An advertising placement and user behavior interaction system

Abstract

The present invention discloses an advertising delivery and user behavior interaction system, which includes an advertising management system module, an advertising delivery module, a data collection module, a data storage module, a user behavior analysis module, a behavior pattern fusion module, an advertising effect feedback and display module, and an advertising delivery strategy optimization and effect tracking module. By using the cosine similarity algorithm to deeply fuse the online and offline user behavior portraits, the present invention solves the problem that specific advertisements cannot be delivered online to users. By using the offline age and appearance characteristics to obtain the consumption behaviors of the corresponding online user groups, the integration of online user consumption and offline user behaviors is realized, so as to perform precise advertising delivery. By using the whale algorithm to deeply mine specific user behaviors, the effective integration of online and offline user behaviors and precise advertising delivery are realized. Through the feedback and re-optimization of advertising effects, it can help advertisers optimize advertising delivery strategies more targeted and reduce advertising delivery costs.

Description

Technical Field

[0001] The present invention belongs to the technical field of advertising placement, and specifically refers to an advertising placement and user behavior interaction system. Background Art

[0002] Currently, there are usually two ways of advertising placement: one is through offline means, such as outdoor billboards, in - elevator terminal advertising machines, etc., and the other is through online means, such as specific web pages on the Internet, platform APPs, etc. Achieving precise advertising placement has become the goal of major advertising platforms. Whether the advertisement is placed precisely directly affects the advertising effect and the advertising conversion rate.

[0003] The existing advertising placement systems have the following disadvantages regarding user interaction behavior:

[0004] (1) Online placement and offline placement do not match the user behavior, and online and offline placements are not integrated, so precise target audience positioning cannot be achieved;

[0005] (2) Online advertisements are usually placed in a large - scale and fixed - time manner, which cannot meet the requirement of targeted and precise advertising placement for specific target personnel in a specific small range, reducing the advertising placement effect, having a high placement cost, and being unable to convert online advertisement content into offline placement;

[0006] (3) Offline user behavior focuses on appearance, dressing, and emotions, which are difficult for advertisers to define, and it is difficult to implement precise placement for users;

[0007] Therefore, how to provide an advertising placement and user behavior interaction system to solve the defects in the prior art, such as the inability to integrate online and offline placements, the inability of online advertisements to be precisely targeted at small - range users, the inability of offline advertisements to be intelligently placed, and the low - efficiency advertising placement algorithm, has become an urgent technical problem for those skilled in the art. Summary of the Invention

[0008] In order to solve the problems in the above - mentioned prior art that existing online advertisements cannot meet the requirement of targeted and specific - target - personnel advertising placement in a specific small range, and offline user behavior focuses on appearance, dressing, and emotions, which are difficult for advertisers to define and it is difficult to implement precise placement for users, etc., the present invention proposes an advertising placement and user behavior interaction system.

[0009] To solve the above - mentioned technical problems, the technical solutions adopted by the present invention are as follows:

[0010] An advertising delivery and user behavior interaction system, characterized in that it includes an advertising management system module, an advertising delivery module, a data collection module, a data storage module, a user behavior analysis module, a behavior pattern fusion module, an advertising effect feedback and display module, and an advertising delivery strategy optimization and effect tracking module;

[0011] The advertising management system module is used to manage the scheduling of the business, data, and communication of each module of the system;

[0012] The data collection module is used to collect the online and offline behavior data of users;

[0013] The data storage module is used to store the data collected by the data collection module;

[0014] The user behavior analysis module is used to obtain the behavior data of online and offline users, and respectively establish an online user portrait and an offline user portrait according to the behavior data;

[0015] The behavior pattern fusion module deeply fuses the online and offline user behavior portraits through the cosine similarity algorithm to achieve the integration of online and offline user behaviors; by using the offline age and appearance characteristics to obtain the consumption behaviors of the corresponding online user groups, it realizes the integration of online consumption behaviors, so as to perform precise advertising delivery;

[0016] The advertising effect feedback and display module is used to evaluate the advertising delivery effects for online and offline, and provide a visual display interface for each analysis result;

[0017] The advertising delivery strategy optimization and effect tracking module is used to further optimize the advertising delivery strategy according to the online or offline user behavior preferences. The optimization of the strategy at least includes the integration of online and offline two-way users, the diversification change of offline advertising display content, the novelty of offline advertising display content, and the adjustment of the online and offline user behavior weight values, so as to achieve precise and efficient advertising delivery;

[0018] The advertising delivery module is used for online and offline advertising delivery methods.

[0019] Furthermore, the data storage module includes a distributed storage unit, a data backup unit, and a data recovery unit;

[0020] The distributed storage unit is used to perform distributed storage on the database, store it on multiple servers, and also store it in the cloud to ensure the security and reliability of the data;

[0021] The data backup unit is used to regularly back up the data to prevent data loss;

[0022] The data recovery unit is used to quickly recover the data when the data is lost or damaged;

[0023] The data acquisition module includes online data acquisition and offline data acquisition;

[0024] The online data acquisition includes age, gender, obesity level, dressing style and hairstyle, advertisement display frequency, click count, download volume, usage duration, viewing frequency, social data, shopping records and consumption habits, and this data is stored in the data storage module;

[0025] The age constructs a feature vector based on the shopping records used online. The records at least include the birth date records filled in each APP, and the value of the feature vector is 0 ≤ a1 ≤ 100;

[0026] The gender constructs a feature vector based on the shopping records used online. The records at least include shopping goods that distinguish genders. The value for male is 1, and the value for female is 0;

[0027] The obesity level constructs a feature vector based on the user behavior and shopping records used online. The records at least include the sizes of clothes purchased, and the value of the feature vector is 0 ≤ a2 ≤ 200;

[0028] The dressing style and hairstyle constructs a feature vector based on the user behavior and shopping records used online. The records at least include the styles of clothes purchased, and the value of the feature vector is 0 ≤ a3 ≤ 100;

[0029] The offline data acquisition includes age, gender, obesity level, dressing style and hairstyle, viewing frequency, viewing duration, user facial content and voice content, and this data is stored in the data storage module;

[0030] Further, the advertisement delivery module includes an online advertisement delivery unit and an offline advertisement delivery unit;

[0031] The online advertisement delivery unit is used for online advertisement delivery and delivers advertisements online in a time - period - based manner;

[0032] The offline advertisement delivery unit is used for offline advertisement delivery and delivers advertisements by using a terminal advertisement machine. The terminal advertisement machine is installed with a camera that can capture user behavior and appearance features, and this camera has a voice sensing function; the terminal advertisement machine has the function of playing multiple advertisement pictures simultaneously.

[0033] Further, the data acquisition module includes a user online data acquisition unit and a user offline data acquisition unit;

[0034] The user online data collection module is used to collect the user's online behavior data, and the online behavior data includes the user's browsing advertisement records, searching advertisement records, clicking advertisement records, commenting advertisement records, social records, and purchasing records on the website or application;

[0035] The user offline data collection unit includes a terminal advertising machine data collection unit and a terminal advertising machine data processing unit;

[0036] The terminal advertising machine data collection unit uses the camera to collect the user's behavior data, and the data at least includes the user's dressing behavior and facial expressions, and uses a voice sensing device to collect the user's behavior data, and the data at least includes the user's comments on the advertisement and the user's interest and hobby comments. The above collected data is stored in the data storage module;

[0037] The terminal advertising machine data processing unit includes an image processing unit and a voice processing unit;

[0038] The image processing unit is used to perform image recognition using a convolutional neural network of deep learning, generate an image recognition result, and store it in the data storage module;

[0039] The voice processing unit is used to perform voice analysis using natural language processing and speech recognition algorithms, generate a voice analysis result, and store it in the data storage module;

[0040] After the image and voice are processed, they include the frequency record of watching advertisements, the duration record of watching advertisements, the comment record of watching advertisements, the facial expression record of watching, the voice record of watching advertisements, the user's dressing, interests and hobbies, and other behavior action records;

[0041] Furthermore, the user behavior analysis module includes a user behavior online analysis unit, a user behavior offline analysis unit, and an offline advertisement placement strategy unit;

[0042] The user behavior online analysis unit is used to obtain the user's online behavior characteristics and create a user online portrait according to the user's online behavior characteristics. The user's online behavior characteristics at least include age, gender, height and obesity, dressing, hairstyle, online interests and hobbies, online purchasing habits, online social hobbies, and online search habits;

[0043] The online purchasing habits include the types of online purchased goods and the frequency of purchasing goods;

[0044] The offline user behavior analysis unit is used to obtain the offline behavior characteristics of users and create an offline user profile based on the offline behavior characteristics of users. The offline behavior characteristics of users at least include the level of offline emotion, dressing preferences, offline hobbies, offline hairstyle preferences, and offline personalized appearance.

