An APP advertisement accurate delivery system based on user behavior portrait

Abstract

This invention provides a precise APP advertising delivery system based on user behavior profiles, relating to the field of internet advertising service technology. It includes a data acquisition module for acquiring user behavior data, candidate ad data, and APP ad placement scenario data generated by target users within a target APP. The user behavior data includes browsing behavior, click behavior, dwell time, search behavior, ad closing behavior, or ad skipping behavior. This precise APP advertising delivery system based on user behavior profiles enables the ad matching process to simultaneously consider user interests, ad content attributes, and APP ad placement display scenarios, improving the targeting and accuracy of the ad delivery data processing.

Description

Technical Field

[0001] This invention relates to the field of internet advertising service technology, specifically to an app advertising precision delivery system based on user behavior profiles. Background Technology

[0002] Existing app advertising systems typically display ads to users through the app's splash screen, news feed, pop-ups, content detail pages, or other ad placements. They select ads from a pool of candidates based on user attributes, device information, geographic location, browsing history, or pre-defined audience tags. Some systems also combine advertiser-defined criteria such as ad category, timing, target audience, or ad placement type to filter and display ads, achieving targeted mobile internet advertising.

[0003] However, existing technologies still have the following shortcomings: existing APP advertising delivery systems lack a mechanism to match user behavior profiles with advertising tags and APP advertising placement scenarios to generate target advertising delivery results, resulting in insufficient accuracy of advertising delivery. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a precise APP advertising delivery system based on user behavior profiles. By modifying user behavior profiles for specific advertising scenarios, it achieves coordinated matching between user profiles, advertising tags, and APP advertising scenarios, thus solving the problem of insufficient accuracy in APP advertising delivery.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a precise advertising delivery system for apps based on user behavior profiles, comprising:

[0006] The data acquisition module is used to acquire user behavior data, candidate ad data, and APP ad placement scenario data generated by the target user within the target APP. The user behavior data includes one or more of the following: browsing behavior, clicking behavior, dwell time, search behavior, ad closing behavior, or ad skipping behavior. The candidate ad data includes the ad tags of the candidate ads. The APP ad placement scenario data includes the ad placement type of the target ad placement, the page scenario to which the ad placement belongs, and the ad display trigger scenario.

[0007] The profile building module is used to extract behavioral features from the user behavior data and generate a user behavior profile of the target user based on the behavior type, behavior occurrence time, behavior frequency and behavior duration. The user behavior profile includes multiple user interest tags and an initial profile weight corresponding to each user interest tag. The initial profile weight is used to characterize the intensity of the target user's interest in the advertisements corresponding to the user interest tags.

[0008] The advertising scenario processing module is used to analyze the advertising features of the candidate advertising data, generate advertising tag features corresponding to each candidate advertising, and generate advertising location scenario features based on the APP advertising location scenario data. The advertising tag features are used to characterize the advertising content category, target audience and display format of the candidate advertising, and the advertising location scenario features are used to characterize the page position, advertising triggering method and advertising display format of the target advertising location in the target APP.

[0009] The contextual matching module is used to determine the contextual correction coefficient corresponding to each user interest tag based on the contextual features of the ad slot, and to use the contextual correction coefficient to correct the initial profile weight to obtain a contextual user behavior profile; and to match the contextual user behavior profile with the ad tag features of each candidate ad to calculate the placement fit value of each candidate ad relative to the target user and the target ad slot.

[0010] The delivery result generation module is used to determine the target advertisement from the candidate advertisements according to the delivery adaptation value, and generate the target advertisement delivery result for the target advertisement position. The target advertisement delivery result includes the target advertisement identifier, the target advertisement position identifier, and the display priority of the target advertisement in the target advertisement position.

[0011] Preferably, the data acquisition module is further configured to organize the user behavior data into user behavior records. The user behavior records include target user identifier, behavior type, APP content object identifier, APP content object tag, behavior occurrence time, behavior duration, page scene identifier, and advertising feedback identifier. The APP content object includes APP page content, advertising materials, or search results items. The advertising feedback identifier is used to mark whether the user behavior record corresponds to an ad click, ad closing, or ad skipping.

