A user engagement driven interactive advertisement incentivization method, apparatus, and medium

Abstract

The application discloses a user participation feeling driven interactive advertisement incentive method and device and medium, relates to the technical field of advertisement incentive, and comprises the following steps: performing attenuation trend division and continuous time window determination on an attenuation sequence, generating a trigger window set; performing trigger window level mapping and reward action constraint screening on the trigger window set, and registering the attribution relationship between the trigger window and the reward action to form a reward action pool; performing reward action selection in the reward action pool through an attribution strategy model to generate a reward action sequence; performing advertisement presentation and behavior association processing on the reward action sequence to form a delayed return; performing backtracking attribution assignment and trigger window constraint adjustment on the delayed return, and performing strategy update processing on the attribution strategy model to generate an incentive decision. The application enhances the user participation feeling driven interactive advertisement incentive effect through continuous time sequencing and attention attenuation calculation.

Description

Technical Field

[0001] This invention relates to the field of advertising incentive technology, and in particular to an interactive advertising incentive method, device and medium driven by user engagement. Background Technology

[0002] Interactive advertising incentive technology occupies an important position in the internet advertising technology system. With the development of mobile application platforms and online content platforms, advertising delivery methods are gradually shifting from a one-way display model to an interactive advertising model. Interactive advertising records user interaction data such as dwell time, operation behavior, and response rhythm to form analyzable behavioral sequences. These sequences are then combined with reward actions within the advertising system to provide incentive feedback, thereby enhancing user engagement and advertising effectiveness. With the development of data analysis techniques and machine learning methods, interactive advertising systems are increasingly using behavioral sequence analysis, temporal relationship calculations, and strategy model calculations to characterize user engagement states and dynamically generate advertising incentive decisions based on behavioral changes, driving advertising incentive technology towards a data-driven and strategy-driven direction.

[0003] In the field of interactive advertising incentive technology, existing methods typically trigger reward actions based on the number of clicks, dwell time, or simple behavioral rules. The utilization of information on the temporal relationship and rhythm changes between user interaction behaviors in the incentive decision-making process is limited, and the matching degree between the timing of advertising incentive triggering and the user's participation status is somewhat limited, thus affecting the effect of user engagement modulation. To improve these problems, existing interactive advertising incentive technologies usually adopt behavioral threshold judgment methods or click-through rate statistical models to judge user interaction behavior data and trigger reward actions. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a user-engagement-driven interactive advertising incentive method to address the limitation in the matching degree between advertising incentive triggering timing and user engagement status.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, the present invention provides a user-engagement-driven interactive advertising incentive method, comprising, Acquire user interaction behavior data, sort it over a continuous time period and calculate attention decay value to generate a decay sequence, perform decay status division and continuous time window determination on the decay sequence, and generate a set of trigger windows. Trigger window level mapping and reward action constraint filtering are performed on the trigger window set, and the attribution relationship between trigger windows and reward actions is registered to form a reward action pool. Reward action selection is performed in the reward action pool through the attribution strategy model to generate a reward action sequence. The reward action sequence is processed by ad presentation and behavior association to generate delayed rewards; The system performs backtracking attribution assignment and trigger window constraint adjustment for delayed rewards, updates the attribution strategy model, calculates reward strategies for the trigger window set based on the updated attribution strategy model, and generates incentive decisions.

[0007] As a preferred embodiment of the user-engagement-driven interactive advertising incentive method of the present invention, wherein: the generation of the decay sequence specifically comprises, Get the dwell time, operation behavior and response rhythm in the interactive advertising interface, and write the interactive behavior timestamp and interactive advertising identifier to generate user interactive behavior data; The user interaction data is sorted continuously over time according to the timestamps of the interaction behaviors from earliest to latest, generating an interaction behavior sequence; Read the timestamps of adjacent interactive behaviors one by one from the sequence of interactive behaviors, and form a sequence of adjacent time intervals; Attention decay is calculated for each adjacent time interval sequence to obtain the attention decay value, and the attention decay value is written into the interaction behavior sequence to generate the decay sequence.

[0008] As a preferred embodiment of the user-engagement-driven interactive advertising incentive method of the present invention, wherein: the generation of the trigger window set specifically comprises, Read adjacent attention decay values ​​one by one along the decay sequence, and divide the decay sequence into decay states according to the direction of change between adjacent attention decay values ​​to generate a decay segment sequence; Obtain the slow descent segment and the rapid descent segment from the decay segment sequence, and mark the slow descent segment and the rapid descent segment as decay segment type. At the same time, determine the start time position and end time position of the segment based on the interaction behavior timestamps corresponding to the slow descent segment and the rapid descent segment. The start and end times of a segment are determined using continuous time windows to generate a candidate time window sequence. The start time position, end time position, and decay segment type of the candidate time window sequence are matched with the interactive ad identifier to generate a set of trigger windows.

[0009] As a preferred embodiment of the user-engagement-driven interactive advertising incentive method of the present invention, wherein: a reward action pool is formed, specifically, Perform trigger window level mapping on the attenuation segment types in the trigger window set to obtain the trigger window level; Match the trigger window level with the interactive ad identifier to obtain the set of trigger window reward actions; Perform mutual exclusion constraint verification, frequency constraint verification, and time period constraint verification on the reward actions in the set of reward actions in the trigger window. Retain the reward actions that pass the mutual exclusion constraint verification, frequency constraint verification, and time period constraint verification, and generate a set of executable reward actions. Based on the set of executable reward actions and the trigger window level, establish attribution relationships, write attribution identifiers, and generate an attribution relationship set; The reward actions in the attribution set are aggregated according to the trigger window level to generate a reward action pool.

