Information push method and system applied to intelligent marketing service platform

Abstract

The embodiment of the present application provides an information push method and system applied to an intelligent marketing service platform, which obtains the potential audience groups and their corresponding group feature labels of the target audience objects in multiple marketing interest levels, uses the feature knowledge network to obtain the audience knowledge vectors of each level, and performs weighted fusion to generate a comprehensive target vector. Further combined with the advertising preference feature library, correlation integration is performed to form advertising push features, and finally a personalized advertising information push strategy is determined, thereby effectively improving the accuracy of advertising push and the response rate of the audience, and realizing the efficiency and personalization of intelligent marketing services.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and more specifically, to an information push method and system applied to an intelligent marketing service platform. Background Art

[0002] With the rapid development of the Internet and big data technology, intelligent marketing services have gradually become an important means of market promotion. In order to achieve efficient advertising and accurate reach, information push methods for target audiences have become a research hotspot. However, traditional advertising push methods often ignore the personalized differences of audiences at different levels of marketing interests, resulting in push content that is inconsistent with audience interests, which in turn affects the conversion rate of advertising and audience satisfaction.

[0003] In the existing advertising push technology, most methods only push based on the basic attributes of the audience (such as age, gender, etc.) or simple interest classification. This extensive method is difficult to accurately capture the audience's deep-level interests and needs, making advertising push lack of targeting and refinement. At the same time, due to the lack of comprehensive analysis of the audience at multiple marketing interest levels, traditional methods cannot effectively explore and utilize the correlation between the interests and needs of the audience at different levels, further restricting the improvement of advertising push effects. Summary of the invention

[0004] In order to at least overcome the above-mentioned deficiencies in the prior art, the purpose of this application is to provide an information push method and system applied to an intelligent marketing service platform, which realizes the all-round and accurate capture of audience interests by deeply analyzing the potential audience groups and group feature labels of the target audience in multiple marketing interest levels, combining feature knowledge networks and audience knowledge vectors. At the same time, through technical means such as weighted fusion and relevance integration, the audience's advertising preference characteristics and comprehensive target vectors at all levels are effectively integrated, thereby generating highly personalized advertising push features and strategies, which not only improves the accuracy of advertising push and the audience's response rate, but also greatly improves the overall effect and efficiency of intelligent marketing services.

[0005] In a first aspect, the present application provides an information push method applied to an intelligent marketing service platform, which is applied to an Internet marketing service system, and the method comprises:

[0006] Obtain the target audience object in the marketing advertisement to be pushed, determine the potential audience group associated with the target audience object in K marketing interest levels, and obtain the group feature labels corresponding to the potential audience groups in each marketing interest level; K is a positive integer;

[0007] Based on the group feature labels corresponding to the potential audience groups in the i-th marketing interest level, in the feature knowledge network corresponding to the i-th marketing interest level, an audience knowledge vector corresponding to the potential audience groups in the i-th marketing interest level is obtained; the feature knowledge network corresponding to the i-th marketing interest level includes target vectors corresponding to audience effective response nodes of each audience group in the i-th marketing interest level, where i is a positive integer not greater than K;

[0008] Performing weighted fusion on the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level to generate a comprehensive target vector of the target audience object in the i-th marketing interest level;

[0009] Acquire an advertising preference feature library corresponding to the target audience object, perform correlation integration on the advertising preference feature library and the comprehensive target vectors of the target audience object at each marketing interest level, generate advertising push features corresponding to the target audience object, and determine the advertising information push strategy corresponding to the target audience object based on the advertising push features.

[0010] In a possible implementation of the first aspect, acquiring a target audience object in a marketing advertisement to be pushed, determining potential audience groups associated with the target audience object in K marketing interest levels, and acquiring group feature labels corresponding to the potential audience groups in each marketing interest level include:

[0011] Obtaining audience portrait attributes corresponding to the target audience object in the marketing advertisement to be pushed, and determining the potential audience group associated with the target audience object in each of the K marketing interest levels based on the audience portrait attributes and the interest knowledge links corresponding to the audience effective response nodes of each audience group in the K marketing interest levels;

[0012] Classify the interest knowledge links corresponding to the potential audience groups in each marketing interest level into group characteristic labels, and generate group characteristic labels corresponding to the potential audience groups in each marketing interest level;

[0013] The step of determining the potential audience group associated with the target audience object in each of the K marketing interest levels based on the audience portrait attributes and the interest knowledge links corresponding to the audience effective response nodes of each audience group in the K marketing interest levels includes:

[0014] Matching the audience portrait attributes with the interest knowledge links corresponding to the audience effective response nodes of each audience group in the K marketing interest levels, and calculating the Euclidean distance between the audience portrait attributes of the target audience object and the interest knowledge links of each audience group;

[0015] According to the calculated Euclidean distance, an audience group closest to the target audience object is determined in each of the K marketing interest levels as a potential audience group, and the potential audience group is associated with the target audience object to establish a corresponding mapping relationship.

[0016] In a possible implementation of the first aspect, classifying the interest knowledge links corresponding to the potential audience groups in each marketing interest level to generate group characteristic labels corresponding to the potential audience groups in each marketing interest level includes:

[0017] Obtain the interest knowledge link corresponding to the audience effective response node x of the potential audience group in the i-th marketing interest level, and obtain the model label feature distribution corresponding to M group feature models; M is a positive integer;

[0018] The M interest feature vectors included in the interest knowledge link corresponding to the audience effective response node x are fused with the model label feature distributions corresponding to the M group feature models to generate fused label features of the audience effective response node x on the M group feature models;

[0019] Perform group feature label classification on the fused label features of the audience effective response node x on the M group feature models to generate a group feature label classification result, obtain the classification measurement parameters of the feature knowledge network corresponding to the i-th marketing interest level, and output the screening result of the classification measurement parameters with the group feature label classification result as the group feature label corresponding to the audience effective response node x.

[0020] In a possible implementation of the first aspect, the method further includes:

[0021] Constructing K marketing interest levels for the marketing advertisement to be pushed, obtaining the minimum audience response heat value and the maximum audience response heat value in the K marketing interest levels, determining the audience response growth coefficient based on the minimum audience response heat value and the maximum audience response heat value, and rounding down the product of the minimum audience response heat value and the i-th power of the audience response growth coefficient to generate the audience group size corresponding to the i-th marketing interest level;

[0022] Based on the audience group sizes corresponding to the K marketing interest levels, the marketing advertisement to be pushed is divided into K audience groups to generate multiple audience groups in each marketing interest level;

[0023] Classifying the interest knowledge links corresponding to the multiple audience groups in the i-th marketing interest level by group characteristic labels, and generating group characteristic labels corresponding to the multiple audience groups in the i-th marketing interest level;

[0024] Obtain target vectors corresponding to audience effective response nodes of multiple audience groups in the i-th marketing interest level, edit group feature labels and target vectors corresponding to audience effective response nodes of multiple audience groups in the i-th marketing interest level into a knowledge graph structure, and generate a feature knowledge network corresponding to the i-th marketing interest level.

