parameters and the location optimization parameters, the overall advertisement selection parameters are obtained.

[0084] In this embodiment, the overall advertisement selection parameters adopt a multi-dimensional combination structure, including target orientation parameters, audience matching parameters and location optimization parameters as three main dimensions. The system ensures the coordination between the parameters in each dimension through parameter correlation analysis, and forms a complete parameter system.

[0085] In one embodiment, the location optimization parameters include dynamic line parameters and visual focus parameters, and the crowd matching parameters include consumption ability parameters and shopping habit parameters, with reference to Figure 6 , in step S412, the crowd matching parameters, the commodity association parameters and the location optimization parameters are obtained according to the target residence time and the target audience group, which specifically include the following steps: S4121、According to the target residence time and the target audience group, the dynamic line parameters, the visual focus parameters, the consumption ability parameters, the shopping habit parameters and the commodity association parameters are obtained.

[0086] In this embodiment, the dynamic line parameters describe the moving track characteristics of the user in front of the retail cabinet, including moving speed, residence point distribution and path selection as three dimensions; the visual focus parameters describe the visual attention distribution of the user, including gaze point position, gaze duration and visual line transfer path; the consumption ability parameters reflect the purchasing power level of the user; the shopping habit parameters describe the shopping behavior mode of the user; and the commodity association parameters reflect the associated purchase probability between different commodities.

[0087] Specifically, the system establishes a multi-dimensional parameter extraction rule library. The motion line parameter is obtained by analyzing the typical moving path of the target audience. The system divides multiple moving monitoring areas in front of the retail counter and records the moving characteristics of the user in each area. The line of sight focus parameter is obtained by the hot area analysis technology. The system divides the display area of the retail counter into grids and counts the line of sight dwell characteristics of each grid. The consumption ability parameter is generated based on the historical consumption records of the target audience. The shopping habit parameter is obtained by analyzing the shopping time, frequency, and category preference of the target audience. The commodity association parameter is obtained by analyzing the shopping basket data and calculating the association degree between commodities. The calculation formula is D = a x V + b x S + g x P, where V is the standardized moving speed value, S is the stop point distribution density value, P is the path regularity score, a, b, and g are weight coefficients and satisfy a+b+g=1; the line of sight focus parameter value F = å(Ti x Wi) / T, where Ti is the gaze time of the i-th grid, Wi is the weight coefficient of the grid, and T is the total observation time; the commodity association degree R(A, B) = P(B|A) = N(A∩B) / N(A), where N(A∩B) represents the number of times that commodities A and B appear together, and N(A) represents the total number of times that commodity A appears. The system standardizes the calculation results of each parameter to the [0, 1] interval for subsequent advertisement optimization. Among them, considering that the shopping path of the user has certain regularity and inertia, understanding this regularity helps to predict the moving track and focus of the user, thereby optimizing the advertisement placement strategy. The higher the regularity of the user, the more stable the shopping behavior, and the more controllable the expected effect of the advertisement placement. The specific calculation method of the path regularity score P is: P = a1 x Rs + a2 x Rf + a3 x Rd. Rs is the path similarity, which is obtained by calculating the edit distance between different access paths, Rs = 1-D(Pi, Pj) / max(len(Pi), len(Pj)), where D(Pi, Pj) is the edit distance of path Pi and Pj; Rf is the fixed point frequency, which is obtained by calculating the repeated appearance frequency of the key node, Rf = Nf / Nt, where Nf is the number of fixed points, and Nt is the total access times; Rd is the direction consistency, which is obtained by calculating the frequency of direction change, Rd = 1-Nc / Nt, where Nc is the number of direction changes. a1, a2, and a3 are weight coefficients and satisfy a1+a2+a3=1. This multi-dimensional path regularity scoring method can comprehensively reflect the stability characteristics of the user's shopping path.

[0088] S4122、According to the consumption ability parameter and the shopping habit parameter, the population matching parameter is obtained.

[0089] In this embodiment, the crowd matching parameter adopts a two-dimensional score structure, and the consumption capacity dimension and the shopping habit dimension are combined to form a complete crowd characteristic description. The system designs a crowd matching degree evaluation model, and generates a comprehensive parameter reflecting the user group characteristics by weighted integration of the consumption capacity parameter and the shopping habit parameter.

[0090] Specifically, the system first classifies and quantizes the consumption capacity parameter to establish a consumption capacity score standard, then extracts features from the shopping habit parameter to construct a shopping behavior score system, and finally integrates the parameters of the two dimensions into a unified crowd matching parameter through feature combination.

[0091] In one embodiment, the personalized delivery parameter includes an advertisement display time period parameter and an advertisement display form parameter, with reference to Figure 7 In step S420, the retail cabinet display position information is obtained, and personalized delivery parameters are set for different positions according to the display position information and the overall advertisement selection parameter, specifically including the following steps: S421, obtain retail cabinet passenger flow distribution information and shopping dynamic line information.

