Advertisement online pushing system and method, computer equipment and storage medium

Abstract

The invention provides an advertisement online pushing system and method, computer equipment and a storage medium. The method is applied to an online advertisement pushing system comprising a front-end equipment module, user feature data are obtained through front-end equipment, the user feature data are transmitted, target advertisement content is displayed, and the obtained user feature data can be stored and called through a data bin module. And the target advertisement content is obtained by filtering and processing the user feature data through a data processing module and an information publishing module at the rear end. Therefore, through the front-end equipment module, the functions of online user feature data acquisition and real-time pushing are realized, and the pushed content is obtained through automatic analysis according to the real-time features of the user, so that the timeliness and effectiveness of advertisement pushing are improved, and meanwhile, the advertisement pushing efficiency can be improved according to the track and behavior mode of the whole action process of the user. A data basis is provided for training of an advertisement pushing related model, data analysis and feedback are realized, and the effectiveness of advertisement pushing is further improved.

Description

Technical Field

[0001] The present application relates to the field of information push, and in particular to an online advertisement push system, method, computer device and storage medium. Background Art

[0002] In some industries, there may be application scenarios involving information delivery. For example, in the retail industry, it is necessary to improve the efficiency and effectiveness of advertising delivery to increase sales and provide a basis for subsequent sales decisions.

[0003] However, the current advertising push method usually generates a corresponding advertising push strategy based on the user distribution data and consumption habit data of the target area within a certain period of time after offline analysis. Therefore, due to the low analysis efficiency of this type of advertising push method, the timeliness of the push results is not high and the push effect is not good. Therefore, an online advertising push method is needed to improve the timeliness and effectiveness of advertising push. Summary of the invention

[0004] The purpose of this application is to solve at least one of the above-mentioned technical defects, especially the technical defects of insufficient timeliness and effectiveness of advertising push in the prior art.

[0005] In a first aspect, the present application provides an online advertisement push system, the system comprising:

[0006] Front-end equipment module, data warehouse module, data processing module and information publishing module;

[0007] Wherein, the front-end device module includes at least one image acquisition unit and at least one display unit, which are used to acquire user characteristic data, transmit the user characteristic data, and display target advertisement content generated according to the user characteristic data;

[0008] The data warehouse module is used to receive the user characteristic data;

[0009] The data processing module is used to obtain recommended content label data through a collaborative filtering algorithm according to the user feature data;

[0010] The information publishing module is used to determine the target advertisement content according to the recommended content tag data, and transmit the target advertisement content to the front-end device module for display.

[0011] As an optional implementation, the method is applied to a front-end device module, the front-end device module includes at least one image acquisition module and at least one display module, and the method includes:

[0012] Acquire user feature data of the target user through the image acquisition module;

[0013] Transmitting the user characteristic data to a data warehouse module;

[0014] Receive target advertisement content from the information publishing module, and display the target advertisement content on the display module.

[0015] As an optional implementation manner, obtaining user feature data of the target user through the image acquisition module includes:

[0016] Acquire the user image of the target user within a preset time period through the image acquisition module;

[0017] Extracting image features from the user image;

[0018] The body feature data, clothing feature data and behavior feature data of the target user are determined according to the image features.

[0019] As an optional implementation manner, determining the physical feature data, clothing feature data and behavior feature data of the target user according to the image features includes:

[0020] Determining the gender attribution probability, age regression value, and emotional state classification of the target user according to the image features;

[0021] Determine one or more of the clothing type, clothing style, clothing color and clothing brand classification of the target user through a pre-trained image classification model according to the image features;

[0022] According to the image features, the user actions of the target user are analyzed by the OpenPose algorithm to determine the interest point classification and behavior pattern classification of the target user.

[0023] As an optional implementation manner, the transmitting the user characteristic data to a data warehouse module includes:

[0024] Determining first data in the user characteristic data;

[0025] Performing an encryption operation on the first data to obtain second data, and transmitting the second data to a data warehouse module;

[0026] The encryption operation includes:

[0027] Generate a target key of a preset length;

[0028] Performing data preprocessing and grouping processing on the first data to obtain third data;

[0029] According to the target key, an encryption operation is performed on the third data to obtain the second data.

[0030] As an optional implementation manner, the front-end device module includes a first device, a second device and a third device, and the method further includes:

[0031] Acquire first dynamic data through the image acquisition module of the first device;

[0032] Acquire second dynamic data through the image acquisition module of the second device;

[0033] Determine the track data of the target user according to the first dynamic data and the second dynamic data, and determine the third device according to the track data;

[0034] Transmitting the first dynamic data, the second dynamic data and the trajectory data to a data warehouse module;

[0035] The dynamic advertisement content is received from the information publishing module through the third device, and the dynamic advertisement content is displayed on the display module of the third device.