[0025] In a possible implementation of the first aspect, obtaining target vectors corresponding to audience effective response nodes of multiple audience groups in the i-th marketing interest level includes:

[0026] Obtaining the consumption behaviors, browsing histories, and search preferences corresponding to the audience effective response nodes c of the multiple audience groups in the i-th marketing interest level, and outputting the consumption behaviors, browsing histories, and search preferences corresponding to the audience effective response nodes c as audience behavior features of the audience effective response nodes c;

[0027] The advertisement click rate and conversion rate corresponding to the audience effective response node c are output as the advertisement effect characteristics of the audience effective response node c, and the audience behavior characteristics and the advertisement effect characteristics are output as the target vector corresponding to the audience effective response node c.

[0028] In a possible implementation of the first aspect, weighted fusion of audience knowledge vectors corresponding to potential audience groups in the i-th marketing interest level to generate a comprehensive target vector of the target audience object in the i-th marketing interest level includes:

[0029] Based on the audience portrait attributes corresponding to the target audience object and the group attention portrait attributes corresponding to the potential audience group in the i-th marketing interest level, M attention portrait attributes of the target audience object in the potential audience group in the i-th marketing interest level are obtained; M is a positive integer;

[0030] Based on the M attention portrait attributes, the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level are weightedly fused to generate a comprehensive target vector of the target audience object in the i-th marketing interest level.

[0031] In a possible implementation of the first aspect, weighted fusion of audience knowledge vectors corresponding to potential audience groups in the i-th marketing interest level to generate a comprehensive target vector of the target audience object in the i-th marketing interest level includes:

[0032] Performing regularized conversion on the group attention portrait attributes corresponding to the potential audience group in the i-th marketing interest level to generate regularized portrait attributes corresponding to the potential audience group in the i-th marketing interest level;

[0033] Based on the attention network and the regularized portrait attributes corresponding to the potential audience group in the i-th marketing interest level, the attention coefficient of the target audience object on M group feature models is obtained; M is a positive integer;

[0034] Based on the attention coefficients on the M group feature models, the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level are weightedly fused to generate a comprehensive target vector of the target audience object in the i-th marketing interest level.

[0035] In a possible implementation of the first aspect, acquiring an advertisement preference feature library corresponding to the target audience object, integrating the advertisement preference feature library and the comprehensive target vectors of the target audience object at each marketing interest level, and generating advertisement push features corresponding to the target audience object includes:

[0036] Outputting the consumption behavior, browsing history, search preference, advertisement click-through rate and conversion rate corresponding to the target audience object into an advertisement preference feature library corresponding to the target audience object;

[0037] According to the order in which the audience response heat values ​​arrive from small to large, the comprehensive target vectors of the target audience objects in each marketing interest level are relevance integrated to generate relevance integration features, and the relevance integration features are fused with the advertising preference feature library to output as advertising push features corresponding to the target audience objects.

[0038] In a possible implementation of the first aspect, determining the advertising information push strategy corresponding to the target audience object based on the advertising push feature includes:

[0039] The advertising push feature is loaded into the advertising push decision network, and based on the decision functions corresponding to the K advertising push decision branches in the advertising push decision network, a strategy is defined for the advertising push feature to generate an advertising information push strategy corresponding to the target audience object; K is a positive integer.

[0040] In a second aspect, an embodiment of the present application further provides an Internet marketing service system, comprising a processor and a machine-readable storage medium, wherein the machine-readable storage medium stores a computer program, and the computer program is loaded and executed according to the processor to implement the information push method applied to the intelligent marketing service platform of the above first aspect.

[0041] Based on any one of the above aspects, the beneficial effects of this application are as follows:

[0042] By obtaining the target audience objects in the marketing advertisements to be pushed and determining the associated potential audience groups, in-depth mining and analysis of marketing interests at different levels are achieved, which helps to better understand the needs and preferences of the target audience, thereby improving the accuracy and effectiveness of advertising. Based on the characteristic knowledge network of each marketing interest level, the corresponding audience knowledge vector is obtained, and weighted fusion is performed to generate a comprehensive target vector, which can convert complex user information into numerical data that is easy to process and analyze, facilitating further calculation and processing. Combined with the advertising preference feature library of the target audience objects, the advertising push features are generated, and the advertising information push strategy is determined based on this, which can make the advertising push more in line with the user's personal preferences, enhance the pertinence and attractiveness of the advertising, and thus improve the click-through rate and conversion rate of the advertising. As a result, the accuracy and effectiveness of the advertising are improved, the advertising push is made more in line with the user's personal preferences, and the click-through rate and conversion rate of the advertising are improved.

[0043] That is, the embodiment of the present application obtains the potential audience groups and their corresponding group feature labels of the target audience objects in multiple marketing interest levels, uses the feature knowledge network to obtain the audience knowledge vectors of each level, and performs weighted fusion to generate a comprehensive target vector. Further combined with the advertising preference feature library, correlation integration is performed to form advertising push features, and finally a personalized advertising information push strategy is determined, thereby effectively improving the accuracy of advertising push and the response rate of the audience, and realizing the efficiency and personalization of intelligent marketing services. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 A flowchart of an information push method applied to an intelligent marketing service platform provided in an embodiment of the present application;

[0045] Figure 2 A block diagram showing the functional structure of an Internet marketing service system for implementing the above-mentioned information push method applied to an intelligent marketing service platform provided in an embodiment of the present application. DETAILED DESCRIPTION

[0046] See also Figure 1 As shown, the present application provides an information push method applied to an intelligent marketing service platform, comprising the following steps.

[0047] Step S110, obtaining the target audience object in the marketing advertisement to be pushed, determining the potential audience groups associated with the target audience object in K marketing interest levels, and obtaining group feature labels corresponding to the potential audience groups in each marketing interest level. K is a positive integer.

[0048] The marketing advertisements to be pushed may refer to marketing advertisements that are planned to be pushed to specific audiences. These advertisements usually contain product, service or brand information, aiming to attract the audience's attention and promote purchasing behavior. For example, a travel platform plans to push a summer vacation package advertisement, which contains information such as specific vacation destinations, price discounts and itinerary arrangements.

[0049] The target audience may refer to the main target of the advertisement, that is, the specific group or individual that the advertiser hopes to reach and influence through the advertisement information. For example, in the advertisement of the above-mentioned travel platform, the target audience may be employed persons aged between 25 and 45 who have a hobby of travel and a certain economic foundation.

[0050] The K marketing interest levels may refer to multiple levels that divide the audience according to their interest in different marketing content. These marketing interest levels can help advertisers target the audience more accurately and push advertising content that matches their interests. K represents the number of levels and is a positive integer. For example, assuming K=3, the audience can be divided into three levels: a high interest level (such as users who frequently browse travel websites and often participate in travel activities), a medium interest level (users who occasionally pay attention to travel information and have travel experience), and a low interest level (users who rarely pay attention to travel information and have almost no travel experience). Alternatively, it may also refer to dividing the levels based on the audience's interest in different types of content, with each marketing interest level representing a specific interest or preference.

[0051] The potential audience groups may refer to groups with similar interests and behavioral characteristics as the target audience in each marketing interest level. These groups are the main candidates for advertising push. For example, in the high interest level, the potential audience groups may be those users who frequently browse travel websites, participate in travel forum discussions, and frequently book travel packages.

[0052] The group characteristic labels may refer to labels used to describe common characteristics or behaviors of potential audience groups. These group characteristic labels can help advertisers gain a deeper understanding of the audience and optimize advertising content and push strategies. For example, for potential audience groups with high interest levels, group characteristic labels may include "frequent travelers", "travel enthusiasts", "high-end travel consumers", etc. These labels can be obtained through data analysis, user behavior tracking, market research, etc.