[0010] As a preferred embodiment of the user-engagement-driven interactive advertising incentive method of the present invention, wherein: the generation of the reward action sequence specifically includes, Reward actions are extracted from the reward action pool, and the trigger window level, interactive ad identifier, and attribution identifier corresponding to the reward action are combined to form an attribution index pair sequence; The attribution index is used to map advertisements to the sequence, generating an advertisement sequence. Based on the trigger window level, perform window candidate positioning on the ad sequence and generate a window index sequence; Perform backlink concatenation on the window index sequence to generate a backlink attribution chain sequence; The target reward action is obtained by performing a ranking decision on the backtracking attribution chain sequence through an attribution strategy model. Write the corresponding trigger window identifier and attribution identifier to the target reward action to generate a reward action sequence.

[0011] As a preferred embodiment of the user-participation-driven interactive advertising incentive method of the present invention, wherein: the advertising presentation refers to the timing binding and presentation of the target reward action in the interactive advertising interface based on the correspondence between the target reward action and the trigger window identifier in the reward action sequence; The behavior association processing involves registering user interaction behavior data and target reward actions according to the trigger window identifier to perform attribution association, thereby generating delayed rewards.

[0012] As a preferred embodiment of the user-engagement-driven interactive advertising incentive method of the present invention, the step of updating the attribution strategy model specifically involves: The delayed reward is back-attributed and expanded. Based on the attribution relationship in the attribution relationship set, the attribution identifier in the delayed reward is traced back and the target reward action is linked together to generate a back-attribution chain sequence. Perform reward attribution determination on the backtracking attribution chain sequence, register the reward identifier in the delayed reward to the target reward action of reward attribution determination, and generate an attribution assignment sequence; Write back the trigger window identifier for the attribution assignment sequence, write the reward identifier in the attribution assignment sequence into the corresponding trigger window identifier in the trigger window set, and generate a reward trigger window set; Adjust the trigger window constraints on the reward trigger window set, mark the trigger window identifiers that continuously appear with the reward identifier as constraint release markers, and mark the trigger window identifiers that do not continuously appear with the reward identifiers as constraint tightening markers, and generate constraint adjustment records; The constraint adjustment records and attribution assignment sequences are aligned and encapsulated, and then written into the attribution strategy model to generate the updated attribution strategy model.

[0013] As a preferred embodiment of the user-engagement-driven interactive advertising incentive method of the present invention, the reward strategy calculation involves inputting the set of trigger windows into the updated attribution strategy model, and the updated attribution strategy model performing reward action matching and ranking adjudication on the set of trigger windows to generate the target reward action.

[0014] In a second aspect, the present invention provides a computer device including a memory and a processor, wherein the memory stores a computer program, wherein: when the computer program is executed by the processor, it implements any step of the user engagement-driven interactive advertising incentive method as described in the first aspect of the present invention.

[0015] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, it implements any step of the user engagement-driven interactive advertising incentive method as described in the first aspect of the present invention.

[0016] The beneficial effects of this invention are as follows: By using continuous time sorting and attention decay calculation, the rhythmic changes in user participation are depicted. Combining trigger window level mapping and reward action attribution relationships, reward actions can be matched and triggered during stages of user participation rhythm changes. Furthermore, by using an attribution strategy model to determine the order of the back-attribution chain, the reward action selection is completed. This allows the interactive advertising incentive process to be dynamically adjusted around changes in user participation behavior, improving the matching ability between the timing of interactive advertising incentives and the rhythm of user participation in advertising technology scenarios, thereby enhancing the user-engagement-driven interactive advertising incentive effect. Attached Figure Description

[0017] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart of an interactive advertising incentive method driven by user engagement.

[0019] Figure 2 A flowchart generated to trigger the window.

[0020] Figure 3 A flowchart generated for the reward action.

[0021] Figure 4 A flowchart for interactive incentive decision-making. Detailed Implementation

[0022] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0023] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0024] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0025] Reference Figures 1-4 As one embodiment of the present invention, this embodiment provides a user-engagement-driven interactive advertising incentive method, comprising the following steps: S1. Obtain user interaction behavior data, perform continuous time sorting and attention decay value calculation, generate decay sequence, perform decay status division and continuous time window determination on decay sequence, and generate trigger window set.

[0026] S1.1. Obtain the dwell time, operation behavior, and response rhythm in the interactive advertising interface, and write the interactive behavior timestamp and interactive advertising identifier to generate user interactive behavior data, specifically, During the operation of the interactive advertising interface, the user's dwell time, operation, and response rhythm are recorded; the interactive behavior timestamp corresponding to each record is obtained, and the interactive advertising identifier is written into the corresponding record; the dwell time, operation, and response rhythm of the interactive behavior timestamp and the interactive advertising identifier are aggregated in chronological order to generate user interactive behavior data.