[0045] The offline advertising placement strategy unit is used to perform an intelligent advertising placement strategy only for the offline behavior of users, including a first offline placement strategy and a second offline placement strategy.

[0046] The first offline placement strategy:

[0047] L1: Use the image processing unit to determine whether the user appears on the terminal advertising machine for the first time.

[0048] L2: If the judgment result is the first appearance, further use the image processing unit to analyze the content of the user's eye image and determine whether the content of the eye image has the advertising content played by the terminal advertising machine.

[0049] L3: If the judgment result is the played advertising content, at the same time determine that the user is a newly concerned advertising user.

[0050] L4: For this newly concerned advertising user, change the content played by the terminal advertising machine in real time. The change method uses video playback, and the playback content is the hottest or most popular advertisement in the area.

[0051] The second offline placement strategy is used to determine that the image processing unit analyzes the content of the user's eye image, and if the judgment result is that the user has not paid attention to any advertising content, the method is to increase the sound played by the terminal advertising machine to attract the user's attention and transfer the user's attention to the advertising content of the terminal advertising machine.

[0052] Furthermore, the behavior pattern fusion module includes a first fusion unit and a second fusion unit.

[0053] The first fusion unit fuses the consumption behaviors of specific online groups according to the age and appearance characteristics of offline users. The implementation process is as follows:

[0054] M5: Use the image processing unit to obtain the basic information of the user. The basic information includes age, gender, height, obesity degree, dressing and hairstyle.

[0055] M6: Search the online behavior data of all users in the data storage module, and set the search conditions as: age, gender, height, obesity degree, dressing and hairstyle.

[0056] M7: Search by the stated age. The search results are divided into three categories: young users are those under 30 years old, middle-aged users are those between 30 and 60 years old, and elderly users are those over 60 years old;

[0057] M8: Search by gender. The search results are divided into male users and female users. Male users tend to purchase electronic products and sports goods online, while female users may be more concerned about products in categories such as clothing, beauty, and home decoration;

[0058] M9: Search by height and obesity level. The search results are divided into users with normal height, standard height, non-standard height, users with normal body shape, standard body shape, and non-standard body shape. Users with normal body shape tend to purchase non-customized products, while users with non-standard body shape tend to purchase customized items or medications;

[0059] M10: Search by dressing style and hairstyle. The search results are divided into personalized types and loyal fan types. Users of the loyal fan type may be more inclined to purchase products of specific brands, while users pursuing personalization may prefer customized or uniquely designed products;

[0060] M11: The specific process of the search is as follows:

[0061]

[0062] In the formula, x1 represents the online age feature vector value of the user, y1 represents the online gender feature vector value of the user, z1 represents the online height and obesity feature vector value of the user, d1 represents the online dressing style and hairstyle feature vector value of the user, x2 represents the offline age feature vector value of the user, y2 represents the offline gender feature vector value of the user, z2 represents the offline height and obesity feature vector value of the user, and d2 represents the offline dressing style and hairstyle coefficient feature vector value of the user. a represents the age proportion coefficient, b represents the gender proportion coefficient, c represents the height and obesity proportion coefficient, and d represents the dressing style and hairstyle proportion coefficient;

[0063] Where 0.44 ≤ a ≤ 0.54, 0.33 ≤ b ≤ 0.38, 0.19 ≤ c ≤ 0.24, 0.16 ≤ d ≤ 0.21 and a + b + c + d ≤ 1, and E represents the correction constant;

[0064] Using formula (1), set the specific values of a, b, c, and d, search the online data of all users in the data storage module, and set the threshold of Q to DD;

[0065] M12: If Q is less than the threshold DD, it is the online user found. There are multiple online users found. These online users are grouped into one category. The advertising placement rule for this offline user is: obtain the online user portraits of the online group found through the data storage module, and randomly select N types of products with high purchase frequencies corresponding to the online user portraits;

[0066] M13: The N types of product advertisements are played on the corresponding advertising terminal in a split-screen manner. There is a QR code on each advertisement, and this QR code is presented in a high-brightness stroboscopic manner. The QR code is used to link to the online APP shopping address corresponding to the product. Users can purchase the product by simply scanning this QR code with their mobile phones;

[0067] The advertising terminal determines that the user has not scanned the QR code to purchase the product and performs the operation of the second fusion unit;

[0068] The second fusion unit, according to the offline behavior portrait, matches and searches all the online user portraits in the database module through the cosine similarity algorithm, finds the corresponding online portrait of the user, and thus determines the advertising delivery method;

[0069] Cosine similarity algorithm:

[0070]

[0071] The cosine similarity algorithm can be used to calculate the similarity of two portraits. The larger the value of cosθ, the more similar the two portraits are, and vice versa;

[0072] The processing process of the cosine similarity algorithm:

[0073] M1: Feature vector construction. The online behavior portrait and offline behavior portrait of the user are converted into numerical feature vectors. The features of the user's online behavior portrait include online hobbies, online purchase habits, online social hobbies, and online search habits. The features of the offline behavior portrait include the level of offline emotion, dressing and grooming hobbies, offline hobbies, offline hairstyle hobbies, and offline personalized appearance. The behavior features are quantified as elements in the vector;

[0074] M2: Cosine similarity calculation. The online behavior portrait and offline behavior portrait are calculated using formula (2);

[0075] M4: Similarity evaluation. The larger the value of cosθ, the greater the difference between the online behavior portrait and the offline behavior portrait. By setting a threshold, the similarity degree of the user's online and offline behaviors can be judged;

[0076] Using the cosine similarity algorithm and making the algorithm hold within the threshold range, two online and offline user portraits can be matched, so that the online user portrait and the offline user portrait are fused into the same user;

[0077] The fusion process is as follows: When a user passes by the terminal advertising machine or watches the advertisement on the terminal advertising machine, the terminal advertising machine captures the user's image, and determines whether a user's offline portrait has been established for this user image. If it has been established, the Euclidean similarity algorithm is used to match the online portraits of all users in the database module to find the corresponding online user portrait of this user. According to the fused portrait result, advertisements are accurately placed on this terminal advertising machine. The advertisement management system module adopts four placement methods according to the fusion result, namely the first placement method, the second placement method, the third placement method, and the fourth placement method. These four placement methods use the whale algorithm to search the user data in the database module. The implementation process of the whale algorithm:

[0078] D = ∣CX*(t) - X(t)∣ (Equation 3),

[0079] X(t + 1) = X*(t) - AD (Equation 4),

[0080] t is the current iteration number, A and C are representation coefficients, X*(t) represents the best position so far, that is, the optimal solution, and X(t) represents the current search position;

[0081] A = 2ar 1 -a,

[0082] C = 2r 2 ,

[0083] a = 2 - 2t / Tmax,

[0084] where, r 1 and r 2 are random numbers in (0, 1), the value of a linearly decreases from 2 to 0, t represents the current iteration number, and Tmax is the maximum iteration number,

[0085] X(t + 1) = X*(t) + Dpe bl cos(2πd) (Equation 5),

[0086] where, Dp = ∣X*(t) - X(t)∣, representing the distance between the behavior of the current user and the position of the behavior of the searched user, X*(t) represents the best position vector so far, b is a constant used to define the shape of the spiral, and d is a random number in (-1, 1),