[0053] When implementing the above steps, the Internet marketing service system usually first collects and analyzes a large amount of user data, including browsing history, purchase history, social media interactions, etc., to identify the target audience. Then, the Internet marketing service system can use machine learning algorithms or manually set rules to divide these target audiences into different marketing interest levels. Next, the Internet marketing service system can further analyze the common characteristics and behavior patterns of potential audience groups in each level and label them with corresponding group characteristics. These labels and hierarchical information will serve as an important basis for the formulation of subsequent advertising push strategies.

[0054] Therefore, in this embodiment, the Internet marketing service system first receives a marketing advertisement request for a marketing advertisement to be pushed, and the marketing advertisement request contains basic information about the advertising content and the target audience object. Based on this information, the Internet marketing service system determines the association of the target audience object in K marketing interest levels through data analysis. For example, when the levels are divided based on the audience's interest in different types of content, and each marketing interest level represents a specific interest or preference, assuming K=3, the Internet marketing service system determines that the target audience object A is associated with three marketing interest levels L1, L2, and L3. In the L1 level, A is associated with the potential audience group of "travel enthusiasts"; in the L2 level, A is associated with "healthy lifestyle pursuers"; in the L3 level, A is associated with "technology product enthusiasts".

[0055] The Internet marketing service system further obtains group characteristic labels corresponding to these potential audience groups. For "travel enthusiasts", group characteristic labels may include "like outdoor activities", "frequently browse travel websites", etc.; for "healthy lifestyle pursuers", characteristic labels may include "pay attention to healthy diet", "regularly participate in physical exercise", etc.; for "technology product enthusiasts", characteristic labels may include "pay attention to the release of new technology products", "frequently purchase electronic products", etc.

[0056] Step S120, based on the group feature labels corresponding to the potential audience groups in the i-th marketing interest level, in the feature knowledge network corresponding to the i-th marketing interest level, obtain the audience knowledge vector corresponding to the potential audience groups in the i-th marketing interest level. The feature knowledge network corresponding to the i-th marketing interest level includes the target vector corresponding to the audience effective response node of each audience group in the i-th marketing interest level, where i is a positive integer not greater than K.

[0057] The group characteristic labels may refer to labels that describe the common characteristics or interest preferences of a specific audience group. These group characteristic labels are usually obtained through technical means such as data analysis, market research, and user profiling. For example, in a marketing advertisement for fitness equipment, for people who like outdoor sports, the group characteristic labels may be "outdoor sports enthusiasts", "long-distance runners", "mountaineering enthusiasts", etc.

[0058] The feature knowledge network is a network structure composed of nodes and connections, where nodes represent the concept of knowledge or features, and connections represent the relationship or mutual influence between these concepts. In the field of marketing, the feature knowledge network can represent knowledge about audience characteristics, interests, behaviors, etc. and the associations between them. Taking tourism as an example, the nodes in the feature knowledge network may include "tourist attractions", "travel budget", "travel season", "travel mode" (such as self-driving, group travel), etc., and the connections indicate the relationship between these nodes. For example, the connection between "tourist attractions" and "travel season" may indicate that certain attractions are more popular in certain seasons.

[0059] The audience knowledge vector is a multi-dimensional vector that describes the knowledge or characteristic status of a specific audience group in a certain marketing interest level. Each dimension can represent a characteristic or interest point, and the value of the vector indicates the strength or preference of the audience group on that characteristic or interest point. Taking tourism as an example, the audience knowledge vector may include dimensions such as "interest in natural scenery", "interest in cultural history", "budget level", and "travel frequency". Each dimension has a corresponding value, for example, the value of "interest in natural scenery" may be high, medium, or low.

[0060] In the feature knowledge network, audience effective response nodes are those nodes that are closely related to the characteristics or interests of a specific audience group and show positive responses in past behaviors or feedback. These nodes are the key basis for determining audience knowledge vectors. Continuing with the example of tourism, if the audience group shows a high interest in "beach vacation" and has a high click-through rate or conversion rate in past beach vacation ads, then "beach vacation" can be considered an audience effective response node for this audience group.

[0061] The target vector is a multidimensional vector associated with an audience effective response node, which quantifies the characteristics or interest strength of the audience group at that node. This can be used to accurately define the audience's needs and preferences. Taking "beach vacation" as an example, the target vector may include dimensions such as "preference for beach type" (such as fine sand, rocky coast), "holiday activity preference" (such as surfing, diving), and "accommodation preference" (such as luxury resorts, economical hotels), and each dimension has a quantified value.

[0062] In practical applications, the Internet marketing service system can build a feature knowledge network based on the collected user data and marketing knowledge base, which includes identifying the group feature labels of the audience group, determining the audience effective response nodes, and generating corresponding target vectors for each node. When it is necessary to determine the target audience for a specific marketing advertisement push, the Internet marketing service system can query the nodes that match the audience group feature labels in the feature knowledge network and extract the relevant audience knowledge vectors. These audience knowledge vectors are then used for personalized customization of advertising content and formulation of push strategies.

[0063] Therefore, in this embodiment, the Internet marketing service system obtains audience knowledge vectors in the corresponding characteristic knowledge network for the potential audience groups in each marketing interest level. Taking "travel enthusiasts" in the L1 level as an example, the Internet marketing service system searches for audience effective response nodes that match the characteristic labels of the "travel enthusiasts" group in the characteristic knowledge network of the marketing interest level, and obtains the target vectors corresponding to these audience effective response nodes. These target vectors may include information in multiple dimensions such as "travel destination preference", "travel budget", and "travel frequency".

[0064] Step S130, performing weighted fusion on the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level to generate a comprehensive target vector of the target audience object in the i-th marketing interest level.

[0065] The Internet marketing service system performs weighted fusion on the audience knowledge vectors corresponding to the potential audience groups in each marketing interest level. Taking the L1 level as an example, the Internet marketing service system assigns different weights to the audience knowledge vectors of "travel enthusiasts" according to their importance and relevance, and then performs fusion calculations to generate a comprehensive target vector for the target audience object A in this level. This comprehensive target vector may be a multi-dimensional vector that reflects A's comprehensive preferences and needs in tourism. The Internet marketing service system also performs similar weighted fusion operations on the L2 and L3 levels to generate A's comprehensive target vectors in these two levels.

[0066] For example, suppose that in the i-th marketing interest level, there are two audience knowledge vectors A and B, which respectively represent the characteristics of two different aspects of the target audience object in this level. Through weighted fusion, these two vectors can be combined into a comprehensive target vector C, where each dimension value of C is the weighted sum of the corresponding dimension values ​​of A and B. The weighting coefficient can be set according to the actual situation to reflect the importance of different characteristics.

[0067] Step S140, obtaining an advertising preference feature library corresponding to the target audience object, performing correlation integration on the advertising preference feature library and the comprehensive target vectors of the target audience object in each marketing interest level, generating advertising push features corresponding to the target audience object, and determining an advertising information push strategy corresponding to the target audience object based on the advertising push features.

[0068] The advertising preference feature library refers to a database that stores the advertising preference features of target audiences. These features may include information on audience preferences for advertising content, form, style, etc., and are an important basis for formulating personalized advertising push strategies. For example, the advertising preference feature library may contain data such as the click-through rate, viewing time, and purchase conversion rate of target audiences for different types of advertisements. These data can reflect the audience's preference for different advertisements, thereby helping advertisers to more accurately understand audience needs and optimize advertising content and push strategies.