[0036] As an optional implementation, the method is applied to a data processing module, and the method includes:

[0037] Obtain user feature data in the data warehouse module;

[0038] According to the user feature data, obtaining recommended content tag data through a collaborative filtering algorithm;

[0039] The content tag data is transmitted to the information publishing module.

[0040] As an optional implementation manner, obtaining the recommended content tag data through a collaborative filtering algorithm according to the user feature data includes:

[0041] Performing data cleaning and data segmentation on the user characteristic data;

[0042] The recommendation model is trained by a collaborative filtering algorithm until the target performance of the recommendation model meets the preset conditions;

[0043] The recommendation model generates recommended content label data according to the user feature data.

[0044] In a third aspect, the present application provides a computer device comprising one or more processors and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the method described in the second aspect are performed.

[0045] In a fourth aspect, the present application provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, causes the one or more processors to perform the steps of the method described in the second aspect.

[0046] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0047] Based on any of the above embodiments, the method provided by the present application is applied to an online advertising push system including a front-end device module, and the user characteristic data is obtained, transmitted and the target advertising content is displayed through the front-end device. The obtained user characteristic data can be saved and called through the data warehouse module, and the target advertising content is obtained after filtering and processing the user characteristic data through the back-end data processing module and the information publishing module. The user characteristic data can include characteristic data in aspects such as body characteristic data, clothing characteristic data, behavior characteristic data and environmental context data. Thus, through the front-end device module including at least one image acquisition module and at least one display module, the functions of online acquisition of user characteristic data and real-time push are realized, and the pushed content is obtained by automatic analysis based on the real-time characteristics of the user, which improves the timeliness and effectiveness of advertising push. At the same time, according to the trajectory and behavior pattern of the whole process of user actions, a data basis can be provided for the training of the relevant model of advertising push, and data analysis and feedback can be realized, which further improves the effectiveness of advertising push. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0049] Figure 1 A schematic diagram of the structure of an online advertisement push system provided by an embodiment of the present application;

[0050] Figure 2 A flowchart of an online advertisement push method provided by an embodiment of the present application;

[0051] Figure 3 A flowchart of an online advertisement push method provided by an embodiment of the present application;

[0052] Figure 4 This is a diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0054] In the retail sector, for application scenarios involving advertising, current advertising and push methods are mostly static, with offline adjustments made based on customer distribution and store operations within a certain period of time. The shortcomings of the current advertising delivery process mainly include the inability to analyze user portraits beforehand, the inability to track the entire shopping process of users during the process, and the inability to analyze users' past shopping data afterwards, which makes it difficult to evaluate the effectiveness of advertising. There is a lack of intelligent analysis of real-time data, making it difficult to measure the response of users to advertising, as well as the data value brought to retailers or advertisers, and the inability to accurately deliver advertisements.

[0055] This application adds the corresponding function of online data analysis on the basis of traditional advertising. After structural analysis of the user's human characteristics, the advertisement is automatically delivered. At the same time, the data warehouse can be used as the basis for advertising push and sales data analysis. Retailers can observe and analyze the effect of their operations through data perspective.

[0056] In general, the technical concept of the present application is that the method provided by the present application is applied to an online advertising push system including a front-end device module, and the user characteristic data is obtained, transmitted and the target advertising content is displayed through the front-end device. The obtained user characteristic data can be stored and called through the data warehouse module, and the target advertising content is obtained after filtering and processing the user characteristic data through the back-end data processing module and the information publishing module. The user characteristic data can include characteristic data in terms of body characteristic data, clothing characteristic data, behavior characteristic data and environmental context data. Therefore, through the front-end device module including at least one image acquisition module and at least one display module, the functions of online acquisition of user characteristic data and real-time push are realized, and the pushed content is obtained by automatic analysis based on the real-time characteristics of the user, which improves the timeliness and effectiveness of advertising push. At the same time, according to the trajectory and behavior pattern of the whole process of user actions, it can also provide a data basis for the training of advertising push related models, realize data analysis and feedback, and further improve the effectiveness of advertising push.

[0057] See also Figure 1 , Figure 1 A schematic diagram of the structure of an online advertising push system provided by an embodiment of the present application is shown in FIG. Figure 1 As shown, the system structure includes:

[0058] Front-end equipment module, data warehouse module, data processing module and information publishing module;

[0059] Wherein, the front-end device module includes at least one image acquisition unit and at least one display unit, which are used to acquire user characteristic data, transmit the user characteristic data, and display target advertisement content generated according to the user characteristic data;

[0060] The data warehouse module is used to receive the user characteristic data;

[0061] In the advertising push task, the data warehouse module is used to store user feature data in the form of metadata and provide it to the data processing module. In some feasible application scenarios, the data in the data warehouse module can also be used for sales forecasting or sales model optimization tasks of related merchants or stores.

