Advertisement marketing system for data diversity identification

By introducing data diversity identification module, dual circulation control module and cross-layer information sharing channel module into the advertising and marketing system, the problem of traditional systems being unable to capture dynamic data diversity and lack of a collaboration mechanism in real time is solved, and the market response agility and strategy stability are improved.

CN120198177AInactive Publication Date: 2025-06-24QUANZHOU DIGITAL MEDIA CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202510296451.2
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-03-13
Publication Date
2025-06-24
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Traditional advertising and marketing systems cannot capture dynamic data diversity in real time, resulting in strategy adjustment lag behind market changes, and lack of a synergy mechanism between the external and internal cycles, which often triggers contradictions between long-term goals and short-term execution, and the isolated update of strategy parameters can easily cause strategy oscillation or deviation from the optimal path.

Method used

An advertising and marketing system for data diversity identification is designed, including a data diversity identification module, a dual cycle control module and a cross-layer information sharing channel module. The data diversity identification module calculates dynamic diversity indicators of multi-source heterogeneous data in real time; the dual-cycle control module uses the outer cycle strategy optimization layer and the inner cycle real-time control layer, and uses reinforcement learning and parameter adaptive algorithms for strategy optimization and real-time parameter adjustment; the cross-layer information sharing channel module realizes policy parameter transmission and imposes consistency constraints.

Benefits of technology

Real-time quantification of dynamic characteristics of multi-source data is achieved, and market response agility is enhanced, global planning and immediate feedback are taken into account, target conflicts are avoided, and advertising delivery strategies are maintained in a dynamic environment, so as to achieve efficient coordination between long-term returns and short-term effects.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN120198177A_ABST
    Figure CN120198177A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of advertisement marketing analysis, in particular to a data diversity identification advertisement marketing system, which comprises a data diversity identification module, a double-loop control module and a cross-layer information sharing channel module, and is characterized in that the data diversity identification module is used for calculating dynamic diversity indexes of multi-source heterogeneous data in real time; the double-circulation control module comprises an outer circulation strategy optimization layer unit and an inner circulation real-time control layer unit, the outer circulation strategy optimization layer unit operates at an hour-level time scale, a reinforcement learning algorithm is adopted to carry out long-term advertisement marketing strategy optimization, and an optimization objective function of the outer circulation strategy optimization layer unit is fused with a dynamic diversity index; the internal circulation real-time control layer unit operates at a second-level time scale, and adopts a parameter self-adaptive control algorithm to adjust an advertisement putting strategy in real time; and the cross-layer information sharing channel module is used for realizing strategy parameter transmission between the outer circulation and the inner circulation and applying strategy consistency constraint conditions so as to collaboratively optimize an advertisement marketing strategy.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of advertising marketing analysis, and particularly to an advertising marketing system for data diversity recognition. Background Art

[0002] In traditional advertising marketing systems, their data processing modules mostly rely on offline or static analysis, unable to capture dynamic data diversity in real time, resulting in a lag in strategy adjustment behind market changes; the strategy optimization layer and the real-time control layer are split in terms of time scale, and there is a lack of a coordination mechanism between the outer loop (such as reinforcement learning based on daily data) and the inner loop (second-level feedback control), often leading to contradictions between long-term goals and short-term execution.

[0003] In addition, the policy parameters between the inner and outer loops are updated in isolation, lacking consistency constraints, which easily causes policy oscillations or deviations from the optimal path. Traditional control algorithms have fixed parameters and are difficult to cope with the fluctuations in the real-time bidding environment and sudden changes in user behavior, resulting in low advertising placement efficiency and serious waste of resources. These problems make the system lack stability in complex scenarios and difficult to balance the multi-objective optimization requirements of advertising exposure, conversion rate, and cost control. Summary of the Invention

[0004] The purpose of the present invention is to provide an advertising marketing system for data diversity recognition to solve the problems raised in the above background art.