[0027] S1.2. Sort user interaction data in continuous time order according to the timestamps of the interaction behaviors from earliest to latest, generating an interaction behavior sequence. Specifically, Read the timestamps of user interaction behavior data one by one, and extract the corresponding dwell behaviors, operation behaviors, and response rhythms. Compare the execution time sequence of the read interaction behavior timestamps, and arrange the corresponding dwell behaviors, operation behaviors, and response rhythms in order from earliest to latest interaction behavior timestamps. Collect the dwell behaviors, operation behaviors, and response rhythms arranged in order from earliest to latest interaction behavior timestamps to generate an interaction behavior sequence.

[0028] S1.3 Read the timestamps of adjacent interactive behaviors one by one from the interactive behavior sequence and form an adjacent time interval sequence. Perform attention decay calculation on each adjacent time interval sequence to obtain the attention decay value, and write the attention decay value into the interactive behavior sequence to generate the decay sequence.

[0029] Among them, the first in the adjacent time interval sequence The adjacent time intervals are denoted as The expression is as follows: ; In the formula, The first in the adjacent time interval sequence Adjacent time intervals, The first in the sequence of interactive behaviors Each interactive action corresponds to an interactive action timestamp. The first in the sequence of interactive behaviors Each interactive action corresponds to an interactive action timestamp. This is the sequence index within the interactive behavior sequence. It is the index of the sequence number in the adjacent time interval sequence.

[0030] The expression for attention decay is as follows: ; In the formula, For the first The attention decay value is used to represent the first attention decay value. The degree of attention decay corresponding to each adjacent time interval It is an exponential function. This is a time-scale parameter used to adjust how quickly attention decays with adjacent time intervals, and Units and When the units are consistent, and the timestamps of interactive behaviors are expressed in seconds or milliseconds... The unit is seconds or milliseconds. The corresponding unit is also seconds or milliseconds.

[0031] It should be noted that, in user interaction behavior data, timestamps of adjacent interaction behaviors are extracted to form a sample set of adjacent time intervals. Statistical calculations are performed on the sample set of adjacent time intervals to obtain the median of the time interval distribution, and the median of the time interval distribution is recorded as a time scale parameter. This ensures that the calculation of attention decay values ​​is consistent with the distribution of interaction time intervals; when the distribution of interaction time intervals corresponding to interactive ad icons changes, statistical calculations are re-performed on the adjacent time interval sample sets, and the registered time scale parameters are updated. .

[0032] S1.4. Read adjacent attention decay values ​​one by one along the decay sequence, and divide the decay sequence into decay patterns according to the direction of change between adjacent attention decay values ​​to generate a decay segment sequence. Specifically, The attention decay values ​​are read sequentially from the decay sequence, and adjacent attention decay values ​​are paired to form adjacent attention decay value pairs. For each adjacent attention decay value pair, the difference between the next attention decay value and the previous attention decay value is calculated, and the direction of change between adjacent attention decay values ​​is determined by the sign of the difference. When the difference is positive, the direction of change is registered as the upward direction; when the difference is negative, the direction of change is registered as the downward direction; when the difference is zero, the direction of change is registered as the hold direction. The decay sequence is segmented along the decay sequence according to the continuity of the direction of change, and each segment obtained by the segmentation is registered as a decay segment, generating a decay segment sequence.

[0033] S1.5. Obtain the slow descent segment and the rapid descent segment from the decay segment sequence, and mark the slow descent segment and the rapid descent segment as decay segment types. At the same time, determine the start time position and the end time position of the segment based on the interaction behavior timestamps corresponding to the slow descent segment and the rapid descent segment. Specifically, Read the decay segments one by one from the decay segment sequence, and locate the first and last attention decay values ​​corresponding to the decay segments in the decay sequence. At the same time, read the interaction behavior timestamps corresponding to the first and last attention decay values. Register the interaction behavior timestamps corresponding to the first and last attention decay values ​​as the segment start time position, and register the interaction behavior timestamps corresponding to the last attention decay value as the segment end time position.

[0034] When the direction of change corresponding to the attenuation segment is registered as a downward direction, the absolute value of the attenuation rate of the attenuation segment is calculated. The absolute value of the attenuation rate is the ratio of the absolute value of the difference between the last attention attenuation value and the first attention attenuation value to the time difference between the end time position and the start time position of the segment. The absolute values ​​of the attenuation rates corresponding to all attenuation segments whose direction of change is registered as a downward direction are collected to obtain a set of absolute values ​​of attenuation rates, and the median of the absolute values ​​of attenuation rates is obtained from the set of absolute values ​​of attenuation rates. When the absolute value of the attenuation rate corresponding to the attenuation segment is less than or equal to the median of the absolute values ​​of attenuation rates, the attenuation segment is marked as a slow descent segment and the slow descent segment is marked as an attenuation segment type. When the absolute value of the attenuation rate corresponding to the attenuation segment is greater than the median of the absolute values ​​of attenuation rates, the attenuation segment is marked as a rapid descent segment and the rapid descent segment is marked as an attenuation segment type.

[0035] S1.6. Determine the continuous time window position of the segment's start time position and the segment's end time position, and generate a candidate time window sequence, specifically as follows: Read the start and end times of each segment from the decaying segment sequence one by one, and locate the set of interactive behavior entries whose timestamps are located between the start and end times of the segments in the interactive behavior sequence. When the set of interactive behavior entries is not empty, register the start time of the segment as the start time of the candidate time window and the end time of the segment as the end time of the candidate time window. Gather the registered start and end times of the window in order from earliest to latest according to the start time of the segment to generate the candidate time window sequence.