[0087] D = ∣CXrand - X(t)∣ (Equation 6),

[0088] X(t + 1) = Xrand - AD (Equation 7),

[0089] Among them, Xrand is the randomly selected user behavior position vector. When the algorithm sets A≥1, a search agent is randomly selected, the number of user behavior explorations is set to N, the user behavior exploration boundary is set to the online and offline behavior records of the database module for three months, the number of iterations is set to T, and finally the optimal feature subset is searched out as N user behavior sets. Then, the N user behavior sets are converted into corresponding N advertising placement sets for N-screen placement on the terminal advertising machine;

[0090] The first placement method: The terminal advertising machine plays N advertisements in full-screen polling, and the N advertisements are estimated by the whale algorithm; Conditions for adopting this placement method: The online shopping characteristics are obvious in the recent N days;

[0091] The second placement method: The terminal advertising machine plays N advertisements in split-screen, and the N advertisements are the advertisements that best match the user behavior, estimated by the whale algorithm; Conditions for adopting this placement method: The online shopping characteristics are not obvious in the recent N days;

[0092] The third placement method: The terminal advertising machine plays N advertisements in split-screen, and the N advertisements are novel, not the most popular or booming, and not seen by users usually, estimated by the whale algorithm; Conditions for adopting this placement method: The online shopping characteristics are not obvious in the recent N days, and the online shopping characteristics are obvious in the recent 2N days;

[0093] The above three placement methods are only for the situation where there is only one user in front of the terminal advertising machine;

[0094] The fourth placement method: The terminal advertising machine plays N advertisements in split-screen, and the N advertisements correspond to N users, each user corresponds to one advertisement, and the optimal advertisement for each user is estimated by the whale algorithm, and then the N advertisements are placed through combination;

[0095] The fourth placement method is for N users in front of the terminal advertising machine;

[0096] The presentation method of the above four placement methods on the advertising terminal machine is: There is a QR code on each advertisement, and this QR code is presented in a high-brightness stroboscopic manner. The QR code is used to link to the online APP shopping address of the corresponding commodity, and users can purchase the commodity by simply scanning this QR code with their mobile phones.

[0097] Furthermore, the advertisement effect feedback and display module includes an online advertisement placement cost unit, an online user feedback unit, an online user behavior statistics unit, an offline advertisement placement cost unit, an offline user feedback unit, an offline user behavior statistics unit, an analysis result evaluation unit, and a display unit;

[0098] The online advertisement placement cost unit is used to comprehensively count the online advertisement input cost in a statistical time period;

[0099] The online user feedback unit is used to comprehensively count the online feedback of users on advertisements during a time period, and at least includes the number of times users view advertisements, the time users stay on advertisements, and the comments of users on advertisements;

[0100] The online user behavior statistics unit is used to comprehensively count the purchase strength of users online for the products related to the advertised products during a time period, including whether users purchase the products related to the advertisements, the number of times users purchase the advertised products, and the comments after purchasing the advertised products;

[0101] The offline advertisement investment cost unit is used to comprehensively count the offline advertisement investment cost during a time period;

[0102] The offline user feedback unit is used to comprehensively count the offline feedback of users on advertisements during a time period, and at least includes the time users stay on advertisements and the comments of users on advertisements;

[0103] The offline user behavior statistics unit is used to comprehensively count the purchase strength of users offline for the products related to the advertised products during a time period, including whether users purchase the products related to the advertisements, the number of times users purchase the advertised products, and the comments after purchasing the advertised products;

[0104] The analysis result evaluation unit is used to calculate the relationship between user behavior and advertisement cost, and the evaluation result index Yxcl is obtained by calculating through the following formula:

[0105] Q = Sa*a + Sb*b + Sc*c + Sd*d + E (Formula VIII),

[0106] In the formula, Sa represents the online cost coefficient of the advertisement, Sb represents the online comprehensive performance coefficient of the user, Sc represents the offline cost coefficient of the advertisement, Sd represents the offline comprehensive performance coefficient of the user, a represents the proportional coefficient of the online cost coefficient Sa of the advertisement, b represents the proportional coefficient of the online comprehensive performance coefficient Sb of the user, c represents the proportional coefficient of the offline cost coefficient Sc of the advertisement, d represents the proportional coefficient of the offline comprehensive performance coefficient Sd of the user;

[0107] Among them, 0.24 ≤ a ≤ 0.38, 0.23 ≤ b ≤ 0.31, 0.19 ≤ c ≤ 0.34, 0.16 ≤ d ≤ 0.28 and a + b + c + d ≤ 1, a, b, c, d are the weight values of online and offline user behavior, E represents the correction constant, the larger the Q value, the better the advertisement placement effect, otherwise it is not good;

[0108] The display unit is used to display the online and offline behavior data of users or the data of the results of the analysis result evaluation unit in a graphical interface in real time for further decision-making by advertisers.

[0109] Furthermore, the advertisement placement strategy optimization and effect tracking module includes: an online and offline user integration optimization unit, an online and offline user behavior weight value optimization unit, an offline advertisement placement content optimization unit, and an effect tracking unit;

[0110] The online and offline user integration optimization unit is used to deeply optimize and integrate the online and offline portraits of users, continuously optimize the integration using the cosine similarity algorithm, find the fit point between online and offline behaviors, further integrate the online behavior portrait into the offline behavior portrait, and perform more optimal targeted and precise advertisement placement;

[0111] The online and offline user behavior weight value optimization unit is used to comprehensively analyze the current user behavior, adjust the weight value of the user's online or offline behavior in real time, and optimize the advertisement placement strategy for online and offline conversion;

[0112] The offline advertisement placement content optimization unit is used to deeply analyze according to the integrated portrait, diversify and change the offline display content, and at the same time adjust the search parameters of the search algorithm of the whale algorithm, and bias the adjusted parameters towards the novelty of the user advertisement. The novelty means that the advertisement combination has not been placed within a time period or the user has not been exposed to the advertisement placement within a time period;

[0113] The effect tracking unit is used to track the effect of the new advertisement placement strategy, calculate through the analysis result evaluation unit. If the expected effect is not achieved, the improved whale algorithm is used for advertisement placement. The improved whale algorithm adopts a global search method and can perform a cohesion search for the behavior between users.

[0114] The beneficial effects of an advertisement placement and user behavior interaction system of the present invention are as follows:

[0115] (1) The advertisement placement and user behavior interaction system is a system for advertisement placement and user behavior interaction. The terminal advertisement machine identifies the user behavior and age and appearance characteristics through image processing, and performs intelligent advertisement placement on the user through multiple schemes. The schemes include deeply integrating the online and offline portraits of the same user using the cosine similarity algorithm, obtaining the online corresponding user group consumption behavior using the offline age and appearance characteristics, realizing the integration of online consumption and offline behavior, and helping advertisers more pertinently optimize the design and placement strategy of advertisements through the evaluation and re-optimization of advertisement placement problems;

[0116] (2) The cosine similarity algorithm is used to fuse and match the offline user portrait and the online user portrait, find the corresponding online user portrait of the offline user, fuse them into the same user, and then use the whale algorithm to search for the most suitable advertisement for this user, and intelligently display the advertisement on the advertisement terminal that the user passes by. This algorithm can achieve precise target audience integration and realize precise advertisement placement;

[0117] (3) By using the offline age and appearance characteristics to obtain the consumption behavior of the corresponding online user group, the integration of online consumption and offline behavior is realized, meeting the random intelligent targeted advertisement placement method, reducing the online advertisement placement cost, and well converting the online consumption content into offline advertisement placement, accurately implementing the advertisement placement for offline users. Description of the Drawings

[0118] Figure 1 It is the system architecture diagram of an advertisement placement and user behavior interaction system of the present invention;

[0119] Figure 2 It is the data collection module diagram of an advertisement placement and user behavior interaction system of the present invention;

[0120] Figure 3 It is the behavior pattern fusion module diagram of an advertisement placement and user behavior interaction system of the present invention.