[0012] Preferably, the profile building module is further configured to determine the user interest tag corresponding to each user behavior record based on the mapping table between the APP content object tag and the preset interest tag, and to determine the corresponding behavior type coefficient from the preset behavior type coefficient table based on the behavior type. The browsing behavior, clicking behavior, searching behavior and content dwell behavior correspond to positive behavior type coefficients, and the ad closing behavior and ad skipping behavior correspond to negative behavior type coefficients.

[0013] Preferably, the profile building module is further configured to calculate the behavior contribution value corresponding to each user behavior record based on the behavior type coefficient, behavior occurrence time, behavior frequency, and behavior duration, and to accumulate multiple behavior contribution values ​​corresponding to the same user interest tag to obtain the initial profile weight corresponding to the user interest tag. The behavior occurrence time is used to determine the time decay coefficient, the behavior duration is used to determine the duration contribution coefficient, and the behavior frequency is used to determine the frequency contribution coefficient.

[0014] Preferably, the ad feature parsing includes extracting ad content category tags, ad target audience tags, and ad creative format tags from the ad tags of candidate ads, and generating ad tag features based on the ad content category tags, ad target audience tags, and ad creative format tags. The ad tag features are represented by ad tag vectors, which are vectorized representations of the ad tag features. Each dimension of the ad tag vector corresponds to a different ad tag and its tag weight.

[0015] Preferably, the generation of the ad placement scene features includes performing scene analysis on the APP ad placement scene data to determine ad placement type tags, page scene tags, display trigger tags, and ad placement display format tags from the APP ad placement scene data, and generating the ad placement scene features based on the ad placement type tags, page scene tags, display trigger tags, and ad placement display format tags. The ad placement type tags are used to characterize the target ad placement as a splash screen ad placement, a news feed ad placement, a pop-up ad placement, a search results page ad placement, or a content details page ad placement. The display trigger tags are used to characterize the target ad placement as triggered by APP launch, page switching, search behavior, content browsing, or user dwell.

[0016] Preferably, the determination of the scene correction coefficient includes assigning a coefficient value based on the scene association strength between the ad placement scene features and each user interest tag. The scene association strength is obtained by weighting the association values ​​between the ad placement type tag, page scene tag, display trigger tag and each user interest tag. When the scene association strength is greater than a first association threshold, the scene correction coefficient of the corresponding user interest tag is greater than 1; when the scene association strength is less than a second association threshold, the scene correction coefficient of the corresponding user interest tag is less than 1; when the scene association strength is not less than the second association threshold and not greater than the first association threshold, the scene correction coefficient of the corresponding user interest tag is equal to 1.

[0017] Preferably, the generation of the contextualized user behavior profile includes calculating the contextualized profile weight based on the initial profile weight and the contextual correction coefficient corresponding to each user interest tag, and generating the contextualized user behavior profile based on multiple user interest tags and their corresponding contextualized profile weights. The contextualized profile weight corresponding to any user interest tag is the normalized weight value of the product of the initial profile weight corresponding to the user interest tag and the contextual correction coefficient.

[0018] Preferably, the calculation of the placement adaptation value includes determining the profile ad matching value based on the tag matching result between the contextualized user behavior profile and the ad tag features of each candidate ad, determining the ad placement scene matching value based on the preset ad placement adaptation relationship between the ad tag features and the ad placement scene features, and calculating the placement adaptation value by weighting the profile ad matching value and the ad placement scene matching value. The tag matching result includes the matching result between user interest tags and ad content category tags and ad target audience tags. The preset ad placement adaptation relationship includes the adaptation relationship between ad material format tags and ad placement display format tags.

[0019] Preferably, the generation of the target ad delivery result includes identifying candidate ads with a delivery fit value greater than or equal to a preset delivery threshold as target ads, determining the display priority of the target ads in the target ad slot according to the delivery fit value from high to low, and when the number of target ads is greater than the number of ads that can be displayed in the target ad slot, selecting the target ads with higher display priority to generate the target ad delivery result.