[0005] To achieve the above purpose, the present invention provides the following technical solution: An advertising marketing system for data diversity recognition, including a data diversity recognition module, a dual-loop control module, and a cross-layer information sharing channel module, where:

[0006] The data diversity recognition module is used to calculate the dynamic diversity index of multi-source heterogeneous data in real time;

[0007] The dual-loop control module includes an outer-loop strategy optimization layer unit and an inner-loop real-time control layer unit, where:

[0008] The outer-loop strategy optimization layer unit operates on an hourly time scale and uses a reinforcement learning algorithm to optimize long-term advertising marketing strategies, and its optimization objective function integrates the dynamic diversity index;

[0009] The inner-loop real-time control layer unit operates on a second-level time scale and uses a control algorithm with adaptive parameters to adjust advertising placement strategies in real time;

[0010] The cross-layer information sharing channel module is used to realize the transmission of policy parameters between the outer loop and the inner loop, and impose policy consistency constraint conditions to jointly optimize advertising marketing strategies.

[0011] As a further improvement of this technical solution, the data diversity recognition module calculates the information entropy of multi-source heterogeneous data and generates a dynamic diversity index by weighted summation.

[0012] As a further improvement of this technical solution, the information entropy has the following calculation formula:

[0013] , where is the probability of a certain data in the multi-source heterogeneous data, where represents the th information entropy of the data source, is the index of the data source.

[0014] As a further improvement of this technical solution, the dynamic diversity index has the following calculation formula:

[0015] , where The th information entropy weight of the data source is dynamically updated according to the confidence and real-time coverage rate of the multi-source heterogeneous data. The real-time coverage rate is determined by recording the data update timestamp of each multi-source heterogeneous data and calculating the number of updated data within a unit time.

[0016] As a further improvement of this technical solution, the outer loop strategy optimization layer unit defines a reward function, and the formula is as follows:

[0017] , where is the click-through rate, which measures the proportion of ads being clicked, is the cost of advertising placement, , and respectively represent the initial weight coefficients of the click-through rate, dynamic diversity index, and cost.

[0018] As a further improvement of this technical solution, the optimization process of the advertising marketing strategy specifically includes:

[0019] Initialize the strategy parameters and value function, and set the initial weight coefficients; collect advertising placement data once an hour; construct the current outer layer state vector according to the collected data; select an advertising placement strategy in the action space according to the current outer layer state vector;

[0020] Execute the advertising placement strategy and observe the advertising placement data. Calculate the reward of the current action according to the reward function. Store the current outer state vector, advertising placement strategy, the reward of the current action, and the next outer state vector in the experience replay buffer. Update the value function using the experience replay data and perform iterative optimization until the optimization objective function is maximized using the reward function, and generate a new advertising placement strategy. Among them, the optimized reward function incorporates a dynamic diversity metric, and the optimization objective function indirectly incorporates the dynamic diversity metric;

[0021] The state corresponding to the generated new advertising placement strategy is the target outer state vector, and the target outer state vector is passed to the inner loop real-time control layer unit.

[0022] As a further improvement of this technical solution, the inner loop real-time control layer unit constructs an inner state vector based on the collected advertising placement data and calculates the error between the inner state vector and the target outer state vector.

[0023] As a further improvement of this technical solution, the inner loop real-time control layer unit uses a PID controller to calculate the control signal , and the formula is as follows:

[0024] , where is the error between the inner state vector and the target outer state vector; is the proportional gain; is the integral gain Differential gain.

[0025] As a further improvement of this technical solution, the inner loop real-time control layer unit adjusts the advertising placement strategy according to the control signal , by decomposing the control signal into multiple sub-signals, each sub-signal corresponding to a specific advertising placement strategy parameter, and adjusting the corresponding advertising placement strategy according to the direction and magnitude of each sub-signal.

[0026] As a further improvement of this technical solution, the policy consistency constraint condition in the cross-layer information sharing channel module is realized through an optimization problem with a regularization term to ensure the smoothness of the joint objective function. The specific formula is as follows:

[0027] , where The policy parameter represents the specific parameter of the advertising placement strategy; represents the policy parameter generated by the outer loop policy optimization layer; represents the policy parameter used by the loop real-time control layer; represents the square of the Euclidean distance between the outer loop and inner loop policy parameters; Denotes a dynamically adjusted penalty factor, which is used to balance the trade-off between policy parameter differences and objective function smoothness; Denotes the joint objective function For The trace of the second derivative of; the goal of optimizing this formula is to adjust the above parameters so that the policy parameters of the outer loop and the inner loop tend to be consistent, and the joint objective function Tends to be smooth.