[0062] The data processing module is used to obtain recommended content label data through a collaborative filtering algorithm according to the user feature data;

[0063] The information publishing module is used to determine the target advertisement content according to the recommended content tag data, and transmit the target advertisement content to the front-end device module for display.

[0064] In a specific application scenario, a corresponding description will be given for the coordination and cooperation of each module. For details, please refer to the relevant implementation methods.

[0065] This embodiment provides an online advertising push system, which realizes real-time data acquisition, processing and coordinated cooperation through the front-end device module, data warehouse module, data processing module and information release module, obtains user feature data, transmits user feature data and displays target advertising content through the front-end device, and the obtained user feature data can be saved and called through the data warehouse module, and the user feature data is filtered and processed through the back-end data processing module and information release module to obtain the target advertising content. In this way, the functions of online acquisition of user feature data and real-time push are realized, and the pushed content is obtained by automatic analysis based on the real-time characteristics of the user, which improves the timeliness and effectiveness of advertising push. At the same time, it can also provide a data basis for the training of advertising push related models based on the trajectory and behavior pattern of the entire process of user actions, realize data analysis and feedback, and further improve the effectiveness of advertising push.

[0066] See also Figure 2 , Figure 2 A schematic diagram of a flow chart of an online advertising push method provided by an embodiment of the present application, wherein the method is applied to a front-end device module, wherein the front-end device module includes at least one image acquisition module and at least one display module, such as Figure 2 As shown, the method includes:

[0067] S101, acquiring user feature data of a target user through the image acquisition module;

[0068] User feature data mainly includes data obtained based on image feature analysis and used to indicate or profile users. When transmitted to the data warehouse module, it can usually be transmitted or saved in the form of metadata rather than the data itself, thereby saving communication and storage resources.

[0069] S102, transmitting the user characteristic data to a data warehouse module;

[0070] The data in the data warehouse module can be used in the advertising push scenario described in this application, and can also be used by back-end retailers to call corresponding data for data perspective, analyze operational effects, and perform optimization and other tasks.

[0071] The metadata stored in the data warehouse module may include the following data. In this application, user feature data is mainly used for advertising push.

[0072] 1. User-related metadata:

[0073] User basic information metadata:

[0074] User ID: The unique identifier of each user in the system, such as "U001".

[0075] Name: For example, "Zhang San".

[0076] Gender: "Male" or "Female".

[0077] Age: expressed as a number, such as "28" years old.

[0078] User portrait metadata can be determined based on the data acquired in real time by the image acquisition module, or based on historical purchase records or activity tracks:

[0079] Interest tag: It is a multi-valued attribute, which may include "fashion", "technology", "sports", etc., indicating the user's area of ​​interest.

[0080] Consumption level: divided into "high", "medium" and "low", based on the user's past consumption amount and consumption categories, for example "medium".

[0081] Activity area: Indicated by the area division within the shopping mall, such as "fashion area on the first floor" and "digital area on the third floor", etc., obtained through Wi-Fi positioning or consumption record analysis.

[0082] User behavior metadata:

[0083] Historical ad viewing records: A list of ad content IDs that the user has viewed, such as [“A001 - Men’s Fashion Advertisement”, “A005 - Smartphone Advertisement”].

[0084] Ad viewing time: The length of time each ad is viewed, measured in seconds, such as "15" seconds (corresponding to a certain ad).

[0085] Purchase record association: associated with the user's purchase record, including the name, brand, price, time and other information of the purchased product, such as:

[0086] [{"Product Name": "A Sports Shoes";

[0087] "Brand": "A";

[0088] "price": "800";

[0089] "Purchase time": "2024-10-15";}]

[0090] 2. Metadata related to advertising materials

[0091] Basic advertising information metadata:

[0092] Ad content ID: such as "A010 - High-end beauty advertisement", is the unique identification code of the advertisement in the system.

[0093] Advertisement name: "BBB lipstick new product promotion".

[0094] Ad type: "Fashion and Beauty".

[0095] Delivery time range: from "2024 - 11 - 01" to "2024 - 11 - 30".

[0096] Brand: "B".

[0097] Ad content feature metadata:

[0098] Advertising style labels: such as "luxurious and grand", "fresh and natural", etc., describing the overall style of the advertisement, which can be "luxurious and grand".

[0099] Mainly display product features: including product color, style, function, etc., for example "Lipstick color: bright red; texture: velvet; efficacy: moisturizing, high color rendering."

[0100] Advertising copy keywords: such as "luxury", "durability", "charm", etc., are extracted from the advertising copy to facilitate search and matching.

[0101] Advertising image or video features: If it is an image ad, it may include features such as image resolution, color distribution, and subject object; if it is a video ad, it may include features such as video length, frame rate, and picture style, such as "Image resolution: 1920×1080; Color distribution: red and gold as the main colors; Subject object: close-up of lipstick."