[0020] This invention provides a precise advertising system for apps based on user behavior profiles. It offers the following advantages:

[0021] This app advertising precision targeting system based on user behavior profiles acquires user behavior data generated by target users within the target app and generates user behavior profiles by combining behavior type, occurrence time, frequency, and duration. This allows advertising to move beyond relying solely on basic user attributes or fixed audience tags, instead determining user interest tags and initial profile weights based on actual user interactions within the app. Simultaneously, this technical solution performs ad feature analysis on candidate ads to form ad tag features and performs scene analysis on app ad placement data to form ad placement scene features. This enables the ad matching process to simultaneously consider user interests, ad content attributes, and the app ad placement display scene, improving the targeting and accuracy of ad delivery data processing.

[0022] This invention determines the scene correction coefficient corresponding to each user's interest tag by identifying the scene characteristics of the ad placement. This coefficient is then used to adjust the initial profile weights, generating a scene-based user behavior profile. Finally, based on the scene-based user behavior profile and the ad tag features of candidate ads, a placement fit value is calculated, resulting in a target ad placement result that includes the target ad identifier, the target ad placement identifier, and the display priority. Therefore, the same target user can generate different ad matching results in different app ad placement scenarios, avoiding the problem of insufficient ad placement scene adaptability caused by fixed user profiles, reducing invalid exposures caused by mismatch between ad content and the user's current interests, and improving the accuracy of app ad placement. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the system structure of the present invention;

[0024] Figure 2 This is a flowchart illustrating the precise ad targeting method for apps according to the present invention.

[0025] Figure 3 This is a flowchart illustrating the process of generating advertising tag features and advertising location scene features according to the present invention. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Example 1

[0028] like Figure 1 As shown, this embodiment of the invention provides an APP advertising precision delivery system based on user behavior profiles, including a data acquisition module, used to acquire user behavior data, candidate ad data, and APP ad placement scenario data generated by target users within the target APP. The user behavior data includes one or more of browsing behavior, click behavior, dwell time, search behavior, ad closing behavior, or ad skipping behavior. The candidate ad data includes the ad tags of candidate ads. The APP ad placement scenario data includes the ad placement type of the target ad placement, the page scenario to which the ad placement belongs, and the ad display trigger scenario.

[0029] The data acquisition module is also used to organize user behavior data into user behavior records. User behavior records include target user identifier, behavior type, APP content object identifier, APP content object tag, behavior occurrence time, behavior duration, page scene identifier, and advertising feedback identifier. APP content objects include APP page content, advertising materials, or search results items. Advertising feedback identifiers are used to mark whether user behavior records correspond to ad clicks, ad closing, or ad skipping.

[0030] The profile building module is used to extract behavioral features from user behavior data and generate user behavior profiles of target users based on behavior type, time of occurrence, frequency, and duration. The user behavior profile includes multiple user interest tags and initial profile weights corresponding to each user interest tag. The initial profile weights are used to characterize the intensity of the target user's interest in the advertisements corresponding to the user interest tags.

[0031] The profile building module is also used to determine the user interest tags corresponding to each user behavior record based on the mapping table between APP content object tags and preset interest tags, and to determine the corresponding behavior type coefficients from the preset behavior type coefficient table according to the behavior type. Browsing behavior, clicking behavior, searching behavior, and content dwell behavior correspond to positive behavior type coefficients, while ad closing behavior and ad skipping behavior correspond to negative behavior type coefficients.

[0032] The profile building module is also used to calculate the behavior contribution value corresponding to each user behavior record based on the behavior type coefficient, behavior occurrence time, behavior frequency, and behavior duration. It also accumulates the multiple behavior contribution values ​​corresponding to the same user interest tag to obtain the initial profile weight corresponding to the user interest tag. The behavior occurrence time is used to determine the time decay coefficient, the behavior duration is used to determine the duration contribution coefficient, and the behavior frequency is used to determine the frequency contribution coefficient.

[0033] The advertising scenario processing module is used to analyze the advertising features of candidate advertising data, generate advertising tag features corresponding to each candidate advertising, and generate advertising location scenario features based on APP advertising location scenario data. The advertising tag features are used to characterize the advertising content category, target audience and display format of candidate advertising, and the advertising location scenario features are used to characterize the page position, advertising triggering method and advertising display format of the target advertising location in the target APP.