[0121] Among them, 1 - advertisement management system module, 2 - data collection module, 21 - user online data collection unit, 22 - user offline data collection unit, 221 - terminal advertisement machine data collection unit, 222 - terminal advertisement machine data processing unit, 2221 - image processing unit, 2222 - voice processing unit, 3 - data storage module, 31 - distributed storage unit, 32 - data backup unit, 33 - data recovery unit, 4 - user behavior analysis module, 41 - user behavior online analysis unit, 42 - user behavior offline analysis unit, 43 - offline advertisement placement strategy unit, 5 - behavior pattern fusion module, 51 - first fusion unit, 52 - second fusion unit, 6 - advertisement effect feedback and display module, 61 - online advertisement placement cost unit, 62 - online user feedback unit, 63 - online user behavior statistics unit, 64 - offline advertisement placement cost unit, 65 - offline user feedback unit, 66 - offline user behavior statistics unit, 67 - analysis result evaluation unit, 68 - display unit, 7 - advertisement placement strategy optimization and effect tracking module, 71 - online and offline two-way user fusion optimization unit, 72 - online and offline user behavior weight value optimization unit, 73 - offline advertisement placement content optimization module and effect tracking unit, 74 - effect tracking unit, 8 - advertisement placement module, 81 - online advertisement placement unit, 82 - offline advertisement placement unit. Detailed Embodiment

[0122] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0123] As Figures 1-3 shown, an advertising placement and user behavior interaction system includes an advertising management system module 1, an advertising placement module 8, a data collection module 2, a data storage module 3, a user behavior analysis module 4, a behavior pattern fusion module 5, an advertising effect feedback and display module 6, and an advertising placement strategy optimization and effect tracking module 7;

[0124] The advertising management system module 1 is used to manage the scheduling of the business, data, and communication of each module of the system;

[0125] The data collection module 2 is used to collect the online and offline behavior data of users;

[0126] The data storage module 3 is used to store the data collected by the data collection module 2;

[0127] The user behavior analysis module 4 is used to obtain the behavior data of online and offline users, and respectively establish an online user portrait and an offline user portrait according to the behavior data;

[0128] The behavior pattern fusion module 5 deeply fuses the online and offline user behavior portraits through the cosine similarity algorithm to achieve the integration of online and offline user behaviors; by using the offline age and appearance features to obtain the consumption behaviors of the corresponding online user groups, the online consumption behavior integration is realized, so as to perform precise advertising placement;

[0129] The advertising effect feedback and display module 6 is used to evaluate the advertising placement effects for online and offline, and provide a visual display interface for each analysis result;

[0130] The advertising placement strategy optimization and effect tracking module 7 is used to further optimize the advertising placement strategy according to the online or offline user behavior preferences. The optimization of the strategy at least includes the integration of online and offline two-way users, the diversification change of the offline advertising display content, the novelty of the offline advertising display content, and the adjustment of the online and offline user behavior weight values, so as to achieve precise and efficient advertising placement;

[0131] The advertising placement module 8 is used for the online and offline advertising placement methods.

[0132] Furthermore, the data storage module 3 includes a distributed storage unit 31, a data backup unit 32, and a data recovery unit 33;

[0133] The distributed storage unit 31 is used to perform distributed storage of the database on multiple servers and also in the cloud to ensure data security and reliability;

[0134] The data backup unit 32 is used to regularly back up data to prevent data loss;

[0135] The data recovery unit 33 is used to quickly recover data when data is lost or damaged;

[0136] The data collection module 2 includes online data collection and offline data collection;

[0137] The online collected data includes age, gender, obesity level, dressing style, advertising display frequency, number of clicks, downloads, usage time, viewing frequency, social data, shopping records and consumption habits, and the data is stored in the data storage module 3;

[0138] The age constructs a feature vector based on online shopping records, the records at least include the date of birth records filled in by each APP, and the feature vector value is 0≤a1≤100;

[0139] The gender constructs a feature vector based on online shopping records, where the records at least include shopping items distinguished by gender, with a value of 1 for males and a value of 0 for females;

[0140] The obesity degree constructs a feature vector based on the online user behavior and shopping records, the records at least include the size of the purchased clothes, and the feature vector has a value of 0≤a2≤200;

[0141] The dressing and hairstyle constructs a feature vector based on the online user behavior and shopping records, wherein the records at least include the styles of clothes purchased, and the feature vector has a value of 0≤a3≤100;

[0142] The offline collected data includes age, gender, obesity level, dressing style, viewing frequency, viewing time, user facial content and voice content, and the data is stored in the data storage module 3;

[0143] Further, the advertisement delivery module 8 includes an online advertisement delivery unit 81 and an offline advertisement delivery unit 82;

[0144] The online advertisement delivery unit 81 is used for online advertisement delivery, and delivers advertisements online in a time period manner;

[0145] The offline advertising placement unit 82 is used for offline advertising placement. It uses a terminal advertising machine to place advertisements. The terminal advertising machine is equipped with a camera that can capture users' behaviors and appearance features, and this camera has a voice sensing function. The terminal advertising machine has the function of playing multiple advertisement pictures simultaneously.

[0146] Further, the data acquisition module 2 includes a user online data acquisition unit 21 and a user offline data acquisition unit 22;

[0147] The user online data acquisition module 2 is used to collect users' online behavior data. The online behavior data includes users' browsing advertisement records, searching advertisement records, clicking advertisement records, commenting on advertisements records, social records, and purchase records on websites or applications;

[0148] The user offline data acquisition unit 22 includes a terminal advertising machine data acquisition unit 221 and a terminal advertising machine data processing unit 222;

[0149] The terminal advertising machine data acquisition unit 221 uses the camera to collect users' behavior data, and the data at least includes users' dressing behaviors and facial expressions. It uses a voice sensing device to collect users' behavior data, and the data at least includes users' comments on advertisements and comments on users' interests and hobbies. The above collected data is stored in the data storage module 3;

[0150] The terminal advertising machine data processing unit 222 includes an image processing unit 2221 and a voice processing unit 2222;

[0151] The image processing unit 2221 is used to perform image recognition using a convolutional neural network of deep learning, generate an image recognition result, and store it in the data storage module 3;

[0152] The voice processing unit 2222 is used to perform voice analysis using natural language processing and speech recognition algorithms, generate a voice analysis result, and store it in the data storage module 3;

[0153] After the image and voice are processed, they include the frequency record of watching advertisements, the duration record of watching advertisements, the comment record of watching advertisements, the facial expression record of watching, the voice record of watching advertisements, users' dressing styles, interests and hobbies, and other behavior action records;

[0154] Further, the user behavior analysis module 4 includes a user online behavior analysis unit 41, a user offline behavior analysis unit 42, and an offline advertising placement strategy unit 43;

[0155] The online user behavior analysis unit 41 is used to obtain the online behavior characteristics of users and create an online user profile based on the online behavior characteristics of users. The online behavior characteristics of users at least include age, gender, height and obesity, dressing style and hairstyle, online hobbies, online purchase habits, online social hobbies, and online search habits;

[0156] The online purchase habits include the types of online purchased goods and the frequency of purchasing goods;

[0157] The offline user behavior analysis unit 42 is used to obtain the offline behavior characteristics of users and create an offline user profile based on the offline behavior characteristics of users. The offline behavior characteristics of users at least include the level of offline emotion, dressing style hobbies, offline hobbies, offline hairstyle hobbies, and offline personalized appearance;

[0158] The offline advertising placement strategy unit 43 is used to perform an intelligent advertising placement strategy only for the offline behavior of users, including a first offline placement strategy, a second offline placement strategy, and a third offline placement strategy;

[0159] The first placement strategy:

[0160] L1: Use the image processing unit 2221 to determine whether the user appears at the terminal advertising machine for the first time;

[0161] L2: If the judgment result is the first appearance, further use the image processing unit 2221 to analyze the content of the user's naked-eye image and determine whether the content of the naked-eye image has the advertising content played by the terminal advertising machine;

[0162] L3: If the judgment result is the played advertising content, and at the same time determine that the user is a newly concerned advertising user;

[0163] L4: For this newly concerned advertising user, the content played by the terminal advertising machine is changed in real time. The change method uses video playback, and the playback content is the hottest or most popular advertisement in the area;

[0164] The process of the second placement strategy:

[0165] L5: Use the image processing unit 2221 to determine whether the user appears at the terminal advertising machine for the first time;

[0166] L6: If the judgment result is the first time, further use the image processing unit 2221 to analyze the content of the user's naked-eye image and determine whether the content of the naked-eye image has the advertising content played by the terminal advertising machine;

[0167] L7: If the judgment result is the played advertising content, and at the same time determine that the user is an old concerned advertising user;

[0168] L8: For the old ad - focused users, the content played on the terminal advertising machine is changed in real - time. The change method uses picture playback, with M ads played in a split - screen manner at a time, and the ad content is ads that the user does not often pay attention to.