[0102] 3. Environment-related metadata:

[0103] Ad device location metadata:

[0104] Advertising machine number: such as "D005", which uniquely identifies each advertising machine.

[0105] Location: "Next to the elevator entrance on the second floor of the shopping mall."

[0106] Longitude and latitude coordinates: If the mall has precise geographic positioning requirements, such as "longitude: 120.5678, latitude: 30.1234".

[0107] Environmental Condition Metadata:

[0108] Light intensity: measured in Lux, for example, "300 Lux", which indicates the brightness of the light at the location of the advertising machine, and is collected by the light sensor on the advertising machine.

[0109] Ambient temperature: such as "25℃", collected by a temperature sensor, can be used for certain special advertisements (such as considering the impact of ambient temperature on products in refrigerated food advertisements) or equipment operating status monitoring.

[0110] Time information: includes the current year, month, day, week, hour, minute, and second, such as "2024 - 11 - 15, Wednesday, 15:30:00", which is used to push different advertisements according to different times.

[0111] S103: Receive target advertisement content from the information publishing module, and display the target advertisement content on the display module.

[0112] The target advertisement content can be pre-read through the information publishing module and directly called after the required target advertisement content is determined. For details, please refer to the implementation method related to the recommended content tag data.

[0113] The method provided in this embodiment is applied to an online advertising push system including a front-end device module, and the user characteristic data is obtained, transmitted and the target advertising content is displayed through the front-end device. The obtained user characteristic data can be saved and called through the data warehouse module, and the target advertising content is obtained after filtering and processing the user characteristic data through the back-end data processing module and information publishing module. Thus, the functions of online acquisition of user characteristic data and real-time push are realized through the front-end device module including at least one image acquisition module and at least one display module, and the pushed content is obtained by automatic analysis based on the real-time characteristics of the user, which improves the timeliness and effectiveness of advertising push. At the same time, it can also provide a data basis for the training of advertising push related models based on the trajectory and behavior pattern of the entire process of user actions, realize data analysis and feedback, and further improve the effectiveness of advertising push.

[0114] As an optional implementation manner, obtaining user feature data of the target user through the image acquisition module includes:

[0115] Acquire the user image of the target user within a preset time period through the image acquisition module;

[0116] Extracting image features from the user image;

[0117] The body feature data, clothing feature data and behavior feature data of the target user are determined according to the image features.

[0118] This embodiment can extract image features from user images by acquiring a combination of user images of the target user within a preset time period, thereby determining the target user's physical feature data, clothing feature data, and behavioral feature data, and can further obtain environmental data around the target user. The target user can be a single user or a combination of multiple users, thereby providing a data basis in the form of feature values ​​for subsequent advertising push, thereby improving the effectiveness of advertising push.

[0119] As an optional implementation manner, determining the physical feature data, clothing feature data and behavior feature data of the target user according to the image features includes:

[0120] Determining the gender attribution probability, age regression value, and emotional state classification of the target user according to the image features;

[0121] Determine one or more of the clothing type, clothing style, clothing color and clothing brand classification of the target user through a pre-trained image classification model according to the image features;

[0122] According to the image features, the user actions of the target user are analyzed by the OpenPose algorithm to determine the interest point classification and behavior pattern classification of the target user.

[0123] This implementation provides a specific method for extracting and analyzing various aspects of the target user through image features. The characteristics of the target user include physical feature data, clothing feature data and behavioral feature data. Specifically, the user's gender is determined by a classification algorithm, the user's age is predicted by a regression model, and the user's emotional state is identified by a sentiment analysis model, thereby completing the process of determining the target user's physical feature data. The target user's clothing can be determined through a pre-trained image classification model. The target user's user actions and corresponding tendencies can be analyzed through the OpenPose algorithm to identify the user's interests and behavior patterns, thereby completing the accurate identification of various aspects of the target user's characteristics and improving the effectiveness of advertising push. The corresponding data can also be used for the optimization of the advertising push process or the distribution of stores and products, thereby improving the practicality of the corresponding image data.

[0124] As an optional implementation manner, the transmitting the user characteristic data to a data warehouse module includes:

[0125] Determining first data in the user characteristic data;

[0126] Performing an encryption operation on the first data to obtain second data, and transmitting the second data to a data warehouse module;

[0127] The encryption operation includes:

[0128] Generate a target key of a preset length;

[0129] Performing data preprocessing and grouping processing on the first data to obtain third data;

[0130] According to the target key, an encryption operation is performed on the third data to obtain the second data.

[0131] In the advertising push system, a large amount of sensitive data related to users is involved, such as user personal information, user behavior data, and image data collected by AI passenger flow cameras that may involve user privacy. If these data are not encrypted during the process of being transmitted and aggregated from the AI ​​passenger flow camera to the data warehouse, user privacy may be leaked, causing unnecessary trouble to users and also damaging the security of the advertising push service.