[0069] Relevance integration refers to the association analysis of the information in the advertising preference feature library with the comprehensive target vectors of the target audience at each marketing interest level to generate advertising push features corresponding to the target audience. This process aims to discover the intrinsic connection between audience characteristics and advertising preferences, and provide a basis for formulating personalized advertising push strategies. Through relevance integration, it can be found that the comprehensive target vector of the target audience at a certain marketing interest level is highly correlated with certain features in the advertising preference feature library. For example, if the comprehensive target vector shows that the audience has a high interest in travel content, and the data in the advertising preference feature library also shows that the audience has a high click-through rate for travel ads, then it can be determined that travel ads are the preferred push content for this audience.

[0070] The advertising push features refer to the personalized advertising push features corresponding to the target audience objects generated according to the relevance integration results. These features comprehensively consider factors such as the audience's interests, preferences, and historical behaviors, and are the basis for formulating advertising information push strategies. For example, advertising push features may include the target audience's acceptance of different types of advertisements, the advertising content they expect to see, and the most suitable advertising push time. These advertising push features can help advertisers more accurately target audience needs and improve advertising delivery effects and conversion rates.

[0071] The advertising information push strategy refers to a personalized advertising push plan for the target audience object formulated according to the advertising push characteristics, thereby executing the advertising information push task based on the advertising information push strategy. The advertising information push strategy aims to ensure that the advertising content is highly matched with the audience's needs and improve the attractiveness and effectiveness of the advertisement. For example, based on the advertising push characteristics, a series of specific advertising information push strategies can be generated, such as for the audience group of travel enthusiasts, push the preferential activity information of relevant tourist destinations before the arrival of the tourist season; for the audience group of technology product enthusiasts, timely push the detailed introduction and purchase link of relevant products when new products are released. These advertising information push strategies are designed to meet the audience's needs to the greatest extent, improve advertising effects and user satisfaction.

[0072] Based on the above steps, by obtaining the target audience objects in the marketing advertisements to be pushed and determining the associated potential audience groups, in-depth mining and analysis of marketing interests at different levels are achieved, which helps to better understand the needs and preferences of the target audiences, thereby improving the accuracy and effectiveness of the advertisements. Based on the characteristic knowledge network of each marketing interest level, the corresponding audience knowledge vector is obtained, and weighted fusion is performed to generate a comprehensive target vector, which can convert complex user information into numerical data that is easy to process and analyze, and is convenient for further calculation and processing. Combined with the advertising preference feature library of the target audience objects, the advertising push features are generated, and the advertising information push strategy is determined based on this, which can make the advertising push more in line with the user's personal preferences, enhance the pertinence and attractiveness of the advertisement, and thus improve the click-through rate and conversion rate of the advertisement. As a result, the accuracy and effectiveness of the advertisement are improved, the advertising push is made more in line with the user's personal preferences, and the click-through rate and conversion rate of the advertisement are improved.

[0073] That is, the embodiment of the present application obtains the potential audience groups and their corresponding group feature labels of the target audience objects in multiple marketing interest levels, uses the feature knowledge network to obtain the audience knowledge vectors of each level, and performs weighted fusion to generate a comprehensive target vector. Further combined with the advertising preference feature library, correlation integration is performed to form advertising push features, and finally a personalized advertising information push strategy is determined, thereby effectively improving the accuracy of advertising push and the response rate of the audience, and realizing the efficiency and personalization of intelligent marketing services.

[0074] In a possible implementation, step S110 may include:

[0075] Step S111, obtaining the audience portrait attributes corresponding to the target audience object in the marketing advertisement to be pushed, and determining the potential audience group associated with the target audience object in each of the K marketing interest levels based on the audience portrait attributes and the interest knowledge links corresponding to the audience effective response nodes of each audience group in the K marketing interest levels.

[0076] In this embodiment, when analyzing the audience portrait attributes (such as age, gender, occupation, interests, etc.) of the target audience object, the Internet marketing service system will match these audience portrait attributes with the audience effective response nodes of each audience group in the pre-constructed K marketing interest levels. The audience effective response nodes are derived based on historical data and user behavior analysis, and can reflect the interest and response level of the audience group to specific marketing content. Through matching, the Internet marketing service system can find the audience group most relevant to the target audience object, thereby determining its associated potential audience group in each marketing interest level.

[0077] For example, the audience portrait attributes can be matched with the interest knowledge links corresponding to the audience effective response nodes of each audience group in the K marketing interest levels, and the Euclidean distance between the audience portrait attributes of the target audience object and the interest knowledge links of each audience group can be calculated. Based on the calculated Euclidean distance, the audience group closest to the target audience object can be determined in each of the K marketing interest levels as the potential audience group, and the potential audience group can be associated with the target audience object to establish a corresponding mapping relationship.

[0078] Step S112, classifying the interest knowledge links corresponding to the potential audience groups in each marketing interest level by group characteristic labels, and generating group characteristic labels corresponding to the potential audience groups in each marketing interest level.

[0079] After determining the potential audience groups associated with the target audience in each marketing interest level, the Internet marketing service system can further analyze the interest knowledge links corresponding to these groups. The interest knowledge link is a path that describes the evolution and association of the interests of the audience group, and it contains various interests and preference information of the audience group. By classifying these interest knowledge links with group characteristic labels, the Internet marketing service system can generate more refined group characteristic labels, which will be used for subsequent advertising content customization and push strategy formulation.

[0080] In a possible implementation, step S112 may include:

[0081] Step S1121, obtaining the interest knowledge link corresponding to the audience effective response node x of the potential audience group in the i-th marketing interest level, and obtaining the model label feature distribution corresponding to M group feature models. M is a positive integer.

[0082] When processing the marketing advertising request of the travel company, the Internet marketing service system has already identified the potential audience groups associated with the target audience in multiple marketing interest levels (such as travel interest level, outdoor sports interest level, etc.). Now, the Internet marketing service system needs to conduct a more in-depth analysis of the potential audience groups in the i-th marketing interest level (assuming it is the "outdoor sports interest level").

[0083] First, the Internet marketing service system can obtain the interest knowledge link corresponding to the audience effective response node x of this potential audience group. This interest knowledge link is a data structure composed of multiple interest feature vectors, which reflects the evolution and correlation of the audience group's interest in outdoor sports. For example, this interest knowledge link may contain the audience group's interest level and preference information for outdoor sports such as hiking, mountaineering, and rock climbing.

[0084] At the same time, the Internet marketing service system will also obtain the model label feature distributions corresponding to the M group feature models. These group feature models are pre-trained machine learning models that can classify and label audience groups based on different feature vectors. Each model corresponds to a specific label feature distribution, which describes the probability of different feature vectors being classified into different labels under the model.

[0085] Step S1122, the M interest feature vectors covered in the interest knowledge link corresponding to the audience effective response node x are merged with the model label feature distributions corresponding to the M group feature models to generate a fused label feature of the audience effective response node x on the M group feature models.

[0086] The Internet marketing service system will then fuse the M interest feature vectors in the interest knowledge link corresponding to the audience's effective response node x with the model label feature distribution corresponding to the M group feature models. This process can be understood as matching and combining the interest features of the audience group with the pre-defined label features.