[0028] Compared with the prior art, the beneficial effects of the present invention are:

[0029] The advertising marketing system for data diversity recognition quantifies the dynamic characteristics of multi-source data in real time, provides environmental perception ability for policy optimization, and enhances market response agility; in the double-loop control module, the outer loop optimizes the long-term policy based on hourly reinforcement learning by integrating diversity metrics, and the inner loop realizes second-level real-time parameter tuning through the parameter adaptive algorithm, taking into account both global planning and instant feedback; the cross-layer information sharing channel module ensures the bidirectional synchronization of policy parameters, and enforces the consistency of the inner and outer loop policies through regularization constraints to avoid target conflicts; this design enables the advertising placement policy to maintain stability in a dynamic environment, achieving efficient coordination between long-term benefits and short-term effects. BRIEF DESCRIPTION OF THE DRAWINGS

[0030] Figure 1 Is the overall module schematic diagram of the present invention;

[0031] Figure 2 Is the schematic diagram of the double-loop control module unit of the present invention.

[0032] In the figure: 100, data diversity recognition module; 200, double-loop control module; 201, outer loop policy optimization layer unit; 202, inner loop real-time control layer unit; 300, cross-layer information sharing channel module. DETAILED DESCRIPTION OF THE INVENTION

[0033] Next, with reference to

[0034] Next, please refer to Figure 1 - Figure 2 , the present invention provides a technical solution: an advertising marketing system for data diversity recognition, including a data diversity recognition module 100, a double-loop control module 200, and a cross-layer information sharing channel module 300.

[0035] The data diversity recognition module 100 is used to calculate the dynamic diversity metrics of multi-source heterogeneous data in real time, specifically including:

[0036] Collect multi-source heterogeneous data from multiple different sources, such as website clickstreams, social media interactions, mobile app usage, offline sales data, etc.; among them, multi-source heterogeneous data includes but is not limited to structured data (such as database tables), semi-structured data (such as JSON, XML), and unstructured data (such as text, images, videos);

[0037] To process multi-source heterogeneous data in real time, a sliding window technique is adopted. The sliding window is a time window of a fixed size that slides over the data stream and continuously updates the data samples within the window; the size of the time window is adaptively adjusted according to the data update frequency;

[0038] Information entropy is a measure of data uncertainty. For each multi-source heterogeneous data, its information entropy is calculated , and the formula is as follows:

[0039] , where is the probability of a certain data item appearing in the multi-source heterogeneous data. The higher the information entropy, the more diverse the data of the multi-source heterogeneous data; among them represents the -th information entropy of the data source, is the index of the data source;

[0040] Generate a dynamic diversity metric by weighted summation of the information entropy of each multi-source heterogeneous data , and the formula is as follows:

[0041] , where The -th information entropy weight of the data source is dynamically updated according to the confidence and real-time coverage rate of the multi-source heterogeneous data, where the real-time coverage rate is determined by recording the data update timestamps of each multi-source heterogeneous data and calculating the number of updated data items per unit time;

[0042] Among them represents time throughout the text. Specifically, is a time point used to represent the current moment of the data, and in the afternoon has this meaning.

[0043] The double-loop control module 200 includes an outer-loop strategy optimization layer unit 201 and an inner-loop real-time control layer unit 202, where:

[0044] The outer loop policy optimization layer unit 201 operates on an hourly time scale and uses reinforcement learning algorithms to optimize long-term advertising marketing strategies. Its optimization objective function incorporates dynamic diversity metrics, specifically including:

[0045] Define the state space. The state space represents the state of the system at a certain moment. In the outer loop policy optimization layer unit 201, the state space includes the current advertising placement situation, user behavior data, market feedback data, etc.;

[0046] Define the action space. The action space is the set of all possible actions that the system can take. In the outer loop policy optimization layer unit 201, the action space includes different advertising placement strategies;

[0047] Define the reward function: The reward function is used to evaluate the goodness or badness of the system after taking a certain action. In the outer loop policy optimization layer unit 201, the reward function combines multiple factors, including click-through rate, conversion rate, cost, and dynamic diversity metrics. The formula is as follows:

[0048] , where is the click-through rate, which measures the proportion of ads being clicked, is the cost of advertising placement, , and respectively represent the initial weight coefficients of the click-through rate, dynamic diversity metric, and cost.