[0132] Specifically, the encryption operation may generally include AES or SM4, and its principle is summarized as follows:

[0133] AES (Advanced Encryption Standard) is a symmetric encryption algorithm. Symmetric encryption means that the same key is used for encryption and decryption. The AES algorithm converts plaintext data into ciphertext data based on a specific key through a series of complex mathematical transformation operations. At the receiving end, the ciphertext can be restored to the original plaintext data using the same key.

[0134] The AES algorithm supports different key lengths, the most common of which are 128 bits, 192 bits, and 264 bits. The longer the key length, the higher the encryption security is generally, but the computational complexity of encryption and decryption will also increase accordingly.

[0135] Encryption process example:

[0136] First, you need to generate a key of appropriate length, such as a 128-bit key. This key can be generated by a random number generator and its confidentiality must be ensured so that only the authorized encryption and decryption parties know the key.

[0137] Then, the plaintext data is grouped and processed according to the specifications of the AES algorithm, usually dividing the data into blocks of fixed length (such as 128 bits per block). A series of round-trip operations are performed on each block of data, including byte substitution, row shift, column confusion and other steps, through which each block of plaintext data is converted into ciphertext data.

[0138] After processing all the grouped data, the ciphertext data corresponding to the entire plaintext data is finally obtained. For example, a seemingly irregular sequence of hexadecimal characters may be obtained, such as "5A8B3C1D2E9F4A0B6C7D8E9F0A1B2C3D", which is the result after encryption.

[0139] And, SM4 has a similar principle, based on the principle of symmetric key, that is, encryption and decryption use the same key. The SM4 algorithm transforms the plaintext through specific mathematical operations to generate ciphertext.

[0140] The key length of the SM4 algorithm is fixed at 128 bits, and it also has high performance in ensuring data security.

[0141] Likewise, an example of the encryption process:

[0142] First, generate a 128-bit key, which needs to be kept strictly confidential.

[0143] Then, according to the requirements of the SM4 algorithm, the plaintext data is preprocessed, which may involve operations such as data padding and grouping to make it conform to the input format of the algorithm.

[0144] After that, each group of data is encrypted, and each group of plaintext is converted into ciphertext through a series of specific functions and operation steps. For example, after encryption by the SM4 algorithm, another form of ciphertext data may be obtained, such as "3E5F6G7H8I9J0K1L2M3N4O5P6Q7R8S" (this is just an illustrative example, the actual ciphertext will depend on the specific data and algorithm operation results).

[0145] Encryption timing selection:

[0146] In the advertising push system, when the AI ​​passenger flow camera collects sensitive data, it will be encrypted immediately when the data is transmitted and aggregated to the data warehouse through the IoT communication technology protocol. This ensures that the data is protected at the first moment of transmission, reducing the risk of data leakage.

[0147] Key Management:

[0148] The keys used by encryption algorithms such as AES or SM4 need to be strictly managed. The generation of keys should follow the principle of security, such as using a professional random number generator to generate high-quality keys. For the storage and transmission of keys, such as in an encrypted key library, only authorized users can access it. During the operation of the entire system, the keys need to be updated regularly to further improve data security.

[0149] Decryption process:

[0150] After the data is transferred to the data warehouse, the encrypted data needs to be decrypted for subsequent data analysis and processing. On the data warehouse side, the correct key is also required, and decryption is performed according to the inverse operation process of the corresponding encryption algorithm. For example, for AES encrypted data, the same key as that used for encryption is used to gradually restore the ciphertext to the original plaintext data according to the decryption steps of the AES algorithm; for SM4 encrypted data, the same key is also used to restore the ciphertext to plaintext according to the decryption process of the SM4 algorithm. Only the original data after decryption can be effectively processed on the data analysis platform, such as data cleaning, model training and other operations.

[0151] This embodiment processes the data to be encrypted in the user feature data according to a preset encryption algorithm, such as AES or SM4, during the data aggregation process, determines the target key, performs encryption operations according to the target key after preprocessing the data to be encrypted, obtains encrypted second data, and transmits the encrypted data and the corresponding key to the data warehouse module. When applied, the permissions for related data are only open to users with corresponding permissions, thereby improving the security of the user feature data.

[0152] As an optional implementation manner, the front-end device module includes a first device, a second device and a third device, and the method further includes:

[0153] Acquire first dynamic data through the image acquisition module of the first device;

[0154] Acquire second dynamic data through the image acquisition module of the second device;

[0155] Determine the track data of the target user according to the first dynamic data and the second dynamic data, and determine the third device according to the track data;

[0156] Transmitting the first dynamic data, the second dynamic data and the trajectory data to a data warehouse module;

[0157] The dynamic advertisement content is received from the information publishing module through the third device, and the dynamic advertisement content is displayed on the display module of the third device.