[0034] The ad feature parsing process includes extracting ad content category tags, ad target audience tags, and ad creative format tags from the ad tags of candidate ads. Based on these tags, ad label features are generated. The ad label features are represented by ad label vectors, which are vectorized representations of the ad label features. Each dimension in the ad label vector corresponds to a different ad label and its weight.

[0035] The generation of ad placement scene features includes scene analysis of APP ad placement scene data to determine ad placement type tags, page scene tags, display trigger tags, and ad placement display format tags from the APP ad placement scene data. Based on the ad placement type tags, page scene tags, display trigger tags, and ad placement display format tags, ad placement scene features are generated. The ad placement type tag is used to characterize whether the target ad placement is a splash screen ad placement, news feed ad placement, pop-up ad placement, search results page ad placement, or content details page ad placement. The display trigger tag is used to characterize whether the target ad placement is triggered by APP launch, page switching, search behavior, content browsing, or user dwell.

[0036] The contextual matching module is used to determine the contextual correction coefficients corresponding to each user interest tag based on the contextual characteristics of the ad placement, and to use the contextual correction coefficients to correct the initial profile weights to obtain contextual user behavior profiles; and to match the contextual user behavior profiles with the ad tag features of each candidate ad to calculate the placement suitability value of each candidate ad relative to the target user and the target ad placement.

[0037] The determination of the scene correction coefficient includes assigning a coefficient value based on the scene association strength between the ad placement scene characteristics and each user interest tag. The scene association strength is obtained by weighting the association values ​​between the ad placement type tag, page scene tag, display trigger tag and each user interest tag. When the scene association strength is greater than the first association threshold, the scene correction coefficient of the corresponding user interest tag is greater than 1; when the scene association strength is less than the second association threshold, the scene correction coefficient of the corresponding user interest tag is less than 1; when the scene association strength is not less than the second association threshold and not greater than the first association threshold, the scene correction coefficient of the corresponding user interest tag is equal to 1.

[0038] The generation of contextualized user behavior profiles includes calculating contextualized profile weights based on the initial profile weights and contextual correction coefficients corresponding to each user interest tag, and generating contextualized user behavior profiles based on multiple user interest tags and their corresponding contextualized profile weights. The contextualized profile weight corresponding to any user interest tag is the normalized weight value of the product of the initial profile weight corresponding to that user interest tag and the contextual correction coefficient.

[0039] The calculation of the ad placement fit value includes determining the profile ad matching value based on the tag matching results between the contextualized user behavior profile and the ad tag features of each candidate ad, determining the ad placement scenario matching value based on the preset ad placement fit relationship between the ad tag features and the ad placement scenario features, and calculating the ad placement fit value by weighting the profile ad matching value and the ad placement scenario matching value. The tag matching results include the matching results between user interest tags and ad content category tags and ad target audience tags. The preset ad placement fit relationship includes the fit relationship between ad creative format tags and ad placement display format tags.

[0040] The delivery result generation module is used to determine the target ad from the candidate ads based on the delivery adaptation value, and generate the target ad delivery result for the target ad position. The target ad delivery result includes the target ad identifier, the target ad position identifier, and the display priority of the target ad in the target ad position.

[0041] The generation of target ad delivery results includes identifying candidate ads with a delivery fit value greater than or equal to a preset delivery threshold as target ads, determining the display priority of target ads in target ad slots according to their delivery fit values ​​from high to low, and when the number of target ads exceeds the number of ads that can be displayed in a target ad slot, selecting the target ads with higher display priority to generate target ad delivery results.

[0042] Example 2

[0043] like Figure 2 As shown in Example 1, this example uses an in-feed ad slot in a target app as an example to illustrate the process of precise ad delivery based on user behavior profiles. The target user is identified as U001, the target ad slot is identified as FEED-01, the target ad slot is an in-feed ad slot, the page scenario to which the ad slot belongs is a health content page, the ad display is triggered by content browsing, and the ad slot is displayed in the form of an image and text card.