[0036] S1.7. Match the start time position, end time position, and decay segment type of the candidate time window sequence with the interactive ad identifier to generate a set of trigger windows. Specifically, Select the start and end time positions of the segments in the candidate time window sequence, and locate the decay segment type and interactive ad identifier within the same time range in the decay segment sequence and interactive behavior sequence. Then, register the segment start time position, segment end time position, decay segment type and interactive ad identifier as a set of trigger windows.

[0037] S2. Perform trigger window level mapping and reward action constraint filtering on the trigger window set, and register the attribution relationship between trigger windows and reward actions to form a reward action pool. Then, select reward actions from the reward action pool through the attribution strategy model to generate a reward action sequence.

[0038] S2.1 Perform trigger window level mapping on the attenuation segment types in the trigger window set to obtain the trigger window level, specifically, Read each trigger window from the trigger window set and read the attenuation segment type from the trigger window; when the attenuation segment type is a slow descent segment, register the trigger window level as a low level; when the attenuation segment type is a rapid descent segment, register the trigger window level as a high level; write the registered trigger window level into the corresponding trigger window to obtain the trigger window level.

[0039] S2.2, Match the reward action range between the trigger window level and the interactive ad identifier to obtain the set of reward actions for the trigger window, specifically, Read each trigger window from the trigger window set, and read the trigger window level and interactive ad identifier from each trigger window; obtain the reward action range corresponding to the interactive ad identifier, which includes reward actions associated with the interactive ad identifier, and register the applicable trigger window level for each reward action; filter the reward actions in the reward action range that have the same applicable trigger window level as the trigger window level to obtain a subset of trigger window reward actions; write the subset of trigger window reward actions into the corresponding trigger window, and gather the subsets of trigger window reward actions corresponding to each trigger window in the trigger window set to obtain the set of trigger window reward actions.

[0040] It should be noted that the scope of reward actions is based on the available incentive schemes corresponding to the interactive ad icon. That is, the scope is formed by registering the "set of reward actions that can be issued" associated with the interactive ad icon and the applicable trigger window level, cost limit and placement restrictions of each reward action. The reward actions included are, for example: coupons, points, cash red envelopes, lottery chances, membership time, virtual props, acceleration coupons and unlocking content permissions.

[0041] S2.3. Perform mutual exclusion constraint verification, frequency constraint verification, and time period constraint verification on the reward actions in the reward action set of the trigger window. Retain the reward actions that pass the mutual exclusion constraint verification, frequency constraint verification, and time period constraint verification, and generate an executable reward action set, specifically, Read each reward action from the set of reward actions in the trigger window, and extract mutual exclusion constraints, frequency constraints, and time constraints from the corresponding placement restrictions of the reward actions; based on the mutual exclusion constraints, perform mutual exclusion constraint verification on reward actions within the same trigger window and other reward actions, and record the mutual exclusion constraint verification results; based on the frequency constraints, perform frequency constraint verification on the number of trigger windows with the same interactive ad identifier within the frequency constraint-limited time period in the trigger window set, and record the frequency constraint verification results; based on the time constraints, perform time constraint verification on the start time position and end time position of the segment of the trigger window, and record the time constraint verification results; retain reward actions that are simultaneously registered as passing the mutual exclusion constraint verification, frequency constraint verification, and time constraint verification, and gather the retained reward actions to generate an executable reward action set.

[0042] It should be noted that the mutual exclusion constraint verification is performed based on the mutual exclusion constraint to verify that the relationship between reward actions and other reward actions within the same trigger window cannot be established simultaneously. The mutual exclusion constraint verification includes the verification of the relationship that cannot be established simultaneously and the verification of the consistency of the combination of reward actions.

[0043] Frequency constraint verification is based on frequency constraints and verifies the number of trigger windows corresponding to the interactive ad icon within the frequency constraint-limited time period against the frequency constraint limit. Frequency constraint verification includes time period verification and limit not exceeding relationship verification.

[0044] Time period constraint verification is performed based on time period constraints to verify the start and end time positions of the trigger window segments against the time period constraints. Time period constraint verification includes time period boundary verification and inclusion relationship verification.

[0045] It should also be noted that when a new trigger window is generated in the trigger window set, the reward action set is read from the reward action pool, and the mutual exclusion constraint status, frequency constraint count and time period constraint status are updated according to the reward action execution records registered in the trigger window set. After updating the mutual exclusion constraint status, frequency constraint count, and time period constraint status, the mutual exclusion constraint verification, frequency constraint verification, and time period constraint verification are re-executed for the reward actions in the reward action pool. The reward actions that pass the verification are registered as a set of real-time executable reward actions, thereby ensuring that the reward action selection process is consistent with the real-time status of the trigger window set.

[0046] S2.4. Based on the set of executable reward actions and the trigger window level, establish attribution relationships, write attribution identifiers, and generate an attribution relationship set, specifically as follows: Read each trigger window from the trigger window set, and read the trigger window level and interactive ad identifier from each trigger window; locate the reward actions in the set of executable reward actions that match the trigger window level and the read interactive ad identifier, thus obtaining a subset of executable reward actions for the trigger window; combine the trigger window level and the reward actions in the subset of executable reward actions for the trigger window one by one and register them as attribution relationship entries; write attribution identifiers to the attribution relationship entries, which are generated by combining the interactive ad identifier, the trigger window level, and the order of the reward action in the subset of executable reward actions for the trigger window; collect the registered attribution relationship entries to generate an attribution relationship set.