[0049] Outer loop policy optimization process:

[0050] Initialize the policy parameters and value function, and set the initial weight coefficients; collect advertising placement data once an hour, including information such as user clicks, conversions, and costs; construct the current outer layer state vector based on the collected data; select an advertising placement strategy in the action space according to the current outer layer state vector;

[0051] Execute the advertising placement strategy and observe the advertising placement data, calculate the reward of the current action according to the reward function, store the current outer layer state vector, advertising placement strategy, reward of the current action, and the next outer layer state vector in the experience replay buffer, update the value function using the experience replay data, and iterate and optimize until the optimization objective function is maximized using the reward function and generate a new advertising placement strategy. Since its optimization reward function incorporates dynamic diversity metrics, the optimization objective function indirectly incorporates dynamic diversity metrics;

[0052] The state corresponding to the generated new advertising placement strategy is the target outer layer state vector, and the target outer layer state vector is passed to the inner loop real-time control layer unit 202.

[0053] The inner-loop real-time control layer unit 202 operates on a second-level time scale and uses a control algorithm with parameter adaptation to adjust the advertising placement strategy in real time, specifically including:

[0054] Collect advertising placement data once per second, including information such as user clicks, conversions, costs, etc.;

[0055] Construct an inner-layer state vector based on the collected advertising placement data, and calculate the error between the inner-layer state vector and the target outer-layer state vector;

[0056] Use a PID controller to calculate the control signal , the formula is as follows:

[0057] , where is the error between the inner-layer state vector and the target outer-layer state vector; is the proportional gain; is the integral gain Differential gain;

[0058] According to the control signal , adjust the advertising placement strategy by decomposing the control signal into multiple sub-signals, each sub-signal corresponding to a specific advertising placement strategy parameter, and adjust the corresponding advertising placement strategy according to the direction and magnitude of each sub-signal to make it closer to the target state.

[0059] The cross-layer information sharing channel module 300 is used to implement the transmission of policy parameters between the outer loop and the inner loop, and impose policy consistency constraint conditions to jointly optimize the advertising marketing strategy, specifically including:

[0060] Define a standardized data interface to ensure that the outer-loop policy optimization layer unit 201 and the inner-loop real-time control layer unit 202 can efficiently exchange data, and use a message queue to achieve asynchronous communication to ensure the reliability and real-time nature of data transmission;

[0061] The policy consistency constraint condition is implemented through an optimization problem with a regularization term. This optimization problem aims to minimize the policy parameter difference between the outer-loop policy optimization layer unit 201 and the inner-loop real-time control layer unit 202, and uses a regularization term to ensure the smoothness of the joint objective function. The specific formula is as follows:

[0062] , where is the policy parameter representing the specific parameters of the advertising placement strategy, such as CTR (click-through rate), CVR (conversion rate), COST (cost), etc.; represents the policy parameter generated by the outer-loop policy optimization layer; represents the policy parameter used by the loop real-time control layer; Represents the square of the Euclidean distance between the outer loop and inner loop policy parameters; Represents a dynamically adjusted penalty factor used to balance the trade-off between policy parameter differences and objective function smoothness. A larger will make the optimization pay more attention to the smoothness of the objective function, and a smaller will make the optimization pay more attention to the consistency of the policy parameters; Represents the combined objective function of The trace of the second derivative; the goal of this formula optimization is to adjust the above parameters so that the policy parameters of the outer loop and inner loop tend to be consistent, and the combined objective function tends to be smooth, ultimately co-optimizing the advertising marketing strategy.

[0063] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions in the specification are only preferred examples of the present invention and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. An advertising marketing system for data diversity identification, characterized in that: It comprises a data diversity identification module (100), a dual-loop control module (200) and a cross-layer information sharing channel module (300), wherein: The data diversity identification module (100) is used to calculate the dynamic diversity index of multi-source heterogeneous data in real time; The dual-loop control module (200) comprises an outer-loop strategy optimization layer unit (201) and an inner-loop real-time control layer unit (202), wherein: The outer loop strategy optimization layer unit (201) operates on an hourly time scale and uses a reinforcement learning algorithm to optimize long-term advertising and marketing strategies, and its optimization objective function incorporates the dynamic diversity index; The inner loop real-time control layer unit (202) operates at a time scale of seconds and uses a parameter adaptive control algorithm to adjust the advertising delivery strategy in real time; The cross-layer information sharing channel module (300) is used to realize the transmission of strategy parameters between the outer loop and the inner loop, and impose strategy consistency constraints to collaboratively optimize the advertising marketing strategy.