[0044] The data acquisition module obtains user behavior data generated by target user U001 within the past 7 days and organizes it into user behavior records. For example, user behavior records include: the target user repeatedly browsed content on "fitness training" apps, with a cumulative duration of 420 seconds; clicked on "sports and outdoor" ad creatives twice, with a cumulative duration of 65 seconds; searched for "running plans" related search results three times; browsed content on "travel" apps three times, with a cumulative duration of 130 seconds; and skipped "food and beverage" ad creatives four times. The profile building module generates a user behavior profile based on the app content object tags, behavior types, behavior times, behavior frequencies, and behavior durations in the above user behavior records.

[0045] In this embodiment, the user profile building module maps the aforementioned user behavior records to user interest tags according to a preset interest tag mapping table, and calculates the initial profile weight corresponding to each user interest tag. For example, the generated user behavior profile includes user interest tags such as "Fitness & Health," "Sports & Outdoors," "Travel & Travel," "Automotive Information," and "Food & Beverages." Specifically, the initial profile weight for "Fitness & Health" is 0.38, for "Sports & Outdoors" it is 0.30, for "Travel & Travel" it is 0.18, for "Automotive Information" it is 0.09, and for "Food & Beverages" it is 0.05.

[0046] The advertising scenario processing module generates advertising scenario features based on the advertising type tag, page scenario tag, display trigger tag, and advertising display format tag of the target advertising slot FEED-01. Since the target advertising slot is an in-feed advertising slot within the health content page and is triggered by content browsing, the scenario matching module determines that the user interest tags "Fitness & Health" and "Sports & Outdoors" have a high scenario correlation with the current advertising slot scenario and assigns them scenario correction coefficients greater than 1. User interest tags with lower correlation to the current advertising slot scenario, such as "Travel & Travel," "Automotive News," and "Food & Beverages," are assigned scenario correction coefficients no greater than 1. After scenario correction and normalization, a scenario-based user behavior profile is obtained, with the following weights: "Fitness & Health" (0.43), "Sports & Outdoors" (0.31), "Travel & Travel" (0.15), "Automotive News" (0.07), and "Food & Beverages" (0.04).

[0047] The candidate ad data includes ad tags for multiple candidate ads. The ad scenario processing module analyzes the ad features of each candidate ad and generates ad tag features. For example, the candidate ads include ad AD01, ad AD02, and ad AD03. Among them, ad AD01 has the ad content category tag "fitness and health", the ad target audience tag "sports enthusiasts", and the ad creative format tag "image and text card"; ad AD02 has the ad content category tag "travel", and the ad creative format tag "short video"; ad AD03 has the ad content category tag "automotive information", and the ad creative format tag "banner image".

[0048] The contextual matching module matches the contextualized user behavior profile with the ad tag features of each candidate ad, and calculates the placement adaptation value by combining the preset ad placement adaptation relationship between the ad creative format tag and the ad placement display format tag. In this embodiment, the profile ad matching value for ad AD01 is 0.86, the ad placement context matching value is 0.92, and the placement adaptation value is 0.88; the placement adaptation value for ad AD02 is 0.48; and the placement adaptation value for ad AD03 is 0.29. If the preset placement threshold is 0.60, the placement result generation module determines ad AD01 as the target ad and generates the target ad placement result, which includes the target ad identifier AD01, the target ad placement identifier FEED-01, and the display priority 1.

[0049] Through the above process, the system can modify the user behavior profile according to the specific ad placement scenario within the target app when the same target user has multiple interest tags, so that the final target ad is more suitable for the current ad placement display environment and the user's current ad interest tendency.

[0050] Example 3

[0051] like Figure 3 As shown, based on Examples 1 and 2, this example explains the generation process of advertising tag features and advertising space scene features.

[0052] The candidate ad data includes ad tags for multiple candidate ads. The ad scenario processing module performs ad feature analysis on the ad tags of the candidate ads to determine ad content category tags, target audience tags, and ad creative format tags. For example, if the ad tags for candidate ad AD01 include "fitness and health," "sports enthusiasts," and "image and text card," then the ad scenario processing module will determine "fitness and health" as the ad content category tag, "sports enthusiasts" as the target audience tag, and "image and text card" as the ad creative format tag, and generate the ad tag features corresponding to ad AD01 based on these tags.