[0047] S2.5. Perform aggregation processing on the reward actions in the attribution set according to the trigger window level to generate a reward action pool. The aggregation processing is to collect and encapsulate the reward actions in the attribution set according to the trigger window level into reward action sets corresponding to the trigger window level, thereby generating the reward action pool.

[0048] S2.6 Extract reward actions from the reward action pool, and combine the trigger window level, interactive ad identifier, and attribution identifier corresponding to the reward actions to form an attribution index pair sequence; perform ad mapping on the attribution index pair sequence to generate an ad sequence, specifically, The attribution index is expanded line by line in the sequence to obtain the reward action, trigger window level, interactive ad identifier, and attribution identifier. The trigger window with the same interactive ad identifier and trigger window level is located in the trigger window set to obtain the segment start time position, segment end time position, and decay segment type. The reward action, trigger window level, interactive ad identifier, attribution identifier, segment start time position, segment end time position, and decay segment type are encapsulated into ad entries. The ad entries are then aggregated in sequence according to the attribution index to generate the ad sequence.

[0049] S2.7. Based on the trigger window level, perform window candidate positioning on the ad sequence and generate a window index sequence, specifically as follows: Register the trigger window index in the trigger window set according to the segment start time position from earliest to latest; for the ad entries in the ad sequence, use the trigger window level, interactive ad identifier, segment start time position and segment end time position registered in the ad entry as positioning conditions to locate the trigger window index in the trigger window set; register the located trigger window index in the original order of the ad sequence to generate the window index sequence.

[0050] S2.8. Perform backtracking chain concatenation on the window index sequence to generate a backtracking attribution chain sequence, specifically as follows: Based on the one-to-one correspondence between the window index sequence and the attribution index pair sequence, the attribution identifier at the corresponding position of the window index sequence is located in the attribution index pair sequence, and the correspondence between the window index and the attribution identifier is established. The window index sequence is processed one by one according to the order of the window index in the window index sequence. For window indexes with the same attribution identifier, backlinks are performed in the order of their appearance, and a backlink link is established between the later window index and the previous window index. The backlink links with the same attribution identifier are concatenated end to end to form a backlink attribution chain entry. The backlink attribution chain entries are collected according to the order of the window index in the window index sequence to generate a backlink attribution chain sequence.

[0051] S2.9. Perform a ranking decision on the backtracking attribution chain sequence using the attribution strategy model to obtain the target reward action; Locate the back attribution chain entries in the back attribution chain sequence, and locate the reward action candidate set with the same attribution identifier in the attribution index pair sequence based on the attribution identifier registered in the back attribution chain entries; for each reward action in the reward action candidate set, aggregate the trigger window level, interactive ad identifier, decay segment type, segment start time position, and segment end time position to form a ranking adjudication feature combination; the attribution strategy model calculates the ranking score corresponding to the reward action candidate set based on the ranking adjudication feature combination and outputs the ranking position; register the reward action with the first ranking position as the target reward action.

[0052] It should be noted that the ranking score is formed by combining the candidate set of reward actions with the trigger window level, interactive ad identifier, decay segment type, segment start time position, and segment end time position to form a ranking adjudication feature combination. The ranking adjudication feature combination is then input into the attribution strategy model, which calculates the score of the ranking adjudication feature combination at the ranking adjudication output layer to generate the corresponding ranking score.

[0053] The training process of the attribution strategy model is as follows: The attribution strategy model adopts a ranking adjudication structure, which includes a link encoding network, an action encoding network, a feature fusion layer, and a ranking adjudication output layer. The link encoding network receives backtracking attribution chain entries and encodes the link order information, attribution identifier association information, and window index connection information in the backtracking attribution chain entries to form a feature representation of the backtracking attribution chain entries. The action encoding network receives a reward action candidate set and encodes the reward action attribute information, trigger window level adaptation information, and interactive ad identifier association information in the reward action candidate set to form a feature representation of the reward action candidate set. The encoding processing in the link encoding network and the action encoding network adopts vector embedding encoding to encode the link order information, attribution identifier association information, and other relevant information. The system maps association information, window index connection information, and reward action attribute information into a unified-dimensional feature vector representation. The feature fusion layer concatenates and projects the feature representations of the backtracking attribution chain entries and the reward action candidate set to form a ranking adjudication feature combination. The ranking adjudication output layer receives the ranking adjudication feature combination and performs score calculation on each reward action candidate corresponding to the ranking adjudication feature combination to generate a ranking score for each reward action candidate. The score calculation in the ranking adjudication output layer adopts a linear weighted calculation method. The ranking score is generated by performing weight matrix multiplication and bias superposition operations on the ranking adjudication feature combination, and the ranking positions are formed according to the ranking scores from high to low. The reward action corresponding to the first position in the ranking position is registered as the target reward action.