[0169] The third offline placement strategy is used to judge the content of the user's eye image analyzed by the image processing unit 2221. If the judgment result is that the user does not pay attention to any ad content, the method adopted is to increase the sound played on the terminal advertising machine to attract the user's attention and shift the user's attention to the ad content on the terminal advertising machine.

[0170] Furthermore, the behavior pattern fusion module 5 includes a first fusion unit 51 and a second fusion unit 52;

[0171] The first fusion unit fuses the consumption behaviors of specific online groups according to the age and appearance characteristics of offline users. The implementation process is as follows:

[0172] M5: Use the image processing unit 2221 to obtain the basic information of the user. The basic information includes age, gender, height, obesity degree, dressing style, and hairstyle;

[0173] M6: Search the online behavior data of all users in the data storage module 3, and set the search conditions as: age, gender, height, obesity degree, dressing style, and hairstyle;

[0174] M7: Search by age. The search results are divided into three categories: young users are less than 30 years old, middle - aged users are 30 to 60 years old, and elderly users are 60 years old and above;

[0175] M8: Search by gender. The search results are divided into male users and female users. Male users tend to buy electronic products and sports goods online, while female users may be more concerned about products in categories such as clothing, beauty, and home decoration;

[0176] M9: Search by height and obesity degree. The search results are divided into users with normal height, standard height users, non - standard height users, users with normal body shape, standard body shape users, and non - standard body shape users. Users with normal body shape tend to buy non - customized products, and users with non - standard body shape tend to buy customized things or medicines;

[0177] M10: Search by dressing style and hairstyle. The search results are divided into personalized types and fan types. Fan - type users may be more inclined to buy products of specific brands, while users pursuing personalization may prefer customized or uniquely designed products;

[0178] M11: The specific search process is as follows:

[0179]

[0180] In the formula, x1 represents the online age feature vector value of the user, y1 represents the online gender feature vector value of the user, z1 represents the online height and obesity feature vector value of the user, d1 represents the online dressing, grooming and hairstyle feature vector value of the user, x2 represents the offline age feature vector value of the user, y2 represents the offline gender feature vector value of the user, z2 represents the offline height and obesity feature vector value of the user, and d2 represents the offline dressing, grooming and hairstyle coefficient feature vector value of the user. a represents the age proportion coefficient, b represents the gender proportion coefficient, c represents the height and obesity proportion coefficient, and d represents the dressing, grooming and hairstyle proportion coefficient;

[0181] Among them, 0.44 ≤ a ≤ 0.54, 0.33 ≤ b ≤ 0.38, 0.19 ≤ c ≤ 0.24, 0.16 ≤ d ≤ 0.21 and a + b + c + d ≤ 1, and E represents the correction constant;

[0182] Using formula (1), set the specific values of a, b, c, and d, search the online data of all users in the data storage module 3, and set the threshold of Q to DD;

[0183] M12: If Q is less than the threshold DD, that is, the online user is found. There are multiple online users found. Classify these online users into one group. The advertising placement rule for this offline user is: Obtain the online user portraits of the found online group through the data storage module 3, and randomly select 10 types of goods with high purchase frequency corresponding to the online user portraits;

[0184] M13: Play the advertisements of the 10 types of goods on the current corresponding advertising terminal in a split screen manner. There is a two-dimensional code on each advertisement, and this two-dimensional code is presented in a high-brightness stroboscopic manner. The two-dimensional code is used to link the online APP shopping address corresponding to the goods. The user can purchase the goods by simply scanning this two-dimensional code with a mobile phone;

[0185] If the advertising terminal determines that the user has not scanned the two-dimensional code to purchase the goods, perform the operation of the second fusion unit;

[0186] The second fusion unit, according to the offline behavior portrait, matches and searches all the online user portraits in the database module through the cosine similarity algorithm to find the corresponding online portrait of the user, so as to perform the advertising placement method;

[0187] Cosine similarity algorithm:

[0188]

[0189] The cosine similarity algorithm can be used to calculate the similarity between two portraits. The larger the value of cosθ, the more similar the two portraits are, and vice versa;

[0190] The processing process of the cosine similarity algorithm is as follows:

[0191] M1: Feature vector construction. Convert the online behavior portrait and offline behavior portrait of the user into numerical feature vectors. The features of the user's online behavior portrait include online hobbies, online purchase habits, online social hobbies, and online search habits. The features of the offline behavior portrait include the level of offline mood, dressing preferences, offline hobbies, offline hairstyle preferences, and offline personalized appearance. These behavior features are quantified as elements in the vector.

[0192] M2: Cosine similarity calculation. Use formula (2) to calculate the online behavior portrait and the offline behavior portrait.

[0193] M4: Similarity evaluation. The larger the value of cosθ, the greater the difference between the online behavior portrait and the offline behavior portrait. By setting a threshold, the similarity degree of the user's online and offline behaviors can be judged.

[0194] Using the cosine similarity algorithm and making the algorithm hold within the threshold range can match two user portraits, one online and one offline, so that the online user portrait and the offline user portrait are fused into the same user.

[0195] The fusion process is as follows: When the user passes by the terminal advertising machine or watches the advertisement on the terminal advertising machine, the terminal advertising machine captures the user's image. Determine whether a user's offline portrait has been established for this user image. If it has been established, use the Euclidean similarity algorithm to match the online portraits of all users in the database module to find the corresponding online user portrait of this user. According to the fused portrait result, accurately deliver advertisements on this terminal advertising machine. The advertisement management system module adopts four delivery methods according to the fusion result: the first delivery method, the second delivery method, the third delivery method, and the fourth delivery method. These four delivery methods search the user data in the database module through the whale algorithm. The implementation process of the whale algorithm is as follows:

[0196] D = ∣CX*(t) - X(t)∣ (Formula 3),

[0197] X(t + 1) = X*(t) - AD (Formula 4),

[0198] t is the current iteration number, A and C are representation coefficients, X*(t) represents the best position so far, that is, the optimal solution, and X(t) represents the current search position.

[0199] A = 2ar 1 -a,

[0200] C = 2r 2 ,

[0201] a = 2 - 2t / Tmax,

[0202] where r 1 and r 2 are random numbers in (0, 1), the value of a linearly decreases from 2 to 0, t represents the current iteration number, and Tmax is the maximum iteration number.

[0203] X(t + 1) = X*(t) + Dpe bl cos(2πd) (Equation Five),

[0204] where Dp = ∣X*(t) - X(t)∣, representing the distance between the current user's behavior and the position of the searched user's behavior, X*(t) represents the best position vector so far, b is a constant used to define the shape of the spiral, and d is a random number in (-1, 1).

[0205] D = ∣CXrand - X(t)∣ (Equation Six),

[0206] X(t + 1) = Xrand - AD (Equation Seven),

[0207] where Xrand is the randomly selected user behavior position vector. The algorithm sets that when A ≥ 1, a search agent is randomly selected, the number of user behavior explorations is set to N, the user behavior exploration boundary is set to the online and offline behavior records of the database module for three months, the iteration number T is set, and finally the optimal feature subset is searched as a set of N user behaviors. Again, the set of N user behaviors is converted into corresponding N advertising placement sets for split - screen placement on the terminal advertising machine N - screen.