[0053] Simultaneously, the advertising scenario processing module performs scenario analysis on the APP advertising placement data to determine advertising placement type tags, page scenario tags, display trigger tags, and advertising placement display format tags. For example, if the target advertising placement is an in-feed advertising placement, the page scenario to which the advertising placement belongs is a health content page, the advertising display trigger scenario is content browsing trigger, and the advertising placement display format is an image and text card display format, then the advertising scenario processing module generates corresponding in-feed advertising placement tags, health content page tags, content browsing trigger tags, and image and text card display format tags, and generates advertising placement scenario features corresponding to the target advertising placement based on the above tags.

[0054] In this embodiment, both ad tag features and ad placement scene features are used for subsequent ad placement adaptation value calculation. Ad tag features characterize the ad content attributes and ad creative attributes of candidate ads, while ad placement scene features characterize the display scene and display conditions of the target ad placement. By generating ad tag features and ad placement scene features respectively, the system can determine whether candidate ads simultaneously match the target user's contextualized user behavior profile and the specific ad placement scene within the target app during subsequent matching.

[0055] Example 4

[0056] Based on Examples 1 to 3, this example describes the process of generating scenario-based user behavior profiles.

[0057] The contextual matching module acquires user behavior profiles generated by the profile building module and ad placement scene features generated by the ad scene processing module. The user behavior profile includes multiple user interest tags and the initial profile weights corresponding to each user interest tag. The ad placement scene features include ad placement type tags, page scene tags, display trigger tags, and ad placement display format tags.

[0058] The contextual matching module determines the contextual correction coefficient based on the strength of the contextual association between the ad placement's contextual characteristics and each user's interest tag. For example, when the target ad placement is an in-feed ad placement, the page context is a health content page, and the display trigger is content browsing, user interest tags such as "fitness and health" and "sports and outdoors" have a high contextual association with the current ad placement context, and the corresponding contextual correction coefficient can be set to greater than 1; user interest tags such as "travel" or "automotive information" have a low contextual association with the current ad placement context, and the corresponding contextual correction coefficient can be set to less than or equal to 1.

[0059] The contextual matching module uses the contextual correction coefficients corresponding to each user's interest tags to adjust the initial profile weights, resulting in contextual profile weights. For example, in the target user's user behavior profile, the initial profile weight for "Fitness and Health" is 0.38, for "Sports and Outdoors" it is 0.30, and for "Travel" it is 0.18. In the context of the information flow ad slot on the health content page, after contextual correction, the contextual profile weight for "Fitness and Health" increases to 0.43, for "Sports and Outdoors" it increases to 0.31, and for "Travel" it decreases to 0.15.

[0060] The contextualized matching module generates contextualized user behavior profiles based on multiple user interest tags and corresponding contextualized profile weights. As a result, the same target user can generate different contextualized user behavior profiles in different app ad placement scenarios. This allows the subsequent ad matching process to not only consider the user's historical behavioral interests but also the current app ad placement's page context and display trigger conditions, thereby improving the adaptability of ad matching results to the current delivery scenario.

[0061] Example 5

[0062] Based on Examples 1 to 4, this example explains the process of calculating the placement fit value and generating the target ad placement results.

[0063] The contextual matching module acquires contextual user behavior profiles, ad tag features of candidate ads, and ad placement context features of the target ad placement. Based on the tag matching results between the contextual user behavior profiles and the ad tag features of each candidate ad, it determines the profile ad matching value. For example, if the ad content category tag of candidate ad AD01 is "fitness and health," the ad target audience tag is "sports enthusiasts," and the contextual profile weights corresponding to "fitness and health" and "sports and outdoor activities" are relatively high in the contextual user behavior profile of the target user in the current ad placement context, then the profile ad matching value corresponding to AD01 is relatively high.