[0158] It should be noted that the first device and the second device can be used to determine the user trajectory in this embodiment, but it does not mean that there are only two front-end device modules, and the third device determined here can also be the first device or the second device itself.

[0159] This embodiment is used in scenarios with multiple front-end device modules. Through the image acquisition modules corresponding to the multiple front-end device modules, dynamic data of the same target user at different locations can be obtained, and then the user trajectory and subsequent location tendencies can be determined, thereby determining a third device, and playing the determined dynamic advertising content on the display module of the third device, that is, playing the corresponding content at the location where the target user is about to appear, thereby improving the effectiveness of advertising push.

[0160] As an optional implementation, the method is applied to a data processing module, and the method includes:

[0161] Obtain user feature data in the data warehouse module;

[0162] According to the user feature data, obtaining recommended content tag data through a collaborative filtering algorithm;

[0163] The content tag data is transmitted to the information publishing module.

[0164] This embodiment obtains user feature data stored in the data warehouse through the data processing module to realize data analysis or advertising push functions. In the scenario of advertising push, the recommendation model can be trained through the collaborative filtering algorithm, and the recommended content label data can be obtained in combination with the real-time user feature data, and further transmitted to the information release module. The information release module matches the preset recommended content through the recommended content label data to realize the determination of the advertising content, thereby improving the timeliness and effectiveness of advertising push.

[0165] As an optional implementation manner, obtaining the recommended content tag data through a collaborative filtering algorithm according to the user feature data includes:

[0166] Performing data cleaning and data segmentation on the user characteristic data;

[0167] The recommendation model is trained by a collaborative filtering algorithm until the target performance of the recommendation model meets the preset conditions;

[0168] The recommendation model generates recommended content label data according to the user feature data.

[0169] The following describes a feasible application scenario of collaborative filtering algorithm:

[0170] User-based collaborative filtering:

[0171] The core idea is that if user A and user B have similar preferences for some items (for example, they both like certain advertising content), then the advertisements that user A likes but user B has not yet seen are more likely to be recommended to user B. It determines similar user groups by calculating the similarity between users. Commonly used similarity calculation methods include cosine similarity and Pearson correlation coefficient.

[0172] The cosine similarity between users A and B can be calculated to obtain the similarity value between A and B. Based on this similarity, a set of users similar to the target user is found, and then advertisements that the target user may like are selected from the advertisements liked by these similar users for recommendation.

[0173] Item-based collaborative filtering:

[0174] It focuses on the similarity between items. If Ad A and Ad B are liked by many of the same users, then users who like Ad A may also like Ad B. Take the above rating data as an example to calculate the similarity between ads. For example, to calculate the similarity between Ad A and Ad B, a similar similarity calculation formula is used to get the result based on the ratings of users who like these two ads. Then, when a user likes Ad A, ads similar to Ad A can be recommended to the user.

[0175] ALS (Alternating Least Squares):

[0176] ALS is an optimization algorithm for matrix decomposition, which is used in collaborative filtering to decompose the user-item rating matrix into two low-dimensional matrices (user feature matrix and item feature matrix). For example, for a user-item rating matrix, ALS decomposes it into a user feature matrix and an item feature matrix.

[0177] It minimizes the loss function by alternately fixing the user feature matrix and the item feature matrix. The loss function usually uses the mean square error (MSE). During the training process, one of the user feature matrix and the item feature matrix is ​​first randomly initialized, and then it is fixed to solve the other one. The iterations are repeated until convergence to obtain the optimized user feature matrix and item feature matrix, which can be used to predict the user's rating of unrated items and achieve recommendations.

[0178] Application in advertising push:

[0179] Data preparation:

[0180] Collect data on the interaction between users and advertisements, such as how long users watch the advertisements, whether they purchase related products or visit related stores, and other behavioral data, convert these data into data representations similar to ratings, and determine the user's interaction rating for the advertisements.

[0181] Model training:

[0182] The ALS algorithm is used to decompose and train the constructed user-ad interaction matrix. Appropriate parameters are set, such as the potential feature dimension and regularization parameter. The user feature matrix and the item feature matrix are initialized, and then training is performed according to the iterative steps of ALS. For example, the user feature matrix is ​​fixed first, the item feature matrix is ​​solved, and the update value is calculated by minimizing the loss function; then the item feature matrix is ​​fixed, and the user feature matrix is ​​solved, and this process is repeated until the loss function converges or the set number of iterations is reached.

[0183] Model evaluation and optimization:

[0184] Use indicators such as root mean square error (RMSE) and mean absolute error (MAE) to evaluate the performance of the trained model. If the evaluation indicators are not ideal, you can adjust the parameters of ALS, increase the amount of training data, and perform further preprocessing on the data (such as normalization and outlier processing) to optimize the model until the model performance reaches a satisfactory effect.