[0208] The first placement method: The terminal advertising machine plays N advertisements in full - screen polling, and the N advertisements are estimated by the whale algorithm; Conditions for adopting this placement method: The online shopping characteristics are very obvious in the recent N days;

[0209] The second placement method: The terminal advertising machine plays N advertisements in split - screen, and the N advertisements are the advertisements that best match the user's behavior, estimated by the whale algorithm; Conditions for adopting this placement method: The online shopping characteristics are not obvious in the recent N days;

[0210] The third placement method: The terminal advertising machine plays N advertisements in split - screen, and the N advertisements are novel, not the most popular or booming, and not seen by users usually, estimated by the whale algorithm; Conditions for adopting this placement method: The online shopping characteristics are not obvious in the recent N days and are obvious in the recent 2N days;

[0211] The above three placement methods are only for the case where there is only one user in front of the terminal advertising machine;

[0212] The fourth delivery method: The terminal advertising machine plays N advertisements on split screens. The N advertisements correspond to N users, with each user corresponding to one advertisement. The optimal advertisement for each user is estimated through the whale algorithm, and then the N advertisements are delivered through combination;

[0213] The fourth delivery method targets the first N users in front of the terminal advertising machine;

[0214] The presentation method of the four delivery methods on the advertising terminal machine is as follows: There is a QR code on each advertisement, and this QR code is presented in a highlighted and stroboscopic manner. The QR code is used to link to the online APP shopping address corresponding to the commodity, and users can purchase the commodity by simply scanning this QR code with their mobile phones.

[0215] Furthermore, the advertisement effect feedback and display module 6 includes an online advertisement delivery cost unit 61, an online user feedback unit 62, an online user behavior statistics unit 63, an offline advertisement delivery cost unit 64, an offline user feedback unit 65, an offline user behavior statistics unit 66, an analysis result evaluation unit 67, and a display unit 68;

[0216] The online advertisement delivery cost unit 61 is used to comprehensively count the online advertisement input costs for a statistical period;

[0217] The online user feedback unit 62 is used to comprehensively count the online advertisement feedback situations of users for a statistical period, including at least the number of times users view the advertisement, the time users stay on the advertisement, and the comments of users on the advertisement;

[0218] The online user behavior statistics unit 63 is used to comprehensively count the purchase strength situations of users online related to the advertisements delivered for a statistical period, including whether users purchase advertisement-related products, the number of times users purchase advertisement products, and the comments after users purchase advertisement products;

[0219] The offline advertisement delivery cost unit 64 is used to comprehensively count the offline advertisement input costs for a statistical period;

[0220] The offline user feedback unit 65 is used to comprehensively count the offline advertisement feedback situations of users for a statistical period, including at least the time users stay on the advertisement and the comments of users on the advertisement;

[0221] The offline user behavior statistics unit 66 is used to comprehensively count the purchase strength situations of users offline related to the advertisements delivered for a statistical period, including whether users purchase advertisement-related products, the number of times users purchase advertisement products, and the comments after users purchase advertisement products;

[0222] The analysis result evaluation unit 67 is used to calculate the relationship between user behavior and advertisement cost, and the evaluation result index Yxcl is obtained through the following formula calculation:

[0223] Q = Sa * a + Sb * b + Sc * c + Sd * d + E (Equation 8),

[0224] In the formula, Sa represents the online cost coefficient of the advertisement, Sb represents the comprehensive online performance coefficient of the user, Sc represents the offline cost coefficient of the advertisement, Sd represents the comprehensive offline performance coefficient of the user, a represents the proportionality coefficient of the online cost coefficient Sa, b represents the proportionality coefficient of the comprehensive online performance coefficient of the user to Sa, c represents the proportionality coefficient of the offline cost coefficient Sc, and d represents the proportionality coefficient of the comprehensive offline performance coefficient of the user to Sd;

[0225] Among them, 0.24 ≤ a ≤ 0.38, 0.23 ≤ b ≤ 0.31, 0.19 ≤ c ≤ 0.34, 0.16 ≤ d ≤ 0.28 and a + b + c + d ≤ 1. a, b, c, and d are the online and offline user behavior weight values, E represents the correction constant. The larger the Q value, the better the advertising effect, and vice versa;

[0226] The display unit 68 is used to display the online and offline behavior data of the user or the data of the result of the analysis and evaluation unit 67 in real time in a graphical interface for further decision-making by the advertiser.

[0227] Further, the advertising placement strategy optimization and effect tracking module 7 includes: an online and offline user integration optimization unit 71, an online and offline user behavior weight value optimization unit 72, an offline advertising placement content optimization unit 73, and an effect tracking unit 74;

[0228] The online and offline user integration optimization unit 71 is used to deeply optimize and integrate the online and offline portraits of the user, continuously optimize and integrate using the cosine similarity algorithm, find the fit point of the online and offline behaviors, further integrate the online behavior portrait into the offline behavior portrait, and perform more optimal targeted and precise advertising placement;

[0229] The online and offline user behavior weight value optimization unit 72 is used to comprehensively analyze the current user behavior, adjust the weight value of the online or offline behavior of the user in real time, and optimize the advertising placement strategy for online and offline conversion;

[0230] The offline advertising placement content optimization unit 73 is used to deeply analyze according to the integrated portrait, diversify and change the offline display content, and at the same time adjust the search parameters of the search algorithm of the whale algorithm, and bias the adjusted parameters towards the novelty of the user advertisement. The novelty means that the advertisement combination has not been placed within a time period or the user has not been exposed to the advertisement within a time period;

[0231] The effect tracking unit 74 is used to track the effect of the new advertising placement strategy. Through calculation by the analysis result evaluation unit 67, if the expected effect is not achieved, an improved whale algorithm is used for advertising placement. The improved whale algorithm adopts a global search method and can perform adhesion search on the behaviors between users.

[0232] The beneficial effects of an advertising placement and user behavior interaction system of the present invention are as follows:

[0233] (1) The advertising placement and user behavior interaction system is a system for advertising placement and user behavior interaction. The terminal advertising machine identifies user behaviors and age and appearance characteristics through image processing, and performs intelligent advertising placement for users through various solutions. The solutions include using the cosine similarity algorithm to deeply fuse the online and offline portraits of the same user, and using the offline age and appearance characteristics to obtain the consumption behaviors of the corresponding online user groups, so as to realize the integration of online consumption and offline behaviors. By evaluating and re-optimizing the advertising placement problems, it helps advertisers optimize the design and placement strategies of advertisements more pertinently;

[0234] (2) The cosine similarity algorithm is used to fuse and match the offline user portrait and the online user portrait to find the online user portrait corresponding to the offline user, and fuse them into the same user. Then, the whale algorithm is used to search for the most suitable advertisement for this user, and the advertisement is intelligently placed at the advertising terminal passed by the user. This algorithm can achieve precise target audience integration and realize precise advertising placement;

[0235] (3) By using the offline age and appearance characteristics to obtain the consumption behaviors of the corresponding online user groups, the integration of online consumption and offline behaviors is realized, meeting the random intelligent targeted advertising placement method, reducing the online advertising placement cost, and well converting the online consumption content into offline advertising placement, and precisely implementing the offline user advertising placement.

[0236] The above describes the present invention and its implementation manners. Such description is not restrictive. What is shown in the drawings is only one of the implementation manners of the present invention, and the actual content is not limited thereto. Generally speaking, if those of ordinary skill in the art are inspired by it and without departing from the purpose of the present invention creation, without creative design, structures and embodiments similar to the technical solution are designed, they should all fall within the protection scope of the present invention.

