An advertisement putting dynamic game decision method and system based on multi-objective optimization

Abstract

The application provides an advertisement putting dynamic game decision method and system based on multi-target optimization, acquires user behavior key data, wherein the user behavior key data contains user visual focus position information; transmits the user visual focus position information to a multi-target optimization model in a preset decision server, generates multi-target optimization model input data; analyzes the multi-target optimization model input data and a preset game constraint condition by using the multi-target optimization model, generates multi-target game decision parameters; performs dynamic game deduction calculation based on the multi-target game decision parameters, generates a dynamic advertisement position adjustment scheme, and performs advertisement putting dynamic game decision. The application realizes millisecond-level data transmission to the multi-target optimization model, solves the strategy shock problem caused by data delay and target conflict in the traditional advertisement decision under the real-time attention scene, and realizes the accurate dynamic matching of user focus drift and advertisement display.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of advertisement dynamic game decision, and particularly relates to a multi-objective optimization-based advertisement dynamic game decision method and system. BACKGROUND

[0002] With the rapid development of the Internet advertising industry, in the scenario where user attention presents a high degree of dynamic change, how to timely capture the change trend of user attention and quickly adjust the advertisement display strategy accordingly has become a key technical requirement for improving advertisement conversion rate and user experience. This scenario requires the system to complete the collection, analysis and decision feedback of user behavior data within a millisecond delay.

[0003] The current mainstream solution is a real-time advertisement position dynamic optimization method based on reinforcement learning. By deploying a lightweight neural network model, the user interaction behavior signals from the browser end are received in real time, and the user attention distribution trend is predicted in combination with historical click data. The model takes these prediction results as input features, and uses a deep network to update the online strategy to realize the immediate adjustment of the advertisement display position and content. The existing solution has some inherent defects, including its dependence on the decision mechanism of a single agent, which may cause strategy conflicts or resource mismatches in a multi-agent competitive environment. When processing high-frequency user behavior data, it is limited by model inference efficiency and communication delay, affecting the response ability and stability of the overall system, and it is difficult to meet the online advertisement delivery requirements of high concurrency and low delay. SUMMARY

[0004] The present application provides a multi-objective optimization-based advertisement dynamic game decision method and system to solve the problems in the prior art, such as the dependence on the decision mechanism of a single agent, which may cause strategy conflicts or resource mismatches in a multi-agent competitive environment, and the limitation by model inference efficiency and communication delay when processing high-frequency user behavior data, affecting the response ability and stability of the overall system, and being difficult to meet the online advertisement delivery requirements of high concurrency and low delay.

[0005] In a first aspect, the present application provides a multi-objective optimization-based advertisement dynamic game decision method, comprising:

[0006] Obtaining user behavior key data, wherein the user behavior key data includes user visual focus position information;

[0007] Based on a preset low-delay wireless communication protocol, the user visual focus position information is transmitted to a multi-objective optimization model in a preset decision server to generate multi-objective optimization model input data;

[0008] Based on the dynamic change characteristics of the user visual focus position information, the multi-objective optimization model is used to analyze the multi-objective optimization model input data and the preset game constraint condition, and generate multi-objective game decision parameters;

[0009] Based on the multi-objective game decision parameters, dynamic game deduction calculation is performed to generate a dynamic advertisement position adjustment scheme for dynamic game decision of advertisement placement.

[0010] Optionally, the user behavior key data is obtained, wherein the user behavior key data includes user visual focus position information, which includes:

[0011] The user original visual signal is collected through a preset visual sensor, and the user original visual signal is combined to generate an original visual signal sequence;

[0012] The original visual signal sequence is time-sequentially segmented to generate discrete visual event units;

[0013] The features of each discrete visual event unit are extracted, and the features are combined to generate a visual event feature vector;

[0014] The visual event feature vector is input into a preset focus recognizer to generate a candidate visual focus coordinate set;

[0015] The candidate visual focus coordinate set is spatially clustered to generate final visual focus position information;

[0016] The final visual focus position information is bound to a preset user identity to generate user behavior key data.