[0054] The number of layers in the link coding network is set based on the length of the backtracking attribution chain entries and the complexity of link information coupling, for example, six coding layers, forming separable link representations while maintaining a stable distribution of the feature representations of the backtracking attribution chain entries; the number of layers in the action coding network is set based on the size of the reward action candidate set and the complexity of the reward action attributes, for example, four coding layers, reducing computational fluctuations while maintaining the distinguishability of the feature representations of the reward action candidate set; the projection dimension of the feature fusion layer is set based on the balanced expressive power of the feature representation dimensions of the backtracking attribution chain entries and the feature representation dimensions of the reward action candidate set, for example, 512 dimensions, maintaining the stability of the ranking score formation process within a unified dimensional space.

[0055] The attribution strategy model parameters include link coding network weights, action coding network weights, feature fusion layer weights, and ranking decision output layer parameters. The training data is organized by aligning back attribution chain sequences, attribution index pair sequences, reward action sequences, delayed rewards, attribution assignment sequences, and constraint adjustment records. Positive samples are constructed from the target reward actions corresponding to the attribution identifiers registered as constraint release markers in the constraint adjustment records, and negative samples are constructed from the target reward actions corresponding to the attribution identifiers registered as constraint tightening markers in the constraint adjustment records. Intra-batch negative samples are constructed from non-target reward actions in the same ranking decision feature combination.

[0056] Training employs a mini-batch training method, inputting the sequence decision feature combination and generating a sequence score. The loss function is a combination of sorted cross-entropy loss and a reward weighting term, with the reward weighting term generated by associating the reward identifier with the registration result in the delayed reward. The parameter update operator sequentially updates the weights of the link coding network, the action coding network, the feature fusion layer, and the sequence decision output layer. For example, the momentum parameter of the parameter update operator is set to 90%, used to perform weighted accumulation of historical gradient directions and current gradient directions during the parameter update process, thereby maintaining the continuity of the parameter update direction. If the weight decays to one percent, the scale of the link encoding network weights, action encoding network weights, feature fusion layer weights, and ranking decision output layer parameters, as well as the dimension of the ranking decision feature combination, are set according to the attribution strategy model. This is used to constrain the update magnitude of the link encoding network weights, action encoding network weights, feature fusion layer weights, and ranking decision output layer parameters during training and to keep the parameter range stable. After each round in the round scheduling sequence, the link encoding network weights, action encoding network weights, feature fusion layer weights, ranking decision output layer parameters, and parameter update operator states are retained and used as the initial state for the next round. The training process ends when the decrease in loss value is less than ten percent of the initial decrease, and the attribution strategy model is output.

[0057] It should also be noted that when the attribution strategy model has not yet generated training data, an initial reward action is selected based on the trigger window level and the range of reward actions corresponding to the interactive ad identifier, generating an initial reward action sequence. In the interactive ad interface, the reward action is presented according to the initial reward action sequence, and attribution association registration is performed on the user interaction behavior data corresponding to the initial reward action sequence, forming an initial delayed reward. The initial reward action sequence, initial delayed reward, trigger window set, and attribution index pair sequence are aligned and aggregated to form an initial training data set. This initial training data set is then input into the attribution strategy model to perform initial training, generating an attribution strategy model that can calculate ranking scores for ranking decision feature combinations.

[0058] S2.10 Write the corresponding trigger window identifier and attribution identifier to the target reward action to generate a reward action sequence.

[0059] S3. Perform advertising presentation and behavior association processing on the reward action sequence to form delayed rewards.

[0060] S3.1, Ad presentation refers to the timing-bound presentation of the target reward action in the interactive ad interface based on the correspondence between the target reward action and the trigger window identifier in the reward action sequence. Specifically, Locate the target reward action in the reward action sequence and read the trigger window identifier from the target reward action; locate the corresponding trigger window in the trigger window set based on the trigger window identifier, and determine the segment start time position and segment end time position registered in the corresponding trigger window; perform presentation time binding on the target reward action in the interactive ad interface based on the segment start time position and segment end time position, and register the timing binding of the target reward action with the segment start time position and segment end time position; present the target reward actions that have completed the timing binding registration in the interactive ad interface in the order of the reward action sequence, and complete the timing binding presentation of the trigger window.

[0061] S3.2, Behavior association processing involves attributing and registering user interaction behavior data with target reward actions according to the trigger window identifier, forming a delayed reward. Specifically, In user interaction behavior data, locate interaction behavior items according to the interaction behavior timestamp, and locate the target reward action with the same trigger window identifier in the reward action sequence; establish attribution association registration between interaction behavior items and target reward actions according to trigger window identifier, and associate and write the interaction behavior timestamp registered in the interaction behavior item with the trigger window identifier registered in the target reward action; aggregate and register the interaction behavior items and target reward actions that have completed attribution association registration to form delayed rewards.

[0062] S4. Perform backtracking attribution assignment and trigger window constraint adjustment on delayed rewards, update the attribution strategy model, calculate the reward strategy for the trigger window set based on the updated attribution strategy model, and generate incentive decisions.

[0063] S4.1. Perform back-attribution expansion on delayed rewards. Based on the attribution relationships in the attribution relationship set, perform link tracing on the attribution identifiers in delayed rewards and connect the target reward actions to generate a back-attribution chain sequence. Specifically, Expand the attribution identifiers one by one in the delayed rewards and match the attribution relationship entries with the same attribution identifiers in the attribution relationship set; extract the reward actions corresponding to the attribution relationship entries and determine the attribution correspondence between the reward actions and the attribution identifiers; perform link connection on the reward actions according to the order of the reward actions registered in the attribution relationship entries, and connect the reward actions corresponding to the same attribution identifier in sequence according to the order of appearance to form a backtracking attribution chain entry; gather the backtracking attribution chain entries according to the order of the attribution identifiers in the delayed rewards to generate a backtracking attribution chain sequence.