Claims

1. An advertisement delivery and user behavior interaction system, characterized in that: include Advertisement management system module (1), used to manage the scheduling of services, data and communications of each module of the system; A data collection module (2), used to collect online and offline behavior data of users; A data storage module (3) used for storing the data collected by the data collection module (2); A user behavior analysis module (4) is used to obtain the behavior data of online and offline users, and to establish an online user profile and an offline user profile based on the behavior data; The behavior pattern fusion module (5) deeply integrates the online and offline user behavior portraits through a similarity algorithm to achieve the integration of online and offline user behaviors; Advertisement effect feedback and display module (6), used to evaluate the online and offline delivery effects of advertisements and provide a visual display interface for each analysis result; Advertisement delivery strategy optimization and effect tracking module (7), used to optimize the advertisement delivery strategy according to online or offline user behavior preferences; An advertisement delivery module (8), used to determine online and offline advertisement delivery methods; The behavior pattern fusion module (5) comprises a first fusion unit (51) and a second fusion unit (52); The first fusion unit (51) fuses the consumption behavior of a specific online group according to the age and appearance characteristics of offline users, and the implementation process is as follows: M5: Use the image processing unit (2221) of the terminal advertising machine to obtain the basic information of the user, which includes age, gender, height, obesity level, clothing and hairstyle; M6: Searching the online behavior data of all users of the data storage module (3), setting the search conditions as: age, gender, height, obesity level, dress and hairstyle; M7: Search by age, the search results are divided into three categories, young users are less than 30 years old, middle-aged users are 30 to 60 years old, and elderly users are over 60 years old; M8: Search by gender, the search results are divided into male users and female users; M9: Search by height and obesity level. The search results are divided into users with normal height, users with standard height, users with non-standard height, users with normal body shape, users with standard body shape, and users with non-standard body shape. M10: Search by dress and hairstyle, the search results are divided into personalized type and loyal fan type; M11: The search process is: Where x1, y1, z1, d1 represent the user's online age, online gender, online height and obesity, and online dressing and hairstyle feature vector values, respectively; x2, y2, z2, d2 represent the user's offline age, offline gender, offline height and obesity, and offline dressing and hairstyle feature vector values, respectively; a', b', c', d' represent the corresponding proportional coefficients; Where a'+b'+c'+d'≤1, a'>b'>c' and a'>b'>d'; Use formula 1, set specific values ​​of a', b', c', and d', search the online data of all users in the data storage module (3), and set the threshold of Q to DD; M12: If Q is less than the threshold DD, it is the online user that has been searched. If there are multiple online users, these online users are classified into a group. The advertising rules for the offline users are as follows: the online user portraits of the searched online group are obtained through the data storage module (3), and N1 commodity categories with high purchase frequency corresponding to the online user portraits are randomly selected; M13: Place N1 product category advertisements on the corresponding terminal advertising machine for split-screen playback. Each advertisement has a QR code, which is presented in a high-brightness strobe mode. The QR code is used to link to the online APP shopping address corresponding to the product. The terminal advertising machine determines that the user has not scanned the QR code to purchase the product, and executes the operation of the second fusion unit (52); The second fusion unit (52) matches and searches all online user portraits of the data storage module using a cosine similarity algorithm according to the offline behavior portrait, finds the online portrait corresponding to the user, and thus delivers the advertisement; The processing process is: M1: Feature vector construction, converting the user's online behavior profile and offline behavior profile into numerical feature vectors. The user's online behavior profile features include online interests and hobbies, online purchasing habits, online social hobbies, and online search habits. The offline behavior profile features include offline emotions, dressing preferences, offline interests and hobbies, offline hairstyle preferences, and offline personalized appearance. M2: Use cosine similarity to calculate online and offline behavior profiles; M4: Similarity evaluation, by setting a threshold, can match the online and offline user portraits, so that the online user portrait and the offline user portrait are merged into the same user; Based on the integrated portrait results, advertisements are accurately delivered to the terminal advertising machine.

2. The advertisement delivery and user behavior interaction system according to claim 1, characterized in that: The data storage module (3) comprises a distributed storage unit (31), a data backup unit (32) and a data recovery unit (33); The distributed storage unit (31) is used to perform distributed storage of the database on multiple servers and also store it in the cloud to ensure data security and reliability; The data backup unit (32) is used to regularly back up data to prevent data loss; The data recovery unit (33) is used to quickly recover data when the data is lost or damaged; The data collection module (2) includes online data collection and offline data collection; The online collected data includes age, gender, obesity level, dressing style, advertising display frequency, number of clicks, download volume, usage time, viewing frequency, social data, shopping records and consumption habits, and the data is stored in the data storage module (3); The age constructs a feature vector based on online shopping records, the records at least include the date of birth records filled in by each APP, and the feature vector value is 0≤a1≤100; The gender constructs a feature vector based on online shopping records, where the records at least include shopping items that distinguish between genders, with a value of 1 for males and a value of 0 for females; The obesity degree constructs a feature vector based on online user behavior and shopping records, the records at least include the size of the purchased clothes, and the feature vector has a value of 0≤a2≤200; The dressing and hairstyle constructs a feature vector based on online user behavior and shopping records, wherein the records at least include the styles of clothes purchased, and the feature vector has a value of 0≤a3≤100; The offline collected data includes age, gender, obesity level, clothing and hairstyle, viewing frequency, viewing time, user facial content and voice content, and the data is stored in the data storage module (3).

3. The advertisement delivery and user behavior interaction system according to claim 1, characterized in that: The advertisement delivery module (8) comprises an online advertisement delivery unit (81) and an offline advertisement delivery unit (82); The online advertisement delivery unit (81) is used for online advertisement delivery, and delivers advertisements online in a time period manner; The offline advertising delivery unit (82) is used for offline advertising delivery and uses a terminal advertising machine to deliver advertisements. The terminal advertising machine is equipped with a camera that can capture user behavior and appearance features, and the camera has a voice sensing function. The terminal advertising machine has the function of playing multiple advertising images at the same time.

4. The advertisement delivery and user behavior interaction system according to claim 1, characterized in that: The data collection module (2) comprises a user online data collection unit (21) and a user offline data collection unit (22); The user online data collection unit (21) is used to collect the user's online behavior data, wherein the online behavior data includes the user's browsing advertisement record, searching advertisement record, clicking advertisement record, commenting advertisement record, social record and purchasing record on the website or application; The user offline data collection unit (22) comprises a terminal advertising machine data collection unit (221) and a terminal advertising machine data processing unit (222); The terminal advertising machine data processing unit (222) includes an image processing unit (2221) and a voice processing unit (2222) The terminal advertising machine data collection unit (221) uses a camera to collect user behavior data, the data at least including the user's dressing behavior and facial expressions, and uses a voice sensing device to collect user behavior data, the data at least including the user's comments on the advertisement and the user's interests and hobbies. The above collected data is stored in the data storage module (3); The image processing unit (2221) is used to perform image recognition using a deep learning convolutional neural network to generate image recognition results, which are stored in the data storage module (3); The speech processing unit (2222) is used to perform speech analysis using natural language processing and speech recognition algorithms to generate speech analysis results, which are stored in the data storage module (3); The images and voices, after being processed, include records of the frequency of viewing advertisements, the duration of viewing advertisements, comments on viewing advertisements, facial expressions during viewing, voice records of viewing advertisements, and the user's attire and interests.

5. The advertisement delivery and user behavior interaction system according to claim 1, characterized in that: The user behavior analysis module (4) comprises an online user behavior analysis unit (41), an offline user behavior analysis unit (42) and an offline advertising delivery strategy unit (43); The user behavior online analysis unit (41) is used to obtain the user's online behavior characteristics and create an online user portrait based on the user's online behavior characteristics, wherein the user's online behavior characteristics at least include age, gender, height, obesity, dress and hairstyle, online interests and hobbies, online purchasing habits, online social hobbies and online search habits; The online purchasing habits include the types of goods purchased online and the frequency of purchasing goods; The user behavior offline analysis unit (42) is used to obtain the user's offline behavior characteristics and create an offline user portrait based on the user's offline behavior characteristics, wherein the user's offline behavior characteristics at least include offline mood, dressing preferences, offline hobbies, offline hairstyle preferences and offline personalized appearance; The offline advertising delivery strategy unit (43) is used to implement intelligent advertising delivery strategies only for offline behaviors of users, including a first offline delivery strategy and a second offline delivery strategy; The first offline delivery strategy: L1: using the image processing unit (2221) to determine whether the user appears in the terminal advertising machine for the first time; L2: the result of the judgment is that it occurs for the first time, and the image processing unit (2221) is further used to analyze the user's naked eye image content to judge whether the naked eye image content contains the advertisement content played by the terminal advertising machine; L3: The judgment result is the advertisement content played, and the user is determined to be a new advertisement-focused user; L4: for the new user who follows the advertisement, the content played by the terminal advertisement machine is changed in real time, and the change method is video playback, and the content played is the hottest advertisement in the area; The second offline delivery strategy is used for the image processing unit (2221) to judge and analyze the image content of the user's naked eye, and the judgment result is that the user does not pay attention to any advertising content. The method is to increase the sound played by the terminal advertising machine to attract the user's attention, so that the user's attention is transferred to the advertising content of the terminal advertising machine.