2. The advertising and marketing system for data diversity identification according to claim 1, characterized in that: The data diversity identification module (100) calculates the information entropy of multi-source heterogeneous data and generates a dynamic diversity index by weighted summation.

3. The advertising and marketing system for data diversity identification according to claim 2, characterized in that: Information Entropy The calculation formula is as follows: ,in is the probability of a certain data item appearing in multi-source heterogeneous data, where Indicates The information entropy of the data source is The index of the data source.

4. The advertising and marketing system for data diversity identification according to claim 3, characterized in that: The dynamic diversity index The calculation formula is as follows: ,in No. The information entropy weight of each data source is dynamically updated according to the confidence and real-time coverage of multi-source heterogeneous data, where the real-time coverage is determined by recording the data update timestamp of each multi-source heterogeneous data and calculating the number of data updated per unit time.

5. The advertising and marketing system for data diversity identification according to claim 4, characterized in that: The outer loop strategy optimization layer unit (201) defines a reward function, the formula is as follows: ,in is the click-through rate, which measures the proportion of ads that are clicked. is the cost of advertising. , and They represent the initial weight coefficients of click rate, dynamic diversity index and cost respectively.

6. The advertising and marketing system for data diversity identification according to claim 1, characterized in that: The advertising marketing strategy optimization process specifically includes: Initialize the strategy parameters and value function, and set the initial weight coefficient; collect advertising data every hour; construct the current outer state vector based on the collected data; select an advertising strategy in the action space based on the current outer state vector; Execute the advertising delivery strategy and observe the advertising delivery data, calculate the reward of the current action according to the reward function, store the current outer state vector, advertising delivery strategy, the reward of the current action and the next outer state vector into the experience replay buffer, use the experience replay data to update the value function, iterate and optimize until the reward function is used to maximize the optimization objective function, and generate a new advertising delivery strategy, in which the optimization reward function integrates the dynamic diversity index, and the optimization objective function indirectly integrates the dynamic diversity index; The state corresponding to the new advertisement delivery strategy is generated as a target outer layer state vector, and the target outer layer state vector is transmitted to the inner loop real-time control layer unit (202).

7. The advertising and marketing system for data diversity identification according to claim 1, characterized in that: The inner loop real-time control layer unit (202) constructs an inner layer state vector according to the collected advertisement delivery data, and calculates the error between the inner layer state vector and the target outer layer state vector.

8. The advertising and marketing system for data diversity identification according to claim 7, characterized in that: The inner loop real-time control layer unit (202) uses a PID controller to calculate the control signal , the formula is as follows: ,in is the error between the inner state vector and the target outer state vector; is the proportional gain; is the integral gain Derivative gain.

9. The advertising and marketing system for data diversity identification according to claim 8, characterized in that: The inner loop real-time control layer unit (202) controls the ,adjust the advertising delivery strategy, by decomposing the control signal into multiple sub-signals, each sub-signal corresponds to a specific advertising delivery strategy parameter, and adjusts the corresponding advertising delivery strategy according to the direction and size of each sub-signal.

10. The advertising and marketing system for data diversity identification according to claim 1, characterized in that: The policy consistency constraint in the cross-layer information sharing channel module (300) is implemented through an optimization problem with a regularization term to ensure the smoothness of the joint objective function. The specific formula is as follows: ,in The strategy parameters represent specific parameters of the advertisement delivery strategy; Represents the strategy parameters generated by the outer loop strategy optimization layer; Represents the strategy parameters used by the cyclic real-time control layer; represents the square of the Euclidean distance between the outer loop and inner loop policy parameters; Represents a dynamically adjusted penalty factor used to balance the trade-off between policy parameter differences and objective function smoothness; Denotes the joint objective function right The goal of this formula optimization is to adjust the above parameters so that the strategy parameters of the outer loop and the inner loop are consistent, and the joint objective function Tends to be smooth.

Citation Information

Cited By

  • Advertisement effect evaluation method and system based on reinforcement learning

    CN120996875A

  • An advertisement effect evaluation method and system based on reinforcement learning

    CN120996875B