[0092] In this embodiment, the passenger flow distribution information includes crowd density distribution, residence area distribution and active time period distribution. The shopping dynamic line information includes main shopping path, secondary shopping path and node stopping point. The system collects these information through the front-end sensor network to establish a retail cabinet space behavior database.

[0093] Specifically, the system constructs a passenger flow monitoring network, arranges an infrared sensor array in the retail cabinet area, and divides the monitoring space into multiple grid units. The system records the passenger flow changes of each grid, counts the residence time distribution, and analyzes the crowd movement trajectory. Through processing of the collected data, the system generates a standardized passenger flow distribution map and a shopping dynamic line map.

[0094] S422, according to the position optimization parameter, the passenger flow distribution information and the shopping dynamic line information, obtain the display strategy parameter for the display position.

[0095] In this embodiment, the display strategy parameter includes display priority, display time length and display frequency. The system establishes a position scoring model, combines the position optimization parameter with real-time passenger flow data, calculates the display effect score for each display position, and accordingly formulates differentiated display strategies.

[0096] Specifically, the system designs a display strategy generation rule library to map the position characteristics and passenger flow characteristics into specific display parameters. First, the basic display strategy is determined based on the position optimization parameter, and then the strategy is adjusted according to the passenger flow distribution and the shopping dynamic line. The system sets differentiated display priorities for different display positions, and configures corresponding display time length and frequency parameters according to the position characteristics.

[0097] S423、According to the crowd matching parameter, the commodity association parameter, the passenger flow distribution information and the shopping trajectory information, advertisement display time period parameters and advertisement display form parameters are obtained for the display position.

[0098] In this embodiment, the advertisement display time period parameters include main display time period, secondary display time period and display time period weight. The advertisement display form parameters include display size, display special effect and interaction mode. The system customizes personalized display scheme for each display position through multi-dimensional parameter combination.

[0099] Specifically, the system establishes a display parameter configuration rule library. The configuration of the display time period parameters is based on the passenger flow distribution characteristics. The system analyzes the active time period of the target crowd and determines the best display time in combination with the commodity association. The configuration of the display form parameters considers the position characteristics and the crowd characteristics. The system selects appropriate display size and interaction mode according to the line of sight accessibility of the display position and the surrounding environment.

[0100] It should be understood that the size of the serial number of each step in the above embodiment does not mean the order of execution. The execution order of each process should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0101] In a second aspect, the present application provides a retail cabinet-based advertisement delivery system. The retail cabinet-based advertisement delivery system of the present application is described below in combination with the above retail cabinet-based advertisement delivery method.

[0102] Reference Figure 8 A retail cabinet-based advertisement delivery system, comprising: A data acquisition module for acquiring user purchase history, stay duration and interest preference information to obtain user behavior data; A feature analysis module for analyzing user feature data by using a deep learning algorithm according to the user behavior data; A user portrait module for constructing a target user portrait based on the user feature data; An advertisement matching module for matching advertisement resources in an advertisement content library according to the target user portrait to determine advertisement content to be delivered; A delivery execution module for executing an advertisement delivery operation based on the advertisement content to be delivered to obtain advertisement delivery effect data; A strategy adjustment module for dynamically adjusting an advertisement delivery strategy according to the advertisement delivery effect data.

[0103] In one embodiment, the present application provides an electronic device, which can be a server, and its internal structure diagram can be as shown in Figure 9As shown in the figure. The electronic device includes a processor, a memory and a network interface connected through a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium, an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The database of the electronic device is used to store data. The network interface of the electronic device is used to communicate with the external terminal through the network connection. The computer program is executed by the processor to implement a retail cabinet-based advertisement delivery method.

[0104] Those skilled in the art can understand that, Figure 9 The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. The specific electronic device can include more or fewer components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0105] In one embodiment, an electronic device is also provided, including a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps in the above method embodiments.

[0106] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The above-mentioned computer program can be stored in a non-volatile computer readable storage medium. When the computer program is executed, it can include the processes of the above-mentioned embodiments. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory or optical storage, etc. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not as a limitation, RAM can be in various forms such as static random access memory (SRAM) or dynamic random access memory (DRAM), etc.

[0107] The above are the preferred embodiments of the present application, which do not limit the protection scope of the present application, therefore: any equivalent changes made on the structure, shape, principle of the present application shall be covered within the protection scope of the present application.

Claims

1. A method for placing advertisements based on retail counters, characterized in that: The steps include: Collect user purchase history, length of stay, and interest preference information to obtain user behavior data; Analyze the user behavior data using a deep learning algorithm to obtain user feature data; Building a target user profile based on the user feature data; According to the target user profile, the advertising resources in the advertising content library are matched to determine the advertising content to be delivered; Based on the advertisement content to be delivered, executing an advertisement delivery operation to obtain advertisement delivery effect data; Dynamically adjust the advertising delivery strategy based on the advertising delivery effect data.