6. The advertisement delivery and user behavior interaction system according to claim 1, characterized in that: The fusion process of the second fusion unit (52) is as follows: when a user passes by the terminal advertising machine or watches an advertisement on the terminal advertising machine, the terminal advertising machine collects a user image and determines whether an offline user portrait has been established for the user image. If an offline user portrait has been established, the online portraits of all users in the data storage module are matched using a cosine similarity algorithm to find the online user portrait corresponding to the user. According to the fused portrait result, an advertisement is accurately delivered to the terminal advertising machine. According to the fusion result, the advertisement management system module adopts four delivery methods, namely, a first delivery method, a second delivery method, a third delivery method and a fourth delivery method. The four delivery methods use a whale algorithm to search the user data of the data storage module. The whale algorithm implementation process is as follows: D=|CX*(t)-X(t)|(Formula 3), X(t+1)=X*(t)-AD (Formula 4), t is the current iteration number, A and C represent coefficients, X*(t) represents the best position so far, that is, the optimal solution, and X(t) represents the current search position; A=2ar1-a, C＝2r2， a=2-2t / Tmax, Among them, r1 and r2 are random numbers in (0,1), the value of a decreases linearly from 2 to 0, t represents the current number of iterations, and Tmax represents the maximum number of iterations. X(t+1)=X*(t)+Dpe bd cos(2πd)(Formula 5), Where Dp = |X*(t)-X(t)|, which represents the distance between the current user behavior and the searched user behavior position, X*(t) represents the best position vector so far, b is a constant used to define the shape of the spiral, and d is a random number in (-1,1). D = |CXrand-X(t)| (Formula 6), X(t+1)=Xrand-AD (Formula 7), Among them, Xrand is the randomly selected user behavior position vector, and the algorithm is set to randomly select a search agent when A≥1, set the number of user behavior explorations to N2, set the user behavior exploration boundary to the three-month online and offline behavior records of the data storage module, set the number of iterations to T, and finally search to obtain the optimal feature subset as N3 user behavior sets, and convert the N3 user behavior sets into corresponding N5 advertising delivery sets again, for delivery on the N split screens of the terminal advertising machine; The first delivery method: the terminal advertising machine plays N5 advertisements in full screen polling, and the N5 advertisements are estimated by the whale algorithm; the conditions for adopting this delivery method are: the online shopping characteristics are obvious in the past N4 days; The second delivery method: the terminal advertising machine plays N5 advertisements in split screens, and the N5 advertisements are the advertisements that best match the user behavior, estimated by the whale algorithm; the conditions for using this delivery method: the online shopping characteristics are not obvious in the past N4 days; The third delivery method: the terminal advertising machine plays N5 advertisements in split screens, and the N5 advertisements are novel, not the most popular, and are not usually seen by users, and are estimated by the whale algorithm; the conditions for using this delivery method: the online shopping characteristics are not obvious in the past N4 days, and the online shopping characteristics are obvious in the past 2N4 days; The first delivery method, the second delivery method and the third delivery method are only for the case where there is only one user in front of the terminal advertising machine; The fourth delivery method: the terminal advertising machine plays N5 advertisements in split screens, the N5 advertisements correspond to N5 users, each user corresponds to one advertisement, the optimal advertisement for each user is estimated by the whale algorithm, and then the N5 advertisements are delivered in combination; The fourth delivery method is for the first N5 users of the terminal advertising machine; The four delivery methods are presented in the terminal advertising machine in the following way: each advertisement has a QR code, which is presented in a high-brightness strobe manner. The QR code is used to link to the online APP shopping address corresponding to the product. Users only need to scan the QR code with their mobile phones to purchase the product.

7. The advertisement delivery and user behavior interaction system according to claim 5, characterized in that: The advertising effect feedback and display module (6) comprises an online advertising cost unit (61), an online user feedback unit (62), an online user behavior statistics unit (63), an offline advertising cost unit (64), an offline user feedback unit (65), an offline user behavior statistics unit (66), an analysis result evaluation unit (67) and a display unit (68); The online advertising cost unit (61) is used to comprehensively calculate the online advertising investment cost in a statistical time period; the online user feedback unit (62) is used to comprehensively calculate the user's online feedback on the advertisement in a statistical time period, including at least the number of times the user browses the advertisement, the user's stay time on the advertisement, and the user's comments on the advertisement; The online user behavior statistics unit (63) is used to comprehensively count the purchasing power of users online for the advertisements placed during a certain period of time, including whether the users purchased the products related to the advertisements, the number of times the users purchased the products related to the advertisements, and the comments after purchasing the products related to the advertisements; The offline advertising cost unit (64) is used to comprehensively calculate the offline advertising investment cost in a statistical time period; the offline user feedback unit (65) is used to comprehensively calculate the user's offline feedback on the advertisement in a statistical time period, at least including the user's stay time on the advertisement and the user's comments on the advertisement; The offline user behavior statistics unit (66) is used to comprehensively count the purchase power of users in the offline period corresponding to the advertisements placed, including whether the users purchased the products related to the advertisements, the number of times the users purchased the products related to the advertisements, and the comments after purchasing the products related to the advertisements; The analysis result evaluation unit (67) is used to calculate the relationship between user behavior and advertising cost. The evaluation result index Yxcl is obtained by calculating the following formula: Yxcl=Sa*e+Sb*f+Sc*g+Sd*h+E (Formula 8), In the formula, Sa represents the online advertising cost coefficient, Sb represents the user's online comprehensive performance coefficient, Sc represents the offline advertising cost coefficient, Sd represents the user's offline comprehensive performance coefficient, e represents the proportional coefficient of the online advertising cost coefficient Sa, f represents the proportional coefficient of the user's online comprehensive performance coefficient Sb, g represents the proportional coefficient of the offline advertising cost coefficient Sc, and h represents the proportional coefficient of the user's offline comprehensive performance coefficient Sd; Among them, 0.24≤e≤0.38, 0.23≤f≤0.31, 0.19≤g≤0.34, 0.16≤h≤0.28 and e+f+g+h≤1, E represents the correction constant, the larger the Yxcl value, the better the effect of advertising, otherwise it is not good; The display unit (68) is used to display the user's online and offline behavior data or the result data of the analysis result evaluation unit (67) in real time in a graphical interface for the advertiser to make further decisions.

8. The advertisement delivery and user behavior interaction system according to claim 6, characterized in that: The advertising delivery strategy optimization and effect tracking module (7) comprises: an online and offline user integration optimization unit (71), an online and offline user behavior weight value optimization unit (72), an offline advertising delivery content optimization unit (73) and an effect tracking unit (74); The online and offline user fusion optimization unit (71) is used to perform deep optimization and fusion of the online and offline user portraits, use the cosine similarity algorithm to perform continuous fusion optimization, find the intersection of online and offline behaviors, further integrate the online behavior portrait with the offline behavior portrait, and perform more targeted and accurate advertising delivery; the online and offline user behavior weight value optimization unit (72) is used to perform a comprehensive analysis of the current user behavior, adjust the weight value of the user's online or offline behavior in real time, and optimize the advertising delivery strategy for online and offline conversion; The offline advertising content optimization unit (73) is used to perform in-depth analysis based on the fusion portrait, diversify the offline display content, and adjust the search parameters of the whale algorithm to bias the adjustment parameters towards the novelty of the user's advertisement; the novelty means that the advertisement has not been delivered within a time period or the user has not been delivered the advertisement within a time period; The effect tracking unit (74) is used to track the effect of the new advertising delivery strategy, and calculates it through the analysis result evaluation unit (67). If it fails to meet expectations, an improved whale algorithm is used to deliver advertisements. The improved whale algorithm uses a global search method and can search for adhesion between users.

9. The advertisement delivery and user behavior interaction system according to claim 1, characterized in that: The optimization of the strategy at least includes the integration of online and offline users, the diversification of offline advertising display content, the novelty of offline advertising display content and the adjustment of online and offline user behavior weights to achieve accurate and efficient advertising delivery.

Images