[0087] Specifically, the Internet marketing service system may use a fusion algorithm (such as weighted average, voting mechanism, etc.) to comprehensively consider the label probability of each interest feature vector under each group feature model, thereby generating a fusion label feature of the audience's effective response node x on M group feature models. This fusion label feature is a label vector that integrates multiple models and feature information, which more comprehensively describes the characteristics and preferences of the audience group.

[0088] Step S1123, performing group characteristic label classification on the fused label features of the audience effective response node x on the M group characteristic models, generating a group characteristic label classification result, obtaining the classification measurement parameters of the characteristic knowledge network corresponding to the i-th marketing interest level, and outputting the screening result of the classification measurement parameters by the group characteristic label classification result as the group characteristic label corresponding to the audience effective response node x.

[0089] After obtaining the fused label features, the Internet marketing service system can classify the group feature labels. This process is usually achieved through a classifier (such as a decision tree, neural network, etc.), which divides the audience groups into different label categories based on the information in the fused label features.

[0090] For example, in the outdoor sports interest level, the classifier may divide the audience into different label categories such as "beginner outdoor enthusiasts", "intermediate outdoor enthusiasts" and "advanced outdoor enthusiasts". These label categories reflect the different levels and preferences of the audience in outdoor sports.

[0091] Finally, the Internet marketing service system can obtain the classification measurement parameters of the feature knowledge network corresponding to the i-th marketing interest level (outdoor sports interest level). These parameters are important indicators for evaluating the performance of the classifier, such as accuracy, recall, etc. The Internet marketing service system can screen and optimize the group feature label classification results based on these parameters.

[0092] Specifically, the Internet marketing service system may sort or filter the classification results according to the value of the classification metric parameter, and only retain those label categories that meet certain standards (such as accuracy above a certain threshold) as the final output. In this way, the Internet marketing service system can output a more accurate and reliable group feature label set for the audience's effective response node x. This group feature label set will be used in the subsequent customization of advertising content and push strategy formulation.

[0093] In a possible implementation, the method further includes:

[0094] Step A110, construct K marketing interest levels for the marketing advertisement to be pushed, obtain the minimum audience response thermal value and the maximum audience response thermal value in the K marketing interest levels, determine the audience response growth coefficient based on the minimum audience response thermal value and the maximum audience response thermal value, round down the product of the minimum audience response thermal value and the i-th power of the audience response growth coefficient, and generate the audience group size corresponding to the i-th marketing interest level.

[0095] In this embodiment, after receiving the marketing advertisement request from the travel company, the Internet marketing service system will first construct K marketing interest levels according to the content of the marketing advertisement and the target audience. These marketing interest levels may be divided based on different tourism themes (such as natural scenery, city travel, history and culture, etc.) or tourism activity types (such as hiking, adventure, vacation, etc.). Each marketing interest level represents a specific audience interest and demand area.

[0096] After constructing K marketing interest levels, the Internet marketing service system can analyze historical data and user behavior to obtain the minimum audience response heat value and the maximum audience response heat value in each marketing interest level. These heat values ​​reflect the audience's interest and response level to different marketing interest levels. The minimum heat value represents the audience's lowest interest level for the level, while the maximum heat value represents the highest interest level.

[0097] Next, the Internet marketing service system can calculate an audience response growth coefficient based on the difference between the minimum audience response heat value and the maximum audience response heat value. This audience response growth coefficient reflects the speed and potential of the audience group's interest growth between different marketing interest levels. If the difference between the maximum heat value and the minimum heat value is large, it means that the audience group has a large space for interest growth at this level, and the corresponding growth coefficient will also be high.

[0098] For each marketing interest level i, the Internet marketing service system can multiply the minimum audience response heat value by the audience response growth coefficient to the power of i, and round down the result to generate the audience group size corresponding to this level. This process can be understood as estimating the number of potential audiences at each level based on the initial interest and growth potential of the audience group.

[0099] Step A120, based on the audience group sizes corresponding to the K marketing interest levels, the marketing advertisement to be pushed is divided into K audience groups to generate multiple audience groups in each marketing interest level.

[0100] After determining the audience size at each marketing interest level, the Internet marketing service system can perform K audience segmentation for push marketing ads based on these sizes. Each segmentation will subdivide the target audience of the ad into more specific audience groups based on the interests and characteristics of the audience group. In this way, each marketing interest level will contain multiple audience groups with similar interests and needs.

[0101] Step A130 , classifying the interest knowledge links corresponding to the multiple audience groups in the i-th marketing interest level into group characteristic labels, and generating group characteristic labels corresponding to the multiple audience groups in the i-th marketing interest level.

[0102] For multiple audience groups in each marketing interest level i, the Internet marketing service system can further analyze their corresponding interest knowledge links and classify group characteristic labels. This process is similar to the previous analysis of potential audience groups, but more detailed and specific. Through classification, the Internet marketing service system can generate more accurate and appropriate group characteristic labels for each audience group, which will be used in subsequent advertising customization and push strategies.

[0103] Step A140, obtain the target vectors corresponding to the audience effective response nodes of the multiple audience groups in the i-th marketing interest level, edit the group feature labels and target vectors corresponding to the audience effective response nodes of the multiple audience groups in the i-th marketing interest level into a knowledge graph structure, and generate a feature knowledge network corresponding to the i-th marketing interest level.

[0104] Finally, the Internet marketing service system can obtain the target vectors corresponding to the audience effective response nodes of multiple audience groups in each marketing interest level i. These target vectors are a collection of important nodes and feature vectors of the audience group in the interest knowledge link, which reflect the core interests and needs of the audience group. The Internet marketing service system can combine these target vectors with the corresponding group feature labels and edit them into a knowledge graph structure to generate a feature knowledge network corresponding to each marketing interest level. This feature knowledge network is a complex structure that contains information on the interests, features, and needs of the audience group, which will provide strong support for subsequent advertising customization and push.

[0105] In a possible implementation, step A140 may include:

[0106] Step A141, obtain the consumption behavior, browsing history and search preference corresponding to the audience effective response node c of multiple audience groups in the i-th marketing interest level, and output the consumption behavior, browsing history and search preference corresponding to the audience effective response node c as the audience behavior characteristics of the audience effective response node c.

[0107] Step A142, output the advertisement click rate and conversion rate corresponding to the audience effective response node c as the advertisement effect characteristics of the audience effective response node c, and output the audience behavior characteristics and the advertisement effect characteristics as the target vector corresponding to the audience effective response node c.

[0108] When the Internet marketing service system processes the i-th marketing interest level (assuming it is the "outdoor sports" level), it will conduct in-depth analysis on multiple audience groups in this level. For each audience effective response node c in each audience group (which can be understood as a specific user or user group), the Internet marketing service system will first obtain its consumption behavior data, such as whether the user frequently purchased outdoor equipment, outdoor travel products, etc. in the past purchase records.

[0109] At the same time, the Internet marketing service system will also obtain the browsing history data of the audience's effective response node c, which may include outdoor sports-related websites, forums, social media pages, etc. that the user frequently visits. In addition, search preferences are also an important source of information. The Internet marketing service system will record and analyze the keywords related to outdoor sports entered by the user in the search engine, search frequency, and search result click behavior.