[0064] The contextual matching module also determines the contextual matching value of an ad placement based on the preset ad placement adaptation relationship between ad tag features and ad placement context features. For example, if the target ad placement supports the display format of image and text cards, and the ad creative format tag of candidate ad AD01 is "image and text card", then the contextual matching value of the ad placement corresponding to AD01 is high; if the ad creative format tag of candidate ad AD02 is "short video", but the target ad placement does not support the display format of short videos, then the contextual matching value of the ad placement corresponding to AD02 is low.

[0065] Subsequently, the contextual matching module calculates the placement adaptation value for each candidate ad based on the profile ad matching value and the ad placement context matching value. For example, the profile ad matching value for ad AD01 is 0.86, the ad placement context matching value is 0.92, and the placement adaptation value is 0.88; the placement adaptation value for ad AD02 is 0.48; and the placement adaptation value for ad AD03 is 0.29.

[0066] The campaign delivery result generation module identifies candidate ads with a delivery fit value greater than or equal to a preset delivery threshold as target ads. If the preset delivery threshold is 0.60, then ad AD01 is identified as the target ad. The module further determines the display priority of the target ad in the target ad slot based on the delivery fit value and generates the target ad delivery result. The target ad delivery result includes the target ad identifier, the target ad slot identifier, and the display priority; for example, if the target ad identifier is AD01, the target ad slot identifier is FEED-01, and the display priority is 1.

[0067] Through the above process, the system can filter out target ads from the candidate ad set that simultaneously match the user's current ad interest and the target ad placement display scenario, and generate delivery results for the target APP ad placement.

[0068] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A precise advertising delivery system for apps based on user behavior profiles, characterized in that, include: The data acquisition module is used to acquire user behavior data, candidate ad data, and APP ad placement scenario data generated by the target user within the target APP. The user behavior data includes one or more of the following: browsing behavior, clicking behavior, dwell time, search behavior, ad closing behavior, or ad skipping behavior. The candidate ad data includes the ad tags of the candidate ads. The APP ad placement scenario data includes the ad placement type of the target ad placement, the page scenario to which the ad placement belongs, and the ad display trigger scenario. The profile building module is used to extract behavioral features from the user behavior data and generate a user behavior profile of the target user based on the behavior type, behavior occurrence time, behavior frequency and behavior duration. The user behavior profile includes multiple user interest tags and an initial profile weight corresponding to each user interest tag. The initial profile weight is used to characterize the intensity of the target user's interest in the advertisements corresponding to the user interest tags. The advertising scenario processing module is used to analyze the advertising features of the candidate advertising data, generate advertising tag features corresponding to each candidate advertising, and generate advertising location scenario features based on the APP advertising location scenario data. The advertising tag features are used to characterize the advertising content category, target audience and display format of the candidate advertising, and the advertising location scenario features are used to characterize the page position, advertising triggering method and advertising display format of the target advertising location in the target APP. The contextual matching module is used to determine the contextual correction coefficient corresponding to each user interest tag based on the contextual characteristics of the ad slot, and to use the contextual correction coefficient to correct the initial profile weight to obtain a contextualized user behavior profile. And to match the contextualized user behavior profile with the advertising tag features of each candidate advertisement, and calculate the placement fit value of each candidate advertisement relative to the target user and the target advertisement position; The delivery result generation module is used to determine the target advertisement from the candidate advertisements according to the delivery adaptation value, and generate the target advertisement delivery result for the target advertisement position. The target advertisement delivery result includes the target advertisement identifier, the target advertisement position identifier, and the display priority of the target advertisement in the target advertisement position.

2. The APP advertising precision delivery system based on user behavior profiles according to claim 1, characterized in that: The data acquisition module is also used to organize the user behavior data into user behavior records. The user behavior records include target user identifier, behavior type, APP content object identifier, APP content object tag, behavior occurrence time, behavior duration, page scene identifier, and advertising feedback identifier. The APP content object includes APP page content, advertising materials, or search results items. The advertising feedback identifier is used to mark whether the user behavior record corresponds to an ad click, ad closing, or ad skipping.