[0064] It should be noted that in the backtracking attribution chain sequence, attribution identifiers are used as link limiting identifiers to limit the attribution range of reward actions in the backtracking attribution chain entries; only reward actions with consistent attribution identifiers are subject to reward attribution determination, while reward actions with inconsistent attribution identifiers are excluded from the reward attribution determination range; by limiting the attribution range of reward actions in the backtracking attribution chain sequence through attribution identifiers, delayed rewards are only assigned among reward actions with consistent attribution identifiers, while reward actions with inconsistent attribution identifiers are excluded from the reward attribution determination range, thereby avoiding mixed reward attribution from reward actions with different interactive ad identifiers or different trigger window sources.

[0065] S4.2 Write back the trigger window identifier for the attribution assignment sequence, writing the reward identifier in the attribution assignment sequence to the corresponding trigger window identifier in the trigger window set, generating a reward trigger window set. Specifically, The backtracking attribution chain sequence is expanded line by line according to the order in which the backtracking attribution chain entries are arranged, resulting in backtracking attribution chain entries. The target reward actions in the backtracking attribution chain entries are then expanded according to the order in which they are arranged. In the delayed rewards, the corresponding reward identifiers are matched according to the consistency relationship of the attribution identifiers. The matched reward identifiers are associated with the target reward actions, and the reward identifiers are written into the target reward actions to form target reward actions with determined reward attribution. The target reward actions with determined reward attribution are then aggregated according to the order in which the backtracking attribution chain entries are arranged in the backtracking attribution chain sequence to generate an attribution assignment sequence.

[0066] S4.3. Adjust the trigger window constraints on the reward trigger window set. Mark the trigger window identifiers that continuously appear with the reward identifier as constraint release markers, and mark the trigger window identifiers that do not continuously appear with constraint tightening markers. Generate a constraint adjustment record. Specifically... Extract the correspondence between trigger window identifiers and reward identifiers in the reward trigger window set, and count the occurrence frequency of reward identifiers corresponding to the same trigger window identifier; perform a persistence determination on the continuous occurrence relationship of reward identifiers in the same trigger window identifier based on the occurrence frequency statistics; when reward identifiers form a continuous occurrence relationship in the same trigger window identifier, register the corresponding trigger window identifier as a constraint release mark; when reward identifiers do not form a continuous occurrence relationship in the same trigger window identifier, register the corresponding trigger window identifier as a constraint tightening mark; aggregate the execution records of trigger window identifiers and corresponding constraint release marks or constraint tightening marks to generate constraint adjustment records.

[0067] S4.4 Align and encapsulate the constraint adjustment records and attribution assignment sequences, and write them into the attribution strategy model to generate an updated attribution strategy model; the reward strategy calculation involves inputting the trigger window set into the updated attribution strategy model, and the updated attribution strategy model performs reward action matching and priority adjudication on the trigger window set to generate the target reward action, specifically... The reward strategy calculation involves inputting the set of trigger windows into the updated attribution strategy model. The updated attribution strategy model then performs reward action matching and ranking on the set of trigger windows to generate the target reward action. Specifically: The system retrieves the trigger window level, interactive ad identifier, segment start time position, and segment end time position from the trigger window set, and establishes a correspondence between the trigger window level and interactive ad identifier and the reward actions in the reward action pool to form a reward action candidate set. It then combines the reward action candidate set with the trigger window level, interactive ad identifier, segment start time position, and segment end time position to form a ranking adjudication feature combination. The updated attribution strategy model receives the ranking adjudication feature combination and generates a ranking score for the reward action candidate set corresponding to the ranking adjudication feature combination at the ranking adjudication output layer. The reward action candidate set corresponding to the ranking score is then sorted from high to low according to the ranking score. The reward action at the top of the ranking position is registered as the target reward action.

[0068] This embodiment also provides a computer device applicable to the user-engagement-driven interactive advertising incentive method, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the user-engagement-driven interactive advertising incentive method proposed in the above embodiment. The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0069] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the interactive advertising incentive method driven by user engagement as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0070] In summary, this invention characterizes the rhythmic changes in user engagement through continuous time sequencing and attention decay calculation. By combining trigger window level mapping and reward action attribution relationships, reward actions can be matched and triggered during stages of user engagement rhythm changes. Furthermore, the attribution strategy model uses a backtracking attribution chain to determine the order of reward actions, thereby enabling the interactive advertising incentive process to dynamically adjust around changes in user engagement behavior. This improves the matching ability between the timing of interactive advertising incentives and the rhythm of user engagement in advertising technology scenarios, thus enhancing the user engagement-driven interactive advertising incentive effect.

[0071] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A user-engagement-driven interactive advertising incentive method, characterized in that: include, Acquire user interaction behavior data, sort it over a continuous time period and calculate attention decay value to generate a decay sequence, perform decay status division and continuous time window determination on the decay sequence, and generate a set of trigger windows. Trigger window level mapping and reward action constraint filtering are performed on the trigger window set, and the attribution relationship between trigger windows and reward actions is registered to form a reward action pool. Reward action selection is performed in the reward action pool through the attribution strategy model to generate a reward action sequence. The reward action sequence is processed by ad presentation and behavior association to generate delayed rewards; The system performs backtracking attribution assignment and trigger window constraint adjustment for delayed rewards, updates the attribution strategy model, calculates reward strategies for the trigger window set based on the updated attribution strategy model, and generates incentive decisions.