[0110] After collecting the consumption behavior, browsing history, and search preference data of the audience effective response node c, the Internet marketing service system will integrate and process these data to extract behavioral characteristics that can reflect the user's outdoor sports interests and needs. For example, the Internet marketing service system may find that the user often buys high-end outdoor equipment, frequently browses professional outdoor sports websites, and searches for keywords related to outdoor adventures, mountaineering skills, etc. in search engines. These behavioral characteristics will be output by the Internet marketing service system and saved as the audience behavioral characteristics of the audience effective response node c.

[0111] In addition to the audience behavior characteristics, the Internet marketing service system will also pay attention to the audience's effective response node c's response to the previously pushed outdoor sports related advertisements. Specifically, the Internet marketing service system will record and analyze the user's click-through rate (i.e., the ratio of the number of times the user clicks on the advertisement to the number of times the advertisement is displayed) and conversion rate (i.e., the ratio of users who actually purchase or learn more after clicking on the advertisement). These indicators can directly reflect the effect and appeal of the advertisement to the user. The Internet marketing service system will output and save these data related to the advertising effect as the advertising effect characteristics of the audience's effective response node c.

[0112] Finally, the Internet marketing service system will integrate and encode the audience behavior characteristics and advertising effect characteristics of the audience effective response node c to generate a target vector that can fully describe the user's interests, needs and advertising response in outdoor sports. This target vector is a multi-dimensional data structure that contains information from consumer behavior, browsing history, search preferences to advertising click-through rate and conversion rate. The Internet marketing service system will use this target vector to provide data support for subsequent advertising customization and push strategies.

[0113] In a possible implementation, step S130 may include:

[0114] Step S131, based on the audience portrait attribute corresponding to the target audience object and the group attention portrait attribute corresponding to the potential audience group in the i-th marketing interest level, obtain M attention portrait attributes of the target audience object in the potential audience group in the i-th marketing interest level. M is a positive integer.

[0115] Step S132, based on the M attention portrait attributes, weighted fusion is performed on the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level to generate a comprehensive target vector of the target audience object in the i-th marketing interest level.

[0116] In this embodiment, when processing target audience object A, the Internet marketing service system first obtains the audience portrait attributes of A, which may include age, gender, occupation, income, education level, interests and hobbies, etc. At the same time, the Internet marketing service system also analyzes the group attention portrait attributes of the potential audience group in the i-th marketing interest level (such as the "healthy diet" level), which reflect the themes, brands, products, etc. that the group is concerned about.

[0117] Based on this information, the Internet marketing service system will further analyze the similarities and differences between the target audience object A and the potential audience group in the i-th marketing interest level. By comparing A's audience portrait attributes with the group's attention portrait attributes, the Internet marketing service system can identify M attention portrait attributes that A may pay special attention to in the group. These attention portrait attributes may be topics, brands or products that A and the group pay attention to, or they may be A's unique personal interests or needs.

[0118] For example, if the target audience A is a 30-year-old urban white-collar woman who is interested in healthy eating, and the potential audience groups in the i-th marketing interest level are mainly concerned with organic food, vegetarianism, and fitness meals. By comparison, the Internet marketing service system may find that A is particularly concerned about the two topics of organic food and fitness meals, and therefore use these two topics as the attention profile attributes of A in this group.

[0119] After determining the M attention profile attributes of the target audience object A in the i-th marketing interest level, the Internet marketing service system will further analyze the audience knowledge vector corresponding to these attributes. The audience knowledge vector is a multi-dimensional data structure that contains information about potential audience groups in terms of consumption behavior, browsing history, search preferences, advertising response, etc.

[0120] The Internet marketing service system will perform weighted fusion of these audience knowledge vectors based on the attention profile attributes of A. Specifically, the Internet marketing service system will assign higher weights to audience knowledge vectors related to topics, brands, or products that A pays special attention to, and lower weights to content that A pays little attention to or is irrelevant. Through weighted fusion, the Internet marketing service system can generate a comprehensive target vector that integrates A's multifaceted information and concerns at the i-th marketing interest level.

[0121] This comprehensive target vector not only reflects the common characteristics and interests of A and the potential audience at this level, but also highlights the unique concerns and needs of A. The Internet marketing service system will use this comprehensive target vector to provide accurate data support for subsequent advertising customization and push strategies.

[0122] In a possible implementation, step S130 may include:

[0123] Step S133, performing regularized conversion on the group attention portrait attributes corresponding to the potential audience group in the i-th marketing interest level to generate regularized portrait attributes corresponding to the potential audience group in the i-th marketing interest level.

[0124] Step S134, based on the attention network and the regularized portrait attributes corresponding to the potential audience group in the i-th marketing interest level, the attention coefficient of the target audience object on M group feature models is obtained, where M is a positive integer.

[0125] Step S135, based on the attention coefficients on the M group feature models, weighted fusion is performed on the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level to generate a comprehensive target vector of the target audience object in the i-th marketing interest level.

[0126] In this embodiment, when the Internet marketing service system processes the i-th marketing interest level (such as the "travel adventure" level), it first identifies the potential audience groups in this level. These groups may have common concerns and needs, which are reflected in their group concern profile attributes, such as common travel destinations, adventure activity types, travel equipment brands, etc.

[0127] In order to ensure the consistency and effectiveness of these portrait attributes in subsequent processing, the Internet marketing service system will perform regular conversion. Specifically, the Internet marketing service system may convert different text descriptions (such as "Self-driving Tour in Tibet", "On the way to Tibet", "Exploring the Snowy Plateau", etc.) into a standard label form (such as "Travel to Tibet"). At the same time, for numerical attributes (such as age, income, etc.), the Internet marketing service system will perform normalization processing so that they fall within the same scale range.

[0128] After the regularized transformation, the Internet marketing service system generates regularized portrait attributes corresponding to the potential audience group in the i-th marketing interest level, providing standardized input for subsequent attention coefficient calculation and weighted fusion.

[0129] Next, the Internet Marketing Service System uses a pre-trained attention network to process these regularized portrait attributes. The attention network is a deep learning model that can simulate the human attention mechanism and assign different attention weights according to the importance and relevance of the input information.

[0130] In this scenario, the input of the attention network is the regularized image attributes of the potential audience groups in the i-th marketing interest level, and the output is the attention coefficients of the target audience objects on these group feature models. These coefficients reflect the degree of attention of the target audience objects to different group features.

[0131] For example, if the target audience is a middle-aged person who loves outdoor adventures, then his attention coefficient on the group characteristic model related to outdoor adventures may be high, while the attention coefficient on other less relevant group characteristic models will be low.

[0132] After obtaining the attention coefficients of the target audience objects on the M group feature models, the Internet marketing service system will perform a weighted fusion operation. Specifically, the Internet marketing service system will adjust the weights of different dimensions in the audience knowledge vector corresponding to the potential audience group according to these attention coefficients.

[0133] The audience knowledge vector contains characteristic information of potential audience groups in multiple aspects (such as consumption behavior, browsing history, search preferences, etc.). Through weighted fusion, the Internet marketing service system can organically combine the personal interests and needs of the target audience with these group characteristics to generate a comprehensive target vector that integrates multiple aspects of information.

[0134] This comprehensive target vector not only reflects the overall interests and needs of the target audience in the i-th marketing interest level, but also highlights his personalized attention to the characteristics of different groups. The Internet marketing service system will use this comprehensive target vector to provide accurate data support for subsequent advertising customization and push strategies.