2. The method for placing advertisements based on retail counters according to claim 1, characterized in that: Collecting user purchase history, stay duration, and interest preference information to obtain user behavior data includes the following steps: Extracting commodity category data and consumption amount data from the purchase history; Determine the user's dwell time data in each merchandise area based on the user's dwell time information and the relative positions of each merchandise area in the retail cabinet layout information; Determine the type and degree of user interest preferences based on the purchase history and stay duration information; The commodity category data, the consumption amount data, the residence time data and the interest preference data are associated to obtain the user behavior data.

3. The method for placing advertisements based on retail counters according to claim 2, characterized in that: Based on the user behavior data, a deep learning algorithm is used to analyze and obtain user feature data, which specifically includes the following steps: Encoding the commodity category data to obtain commodity category features; Normalizing the consumption amount data to obtain consumption capacity characteristics; Dividing the residence time data into time periods to obtain time features; Performing quantitative calculation on the interest preference data to obtain preference characteristics; The commodity category characteristics, consumption capacity characteristics, time characteristics and preference characteristics are combined to obtain user characteristic data.

4. The method for placing advertisements based on retail counters according to claim 1, characterized in that: According to the target user profile, the advertising resources in the advertising content library are matched and the advertising content to be delivered is determined, which specifically includes the following steps: Obtaining a target audience group and an expected conversion rate based on the target user portrait, and obtaining overall advertisement selection parameters based on the target audience group and the expected conversion rate; Obtaining retail counter display location information, and setting personalized delivery parameters for different locations based on the display location information and the overall advertisement selection parameters; Screening advertisement content according to the personalized delivery parameters and the overall advertisement selection parameters; Estimating the effect of the screening results to obtain estimated conversion parameters; comparing the estimated conversion parameters with the expected conversion rate to obtain an expected difference value; If the expected difference value exceeds a preset threshold, advertising optimization is triggered; and parameter adjustment suggestions are obtained based on the advertising optimization results; The personalized delivery parameters are updated according to the parameter adjustment suggestions.

5. The method for placing advertisements based on retail counters according to claim 4, characterized in that: Obtaining a target audience group and an expected conversion rate, and obtaining overall advertisement selection parameters based on the target audience group and the expected conversion rate, specifically includes the following steps: Extracting a target residence time from the expected conversion rate and using the target residence time as a main evaluation indicator; According to the target stay time and the target audience group, obtain population matching parameters, product association parameters and location optimization parameters; The overall advertisement selection parameters are obtained by associating the target residence time, the crowd matching parameters, the product association parameters and the location optimization parameters.

6. The method for placing advertisements based on retail counters according to claim 5, characterized in that: The location optimization parameters include movement line parameters and sight focus parameters, and the crowd matching parameters include spending power parameters and shopping habit parameters. According to the target stay time and the target audience group, the crowd matching parameters, product association parameters and location optimization parameters are obtained, which specifically includes the following steps: According to the target stay time and the target audience group, the movement line parameter, the sight focus parameter, the spending power parameter, the shopping habit parameter and the product association parameter are obtained; The population matching parameter is obtained according to the consumption capacity parameter and the shopping habit parameter.

7. The method for placing advertisements based on retail counters according to claim 5, characterized in that: The personalized delivery parameters include advertisement display period parameters and advertisement display format parameters. The retail counter display location information is obtained, and personalized delivery parameters are set for different locations based on the display location information and the overall advertisement selection parameters. Specifically, the steps include: Obtain customer flow distribution information and shopping route information at retail counters; Obtaining display strategy parameters for the display location based on the location optimization parameters, the customer flow distribution information, and the shopping route information; The advertisement display period parameter and the advertisement display format parameter are obtained for the display location according to the crowd matching parameter, the commodity association parameter, the customer flow distribution information and the shopping route information.

8. An advertising system based on a retail counter, characterized in that: include: The data collection module is used to collect users' purchase history, stay time and interest preference information to obtain user behavior data; A feature analysis module is used to analyze the user behavior data using a deep learning algorithm to obtain user feature data; A user portrait module, used to construct a target user portrait based on the user feature data; An advertisement matching module is used to match advertisement resources in the advertisement content library according to the target user profile and determine the advertisement content to be delivered; A delivery execution module, configured to execute an advertisement delivery operation based on the advertisement content to be delivered, and obtain advertisement delivery effect data; The strategy adjustment module is used to dynamically adjust the advertising delivery strategy according to the advertising delivery effect data.

9. An electronic device, characterized in that: The invention comprises a memory, a processor and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the steps of the retail cabinet-based advertising delivery method described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the retail cabinet-based advertising delivery method according to any one of claims 1 to 7 are implemented.

Images