3. The APP advertising precision delivery system based on user behavior profiles according to claim 2, characterized in that: The profile building module is also used to determine the user interest tag corresponding to each user behavior record based on the mapping table between the APP content object tag and the preset interest tag, and to determine the corresponding behavior type coefficient from the preset behavior type coefficient table based on the behavior type. The browsing behavior, clicking behavior, searching behavior and content dwell behavior correspond to positive behavior type coefficients, and the ad closing behavior and ad skipping behavior correspond to negative behavior type coefficients.

4. The APP advertising precision delivery system based on user behavior profiles according to claim 3, characterized in that: The profile building module is also used to calculate the behavior contribution value corresponding to each user behavior record based on the behavior type coefficient, behavior occurrence time, behavior frequency, and behavior duration, and to accumulate multiple behavior contribution values ​​corresponding to the same user interest tag to obtain the initial profile weight corresponding to the user interest tag. The behavior occurrence time is used to determine the time decay coefficient, the behavior duration is used to determine the duration contribution coefficient, and the behavior frequency is used to determine the frequency contribution coefficient.

5. The APP advertising precision delivery system based on user behavior profiles according to claim 1, characterized in that: The ad feature parsing includes extracting ad content category tags, ad target audience tags, and ad creative format tags from the ad tags of candidate ads, and generating ad tag features based on the ad content category tags, ad target audience tags, and ad creative format tags. The ad tag features are represented by ad tag vectors, which are vectorized representations of the ad tag features. Each dimension in the ad tag vector corresponds to a different ad tag and its tag weight.

6. The APP advertising precision delivery system based on user behavior profiles according to claim 1, characterized in that: The generation of the ad placement scene features includes performing scene analysis on the APP ad placement scene data to determine ad placement type tags, page scene tags, display trigger tags, and ad placement display format tags from the APP ad placement scene data, and generating the ad placement scene features based on the ad placement type tags, page scene tags, display trigger tags, and ad placement display format tags.

7. The APP advertising precision delivery system based on user behavior profiles according to claim 6, characterized in that: The determination of the scene correction coefficient includes assigning a coefficient value based on the scene association strength between the ad placement scene characteristics and each user interest tag. The scene association strength is obtained by weighting the association values ​​between the ad placement type tag, page scene tag, display trigger tag and each user interest tag. When the scene association strength is greater than a first association threshold, the scene correction coefficient of the corresponding user interest tag is greater than 1; when the scene association strength is less than a second association threshold, the scene correction coefficient of the corresponding user interest tag is less than 1. When the scene association strength is not less than the second association threshold and not greater than the first association threshold, the scene correction coefficient for the corresponding user interest tag is equal to 1.

8. The APP advertising precision delivery system based on user behavior profiles according to claim 7, characterized in that: The generation of the contextualized user behavior profile includes calculating the contextualized profile weight based on the initial profile weight and the contextual correction coefficient corresponding to each user interest tag, and generating the contextualized user behavior profile based on multiple user interest tags and their corresponding contextualized profile weights. The contextualized profile weight corresponding to any user interest tag is the normalized weight value of the product of the initial profile weight corresponding to the user interest tag and the contextual correction coefficient.

9. The APP advertising precision delivery system based on user behavior profiles according to claim 8, characterized in that: The calculation of the ad placement adaptation value includes determining the profile ad matching value based on the tag matching result between the contextualized user behavior profile and the ad tag features of each candidate ad, determining the ad placement scene matching value based on the preset ad placement adaptation relationship between the ad tag features and the ad placement scene features, and calculating the ad placement adaptation value by weighting the profile ad matching value and the ad placement scene matching value. The tag matching result includes the matching result between user interest tags and ad content category tags and ad target audience tags. The preset ad placement adaptation relationship includes the adaptation relationship between ad material format tags and ad placement display format tags.

10. The APP advertising precision delivery system based on user behavior profiles according to claim 9, characterized in that: The generation of the target ad delivery result includes identifying candidate ads with a delivery fit value greater than or equal to a preset delivery threshold as target ads, determining the display priority of the target ads in the target ad slot according to the delivery fit value from high to low, and when the number of target ads is greater than the number of ads that can be displayed in the target ad slot, selecting the target ads with higher display priority to generate the target ad delivery result.

Images