[0185] Recommended generation:

[0186] The trained model can be used to generate personalized recommendations for each user. For the target user, the predicted rating of each unexposed advertisement is obtained by calculating the feature vector in the user feature matrix and the item feature matrix. The advertisements with the highest ratings are selected as the expected recommended content, and the recommended content label data is generated for subsequent advertising push.

[0187] In the process of obtaining recommended content label data by the data processing module of this embodiment, firstly, data cleaning is performed on the imported user feature data in the form of metadata to delete deviating values ​​and fill missing values, and the data is converted into data in a preset form for easy processing. The data is further divided into a training set and a test set. The recommendation model is trained through a collaborative filtering algorithm until convergence or the model indicators meet the requirements. The trained recommendation model is used to generate recommended content label data based on the user feature data, thereby improving the matching degree between the recommended content and the target users and improving the effectiveness of advertising push.

[0188] See also Figure 3 , Figure 3 A flowchart of an online advertisement push method provided by an embodiment of the present application is shown in FIG. Figure 3 As shown, in this implementation scenario, the information release platform is the aforementioned information release module, the data analysis platform is the aforementioned data processing module, the data warehouse is the aforementioned data warehouse module, and the advertising equipment (AI camera + digital signage) is the aforementioned front-end equipment module, wherein the AI ​​camera is the specific implementation method of the image acquisition unit, and the digital signage is the specific implementation method of the display unit.

[0189] When various embodiments are combined and applied, the method may include:

[0190] First, the information publishing platform prepares the content:

[0191] Preset corresponding material content groups on the information publishing platform.

[0192] The content is grouped according to the user portrait and pre-sent to the advertising device (AI camera + digital signage). After the recommended content label data is confirmed, the corresponding material content is determined as the recommended content and transmitted to the corresponding front-end device module for playback.

[0193] Data collection layer:

[0194] An embedded AI passenger flow camera is integrated into the advertising display screen. When the advertising screen plays an advertisement, when a user passes by or stops to watch the advertisement, this AI passenger flow camera has its own human body collection and recognition algorithm to collect data and conduct simple analysis on users within the current camera's field of view based on human body characteristics.

[0195] Among them, human feature recognition mainly includes the aforementioned body feature data, clothing feature data and behavior feature data, namely:

[0196] Physical characteristics data:

[0197] Gender recognition: Use classification algorithms to determine user gender; Age estimation: Apply regression models to predict user age; Sentiment analysis: Identify the user's emotional state through sentiment analysis models.

[0198] Behavioral characteristic data:

[0199] Use the OpenPose algorithm to analyze user movements and identify points of interest and behavior patterns.

[0200] Wearing characteristic data:

[0201] Apply image classification models to analyze users’ dressing styles (such as clothing type, color, wearing, wearing, and brand logo).

[0202] Data transmission:

[0203] After completing human feature recognition, the data collected by the AI ​​passenger flow camera is aggregated into the data warehouse through the Internet of Things communication technology protocol.

[0204] During the transmission process, sensitive data is encrypted using AES or SM4 to ensure data security. For specific encryption steps, please refer to the relevant descriptions of other implementation methods.

[0205] Data analysis platform processing:

[0206] Data import: The data analysis platform pulls metadata from the data warehouse.

[0207] Data cleaning:

[0208] Handle missing values ​​(delete or fill), and perform feature transformation to a data format suitable for subsequent recommendation models.

[0209] Dataset partitioning: divide the dataset into training set and test set.

[0210] Model training:

[0211] Use collaborative filtering algorithms (such as ALS) to train recommendation models.

[0212] Model evaluation: Use metrics such as RMSE to evaluate model performance.

[0213] Generate recommended content: Generate personalized recommended content label data for each user based on the model results.

[0214] After the recommended content is generated, the generated recommended content tag data is sent to the information publishing platform. The information publishing platform determines the corresponding material content as the recommended content and transmits it to the corresponding front-end device module for playback.

[0215] Thus, the complete online advertising push process is completed.

[0216] The embodiment of the present application also provides an online advertising push device, which completes the relevant steps of data acquisition and indicator determination in the method through a determination module, and completes the relevant steps of data processing in the method through a processing module. The specific implementation method and technical effects are based on the aforementioned method and will not be repeated here.

[0217] It should be noted that it should be understood that the division of the various modules of the above device is only a division of logical functions. In actual implementation, they can be fully or partially integrated into one physical entity, or they can be physically separated. And these modules can all be implemented in the form of software called by processing elements; they can also be all implemented in the form of hardware; some modules can also be implemented in the form of software called by processing elements, and some modules can be implemented in the form of hardware. For example, the processing module can be a separately established processing element, or it can be integrated in a chip of the above device. In addition, it can also be stored in the memory of the above device in the form of program code, and called and executed by a processing element of the above device. The function of the above-determined module. The implementation of other modules is similar. In addition, these modules can be fully or partially integrated together, or they can be implemented independently. The processing element here can be an integrated circuit with signal processing capabilities. In the implementation process, each step of the above method or each module above can be completed by an integrated logic circuit of hardware in the processor element or instructions in the form of software.