[0135] In a possible implementation, step S140 may include:

[0136] Step S141, outputting the consumption behavior, browsing history, search preference, advertisement click rate and conversion rate corresponding to the target audience object into an advertisement preference feature library corresponding to the target audience object.

[0137] Step S142, based on the order in which the audience response heat values ​​arrive from small to large, the comprehensive target vectors of the target audience objects in each marketing interest level are relevance integrated to generate relevance integration features, and the relevance integration features are merged with the advertising preference feature library to output as advertising push features corresponding to the target audience objects.

[0138] In this embodiment, when the Internet marketing service system processes the target audience object B, it will first collect B's consumption behavior data, such as purchase records, purchase frequency, purchase commodity categories, etc. These data can reflect B's consumption habits and preferences. At the same time, the Internet marketing service system will also record B's browsing history, including the websites visited, the pages stayed, the browsing time, etc., to understand B's interests and information acquisition habits.

[0139] In addition, the Internet marketing service system will also analyze B's search preferences, such as search keywords, search frequency, clicked search results, etc. These data can reveal B's needs and intentions. Ad click-through rate and conversion rate are also important indicators that the Internet marketing service system pays attention to. They can directly reflect B's acceptance of advertisements and willingness to buy.

[0140] The Internet marketing service system will integrate and process the above data and output it as an advertising preference feature library corresponding to the target audience B. This feature library contains B's characteristic information in terms of consumption behavior, browsing history, search preferences, advertising click-through rate and conversion rate, etc., providing a rich data foundation for subsequent correlation integration and advertising push.

[0141] When the Internet marketing service system processes the target audience object B, it will calculate the audience response heat value based on B's comprehensive target vector in each marketing interest level. This heat value reflects B's attention and response intensity to different marketing interest levels.

[0142] The Internet marketing service system will integrate the comprehensive target vectors of B in each marketing interest level in the order of audience response heat value from small to large. Specifically, the Internet marketing service system will analyze the correlation and mutual influence between these comprehensive target vectors, such as which levels of interests promote each other, which levels of interests inhibit each other, etc.

[0143] Through relevance integration, the Internet marketing service system can generate a relevance integration feature that integrates B's multi-faceted information and concerns at each level of marketing interest. This feature not only takes into account B's independent interests at each level, but also takes into account the relevance and mutual influence between different levels.

[0144] Finally, the Internet marketing service system will fuse the relevance integration features of the target audience object B with the advertising preference feature library. The fusion method can be simple splicing or complex feature crossover or feature transformation.

[0145] The result of the fusion is the advertising push feature corresponding to the target audience B. This feature integrates B's consumption behavior, browsing history, search preferences, advertising click-through rate and conversion rate, as well as various marketing interest levels. The information and focus points provide comprehensive and accurate data support for the subsequent advertising customization and push of the Internet marketing service system.

[0146] The Internet marketing service system can customize personalized advertising content for B based on this advertising push feature and select the appropriate push timing and channel to improve advertising acceptance and conversion rate.

[0147] In a possible implementation, step S140 may further include:

[0148] Step S143, loading the advertisement push feature into the advertisement push decision network, defining a strategy for the advertisement push feature based on the decision functions corresponding to K advertisement push decision branches in the advertisement push decision network, and generating an advertisement information push strategy corresponding to the target audience object. K is a positive integer.

[0149] In this embodiment, when processing the advertising push strategy of target audience object C, the Internet marketing service system first loads C's advertising push features into a pre-built advertising push decision network. This decision network is a complex deep learning model that contains multiple advertising push decision branches, each of which corresponds to a specific decision function.

[0150] These decision functions are designed based on a large amount of historical data and expert knowledge. They can predict and define the optimal advertising push strategy based on different advertising push characteristics. For example, a decision function may predict the target audience’s willingness to buy a certain type of product based on their consumption behavior and browsing history, and thus define the corresponding advertising push strategy.

[0151] When the Internet marketing service system inputs C's advertising push features into the advertising push decision network, these features will be simultaneously passed to each decision branch. Each decision branch will process and analyze the features according to its own decision function and generate a preliminary advertising push strategy recommendation.

[0152] Then, the Internet marketing service system will combine the suggestions of various decision-making branches and generate a final advertising information push strategy through certain integration and optimization algorithms. This strategy will define in detail the content, form, push timing, push channel and other aspects of the advertisement to ensure that the advertisement can attract the attention of the target audience to the greatest extent and guide them to make purchases.

[0153] During this process, the Internet marketing service system will continuously optimize and adjust the parameters and decision functions in the advertising push decision network based on real-time feedback and historical data to improve the accuracy and effectiveness of advertising push.

[0154] In summary, by loading advertising push features into the advertising push decision network and defining strategies based on decision functions corresponding to multiple decision branches in the advertising push decision network, the Internet marketing service system can generate personalized and accurate advertising information push strategies for target audiences.

[0155] Figure 2 An Internet marketing service system 100 is provided in an embodiment of the present application, including a processor 1001 and a memory 1003 and a program code stored in the memory 1003, and the processor 1001 executes the above program code to implement the steps of the information push method applied to the intelligent marketing service platform.

[0156] Figure 2 The Internet marketing service system 100 shown includes: a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the Internet marketing service system 100 may also include a transceiver 1004, which can be used for data interaction between the Internet marketing service system 100 and other Internet marketing service systems 100, such as data transmission and / or data reception. It should be noted that the transceiver 1004 is not limited to one in actual scheduling, and the structure of the Internet marketing service system 100 does not constitute a limitation on the embodiments of the present application.

[0157] Processor 1001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0158] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 2 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0159] The memory 1003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage medium, other magnetic storage devices, or any other medium that can be used to carry or store program code and can be read by a computer, without limitation herein.

[0160] The memory 1003 is used to store program codes for executing the embodiments of the present application, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the program codes stored in the memory 1003 to implement the steps shown in the above method embodiments.

[0161] An embodiment of the present application provides a computer-readable storage medium having program code stored thereon. When the program code is executed by a processor, the steps and corresponding contents of the aforementioned method embodiment can be implemented.

[0162] The above is only an optional implementation method for some implementation scenarios of the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical concept of the scheme of the present application, other similar implementation methods based on the technical ideas of the present application also belong to the protection scope of the embodiments of the present application.

Claims

1. An information push method applied to an intelligent marketing service platform, characterized in that: The method comprises: Obtain the target audience object in the marketing advertisement to be pushed, determine the potential audience group associated with the target audience object in K marketing interest levels, and obtain the group feature labels corresponding to the potential audience groups in each marketing interest level; K is a positive integer; Based on the group characteristic labels corresponding to the potential audience groups in the ith marketing interest level, in the characteristic knowledge network corresponding to the ith marketing interest level, an audience knowledge vector corresponding to the potential audience groups in the ith marketing interest level is obtained; the characteristic knowledge network corresponding to the ith marketing interest level includes a target vector corresponding to the audience effective response nodes of each audience group in the ith marketing interest level, i is a positive integer not greater than K, the audience knowledge vector is a multi-dimensional vector, describing the knowledge or characteristic state of the audience group in a certain marketing interest level, each dimension represents a characteristic or interest point, and the value of the vector represents the strength or preference of the audience group on the characteristic or interest point, and the audience effective response node refers to a node that is closely related to the characteristics or interests of the audience group and shows a positive response in past behavior or feedback; Performing weighted fusion on the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level to generate a comprehensive target vector of the target audience object in the i-th marketing interest level; Acquire an advertising preference feature library corresponding to the target audience object, perform correlation integration on the advertising preference feature library and the comprehensive target vectors of the target audience object at each marketing interest level, generate advertising push features corresponding to the target audience object, and determine the advertising information push strategy corresponding to the target audience object based on the advertising push features.