2. The user-engagement-driven interactive advertising incentive method as described in claim 1, characterized in that: The generation of the decay sequence specifically involves... Get the dwell time, operation behavior and response rhythm in the interactive advertising interface, and write the interactive behavior timestamp and interactive advertising identifier to generate user interactive behavior data; The user interaction data is sorted continuously over time according to the timestamps of the interaction behaviors from earliest to latest, generating an interaction behavior sequence; Read the timestamps of adjacent interactive behaviors one by one from the sequence of interactive behaviors, and form a sequence of adjacent time intervals; Attention decay is calculated for each adjacent time interval sequence to obtain the attention decay value, and the attention decay value is written into the interaction behavior sequence to generate the decay sequence.

3. The user-engagement-driven interactive advertising incentive method as described in claim 1, characterized in that: The set of trigger windows to be generated specifically includes, Read adjacent attention decay values ​​one by one along the decay sequence, and divide the decay sequence into decay states according to the direction of change between adjacent attention decay values ​​to generate a decay segment sequence; Obtain the slow descent segment and the rapid descent segment from the decay segment sequence, and mark the slow descent segment and the rapid descent segment as decay segment type. At the same time, determine the start time position and end time position of the segment based on the interaction behavior timestamps corresponding to the slow descent segment and the rapid descent segment. The start and end times of a segment are determined using continuous time windows to generate a candidate time window sequence. The start time position, end time position, and decay segment type of the candidate time window sequence are matched with the interactive ad identifier to generate a set of trigger windows.

4. The user-engagement-driven interactive advertising incentive method as described in claim 1, characterized in that: A reward action pool is formed, specifically as follows: Perform trigger window level mapping on the attenuation segment types in the trigger window set to obtain the trigger window level; Match the trigger window level with the interactive ad identifier to obtain the set of trigger window reward actions; Perform mutual exclusion constraint verification, frequency constraint verification, and time period constraint verification on the reward actions in the set of reward actions in the trigger window. Retain the reward actions that pass the mutual exclusion constraint verification, frequency constraint verification, and time period constraint verification, and generate a set of executable reward actions. Based on the set of executable reward actions and the trigger window level, establish attribution relationships, write attribution identifiers, and generate an attribution relationship set; The reward actions in the attribution set are aggregated according to the trigger window level to generate a reward action pool.

5. The user-engagement-driven interactive advertising incentive method as described in claim 1, characterized in that: The generation of the reward action sequence is specifically as follows: Reward actions are extracted from the reward action pool, and the trigger window level, interactive ad identifier, and attribution identifier corresponding to the reward action are combined to form an attribution index pair sequence; The attribution index is used to map advertisements to the sequence, generating an advertisement sequence. Based on the trigger window level, perform window candidate positioning on the ad sequence and generate a window index sequence; Perform backlink concatenation on the window index sequence to generate a backlink attribution chain sequence; The target reward action is obtained by performing a ranking decision on the backtracking attribution chain sequence through an attribution strategy model. Write the corresponding trigger window identifier and attribution identifier to the target reward action to generate a reward action sequence.

6. The user-engagement-driven interactive advertising incentive method as described in claim 1, characterized in that: The advertisement presentation refers to the timing binding and presentation of the target reward action in the interactive advertisement interface based on the correspondence between the target reward action and the trigger window identifier in the reward action sequence; The behavior association processing involves registering user interaction behavior data and target reward actions according to the trigger window identifier to perform attribution association, thereby generating delayed rewards.

7. The user-engagement-driven interactive advertising incentive method as described in claim 1, characterized in that: The specific steps involved in updating the attribution strategy model are as follows: The delayed reward is back-attributed and expanded. Based on the attribution relationship in the attribution relationship set, the attribution identifier in the delayed reward is traced back and the target reward action is linked together to generate a back-attribution chain sequence. Perform reward attribution determination on the backtracking attribution chain sequence, register the reward identifier in the delayed reward to the target reward action of reward attribution determination, and generate an attribution assignment sequence; Write back the trigger window identifier for the attribution assignment sequence, write the reward identifier in the attribution assignment sequence into the corresponding trigger window identifier in the trigger window set, and generate a reward trigger window set; Adjust the trigger window constraints on the reward trigger window set, mark the trigger window identifiers that continuously appear with the reward identifier as constraint release markers, and mark the trigger window identifiers that do not continuously appear with the reward identifiers as constraint tightening markers, and generate constraint adjustment records; The constraint adjustment records and attribution assignment sequences are aligned and encapsulated, and then written into the attribution strategy model to generate the updated attribution strategy model.

8. The user-engagement-driven interactive advertising incentive method as described in claim 1, characterized in that: The reward strategy calculation involves inputting the set of trigger windows into the updated attribution strategy model, and then having the updated attribution strategy model perform reward action matching and ranking on the set of trigger windows to generate the target reward action.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the user-engagement-driven interactive advertising incentive method according to any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the user-engagement-driven interactive advertising incentive method according to any one of claims 1 to 8.

Images