[0218] Indicatively, Figure 4 As shown, Figure 4 This is a schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. The computer device 300 may be provided as a server. Figure 4 , the computer device 300 includes a processing component 302, which further includes one or more processors, and a memory resource represented by a memory 301, for storing instructions that can be executed by the processing component 302, such as an application. The application stored in the memory 301 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 302 is configured to execute instructions to perform the method of any of the above embodiments.

[0219] The computer device 300 may further include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate based on an operating system stored in the memory 301, such as Windows Server TM, Mac OS X TM, Unix TM, Linux TM, Free BSD TM, or the like.

[0220] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0221] An embodiment of the present application provides a storage medium, in which computer-readable instructions are stored. When the computer-readable instructions are executed by one or more processors, the one or more processors execute a method provided in any embodiment.

[0222] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0223] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can refer to each other.

[0224] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An online advertising push system, characterized in that: The system comprises: Front-end equipment module, data warehouse module, data processing module and information publishing module; Wherein, the front-end device module includes at least one image acquisition unit and at least one display unit, which are used to acquire user characteristic data, transmit the user characteristic data, and display target advertisement content generated according to the user characteristic data; The data warehouse module is used to receive the user characteristic data; The data processing module is used to obtain recommended content tag data through a collaborative filtering algorithm according to the user feature data; The information publishing module is used to determine the target advertisement content according to the recommended content tag data, and transmit the target advertisement content to the front-end device module for display.

2. A method for online advertising push, characterized in that: The method is applied to a front-end device module, the front-end device module includes at least one image acquisition module and at least one display module, and the method includes: Acquire user feature data of the target user through the image acquisition module; Transmitting the user characteristic data to a data warehouse module; Receive target advertisement content from the information publishing module, and display the target advertisement content on the display module.

3. The method according to claim 2, characterized in that The step of obtaining user characteristic data of a target user through the image acquisition module includes: Acquire the user image of the target user within a preset time period through the image acquisition module; Extracting image features from the user image; The body feature data, clothing feature data and behavior feature data of the target user are determined according to the image features.

4. The method according to claim 3, characterized in that Determining the target user's body feature data, clothing feature data, and behavior feature data based on the image features includes: Determining the gender attribution probability, age regression value, and emotional state classification of the target user according to the image features; Determine one or more of the clothing type, clothing style, clothing color and clothing brand classification of the target user through a pre-trained image classification model according to the image features; According to the image features, the user actions of the target user are analyzed by the OpenPose algorithm to determine the interest point classification and behavior pattern classification of the target user.

5. The method according to claim 2, characterized in that: The step of transmitting the user characteristic data to a data warehouse module comprises: Determining first data in the user characteristic data; Performing an encryption operation on the first data to obtain second data, and transmitting the second data to a data warehouse module; The encryption operation includes: Generate a target key of a preset length; Performing data preprocessing and grouping processing on the first data to obtain third data; According to the target key, an encryption operation is performed on the third data to obtain the second data.

6. The method according to any one of claims 2 to 5, characterized in that: The front-end device module includes a first device, a second device and a third device, and the method further includes: Acquire first dynamic data through the image acquisition module of the first device; Acquire second dynamic data through the image acquisition module of the second device; Determine the track data of the target user according to the first dynamic data and the second dynamic data, and determine the third device according to the track data; Transmitting the first dynamic data, the second dynamic data and the trajectory data to a data warehouse module; The dynamic advertisement content is received from the information publishing module through the third device, and the dynamic advertisement content is displayed on the display module of the third device.

7. An online advertisement push method, characterized in that: The method is applied to a data processing module, and the method comprises: Obtain user feature data in the data warehouse module; According to the user feature data, obtaining recommended content tag data through a collaborative filtering algorithm; The content tag data is transmitted to the information publishing module.

8. The method according to claim 7, characterized in that The step of obtaining the recommended content tag data through a collaborative filtering algorithm according to the user feature data includes: Performing data cleaning and data segmentation on the user characteristic data; The recommendation model is trained by a collaborative filtering algorithm until the target performance of the recommendation model meets the preset conditions; The recommendation model generates recommended content label data according to the user feature data.

9. A computer device, characterized in that: The method comprises one or more processors and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the one or more processors, the steps of the method according to any one of claims 2 to 6 or any one of claims 7 to 8 are performed.

10. A storage medium, characterized in that: The storage medium stores computer-readable instructions, and when the computer-readable instructions are executed by one or more processors, the one or more processors execute the steps of the method according to any one of claims 2 to 6 or any one of claims 7 to 8.

Images