2. The information push method applied to the intelligent marketing service platform according to claim 1, characterized in that: The step of obtaining the target audience object in the marketing advertisement to be pushed, determining the potential audience groups associated with the target audience object in K marketing interest levels, and obtaining group feature labels corresponding to the potential audience groups in each marketing interest level includes: Obtaining audience portrait attributes corresponding to the target audience object in the marketing advertisement to be pushed, and determining the potential audience group associated with the target audience object in each of the K marketing interest levels based on the audience portrait attributes and the interest knowledge links corresponding to the audience effective response nodes of each audience group in the K marketing interest levels; Classify the interest knowledge links corresponding to the potential audience groups in each marketing interest level into group characteristic labels, and generate group characteristic labels corresponding to the potential audience groups in each marketing interest level; The step of determining the potential audience group associated with the target audience object in each of the K marketing interest levels based on the audience portrait attributes and the interest knowledge links corresponding to the audience effective response nodes of each audience group in the K marketing interest levels includes: Matching the audience portrait attributes with the interest knowledge links corresponding to the audience effective response nodes of each audience group in the K marketing interest levels, and calculating the Euclidean distance between the audience portrait attributes of the target audience object and the interest knowledge links of each audience group; According to the calculated Euclidean distance, an audience group closest to the target audience object is determined in each of the K marketing interest levels as a potential audience group, and the potential audience group is associated with the target audience object to establish a corresponding mapping relationship.

3. The information push method applied to the intelligent marketing service platform according to claim 2, characterized in that: The group characteristic label classification of the interest knowledge links corresponding to the potential audience groups in each marketing interest level to generate group characteristic labels corresponding to the potential audience groups in each marketing interest level includes: Obtain the interest knowledge link corresponding to the audience effective response node x of the potential audience group in the i-th marketing interest level, and obtain the model label feature distribution corresponding to M group feature models; M is a positive integer; The M interest feature vectors included in the interest knowledge link corresponding to the audience effective response node x are fused with the model label feature distributions corresponding to the M group feature models to generate fused label features of the audience effective response node x on the M group feature models; Perform group feature label classification on the fused label features of the audience effective response node x on the M group feature models to generate a group feature label classification result, obtain the classification measurement parameters of the feature knowledge network corresponding to the i-th marketing interest level, and output the screening result of the classification measurement parameters with the group feature label classification result as the group feature label corresponding to the audience effective response node x.

4. The information push method applied to the intelligent marketing service platform according to claim 1, characterized in that: The method further comprises: Constructing K marketing interest levels for the marketing advertisement to be pushed, obtaining the minimum audience response heat value and the maximum audience response heat value in the K marketing interest levels, determining the audience response growth coefficient based on the minimum audience response heat value and the maximum audience response heat value, and rounding down the product of the minimum audience response heat value and the i-th power of the audience response growth coefficient to generate the audience group size corresponding to the i-th marketing interest level; Based on the audience group sizes corresponding to the K marketing interest levels, the marketing advertisement to be pushed is divided into K audience groups to generate multiple audience groups in each marketing interest level; Classifying the interest knowledge links corresponding to the multiple audience groups in the i-th marketing interest level by group characteristic labels, and generating group characteristic labels corresponding to the multiple audience groups in the i-th marketing interest level; Obtain target vectors corresponding to audience effective response nodes of multiple audience groups in the i-th marketing interest level, edit group feature labels and target vectors corresponding to audience effective response nodes of multiple audience groups in the i-th marketing interest level into a knowledge graph structure, and generate a feature knowledge network corresponding to the i-th marketing interest level.

5. The information push method applied to the intelligent marketing service platform according to claim 4, characterized in that: The step of obtaining target vectors corresponding to audience effective response nodes of multiple audience groups in the i-th marketing interest level includes: Obtaining the consumption behaviors, browsing histories, and search preferences corresponding to the audience effective response nodes c of the multiple audience groups in the i-th marketing interest level, and outputting the consumption behaviors, browsing histories, and search preferences corresponding to the audience effective response nodes c as audience behavior features of the audience effective response nodes c; The advertisement click rate and conversion rate corresponding to the audience effective response node c are output as the advertisement effect characteristics of the audience effective response node c, and the audience behavior characteristics and the advertisement effect characteristics are output as the target vector corresponding to the audience effective response node c.

6. The information push method applied to the intelligent marketing service platform according to claim 1, characterized in that: The step of weighting and fusing the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level to generate a comprehensive target vector of the target audience object in the i-th marketing interest level includes: Based on the audience portrait attributes corresponding to the target audience object and the group attention portrait attributes corresponding to the potential audience group in the i-th marketing interest level, M attention portrait attributes of the target audience object in the potential audience group in the i-th marketing interest level are obtained; M is a positive integer; Based on the M attention portrait attributes, the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level are weightedly fused to generate a comprehensive target vector of the target audience object in the i-th marketing interest level.

7. The information push method applied to the intelligent marketing service platform according to claim 1, characterized in that: The step of weighting and fusing the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level to generate a comprehensive target vector of the target audience object in the i-th marketing interest level includes: Performing regularized conversion on the group attention portrait attributes corresponding to the potential audience group in the i-th marketing interest level to generate regularized portrait attributes corresponding to the potential audience group in the i-th marketing interest level; Based on the attention network and the regularized portrait attributes corresponding to the potential audience group in the i-th marketing interest level, the attention coefficient of the target audience object on M group feature models is obtained; M is a positive integer; Based on the attention coefficients on the M group feature models, the audience knowledge vectors corresponding to the potential audience groups in the i-th marketing interest level are weightedly fused to generate a comprehensive target vector of the target audience object in the i-th marketing interest level.

8. The information push method applied to the intelligent marketing service platform according to claim 1, characterized in that: The step of obtaining the advertisement preference feature library corresponding to the target audience object, integrating the advertisement preference feature library and the comprehensive target vectors of the target audience object at each marketing interest level, and generating advertisement push features corresponding to the target audience object includes: Outputting the consumption behavior, browsing history, search preference, advertisement click-through rate and conversion rate corresponding to the target audience object into an advertisement preference feature library corresponding to the target audience object; According to the order of audience response heat value from small to large, the comprehensive target vectors of the target audience objects in each marketing interest level are relevance integrated to generate relevance integration features, and the relevance integration features and the advertising preference feature library are fused and output as advertising push features corresponding to the target audience objects.

9. The information push method applied to the intelligent marketing service platform according to claim 1, characterized in that: The determining of the advertising information push strategy corresponding to the target audience object based on the advertising push characteristics includes: The advertising push feature is loaded into the advertising push decision network, and based on the decision functions corresponding to K advertising push decision branches in the advertising push decision network, a strategy is defined for the advertising push feature to generate an advertising information push strategy corresponding to the target audience object; K is a positive integer.

10. An Internet marketing service system, characterized in that: The Internet marketing service system includes a processor and a machine-readable storage medium, in which machine-executable instructions are stored. The machine-executable instructions are loaded and executed by the processor to implement the information push method applied to the intelligent marketing service platform as described in any one of claims 1-9.

Images