[0017] Optionally, based on a preset low-delay wireless communication protocol, the user visual focus position information is transmitted to a multi-objective optimization model in a preset decision server to generate multi-objective optimization model input data, including:

[0018] The user visual focus position information is data-compressed to generate a compressed visual focus data packet;

[0019] The compressed visual focus data packet is encapsulated through a preset low-delay wireless communication protocol to generate a protocol encapsulation data frame;

[0020] The protocol encapsulation data frame is transmitted to a preset decision server to decompress the compressed visual focus data packet to generate original visual focus position information;

[0021] The original visual focus position information is spatially mapped with a preset advertisement scene space coordinate to generate multi-objective optimization model input data.

[0022] Optionally, based on the dynamic change characteristics of the user visual focus position information, the multi-objective optimization model is used to analyze the multi-objective optimization model input data and the preset game constraint condition, and a multi-objective game decision parameter is generated, including:

[0023] The coordinate offset of the user visual focus position information at the continuous timestamp is extracted, and the coordinate offset is connected to generate a focus movement trajectory;

[0024] The focus movement trajectory and the input data of the multi-objective optimization model are merged to generate trajectory feature data;

[0025] The trajectory feature data is input into the multi-objective optimization model for decision utility calculation to generate a decision utility output value;

[0026] The preset game constraint condition is converted into constraint boundary data, and the decision utility output value is filtered based on the constraint boundary data to generate a Pareto solution set;

[0027] Based on the preset weight configuration scheme, the solution with the highest weight allocation is selected from the Pareto solution set, and the solution with the highest weight allocation is parameterized to generate a multi-objective game decision parameter.

[0028] Optionally, the preset game constraint condition is converted into constraint boundary data, and the decision utility output value is filtered based on the constraint boundary data to generate a Pareto solution set, including:

[0029] The candidate solution vector in the decision utility output value is extracted to generate a candidate solution set;

[0030] The ad position exclusion rule in the preset game constraint condition is analyzed to generate a space conflict boundary;

[0031] The exposure frequency threshold in the preset game constraint condition is extracted to generate a frequency suppression boundary;

[0032] The space conflict boundary and the frequency suppression boundary are fused to generate a dynamic constraint filter;

[0033] Based on the dynamic constraint filter, the candidate solution set is hierarchically filtered to generate an initial Pareto solution set;

[0034] The dominated solution in the initial Pareto solution set is removed to generate a Pareto solution set.

[0035] Optionally, based on the multi-objective game decision parameter, a dynamic game deduction calculation is performed to generate a dynamic ad position adjustment scheme for dynamic ad placement game decision, including:

[0036] construct a dynamic game tree based on the multi-objective game decision parameter, to obtain an initial game structure according to the dynamic game tree;

[0037] traverse a decision path of the initial game structure, to generate a candidate advertisement position strategy set;

[0038] perform conflict resolution processing on the candidate advertisement position strategy set, to generate a strategy set after conflict resolution;

[0039] perform screening on the strategy set after conflict resolution based on a preset Pareto screening rule, to generate a dynamic advertisement position adjustment scheme, for dynamic game decision of advertisement delivery.

[0040] Optionally, the conflict resolution processing on the candidate advertisement position strategy set, to generate a strategy set after conflict resolution, comprises:

[0041] analyze the dynamic change feature of the user visual focus position information, to generate a real-time conflict resolution rule;

[0042] input the real-time conflict resolution rule into the candidate advertisement position strategy set, to generate a strategy set after rule reconstruction;

[0043] perform synchronous conflict detection and elimination operation on the strategy set after rule reconstruction, to generate a preliminary resolution strategy set;

[0044] verify the multi-objective equilibrium of the preliminary resolution strategy set, to generate a strategy set after conflict resolution.

[0045] In a second aspect, the present application provides an advertisement delivery dynamic game decision system based on multi-objective optimization, comprising:

[0046] an acquisition module, configured to acquire user behavior key data, wherein the user behavior key data contains user visual focus position information;

[0047] a transmission module, configured to transmit the user visual focus position information into a multi-objective optimization model in a preset decision server based on a preset low-delay wireless communication protocol, to generate multi-objective optimization model input data;

[0048] an analysis module, configured to analyze the multi-objective optimization model input data and a preset game constraint condition based on the dynamic change feature of the user visual focus position information, by using the multi-objective optimization model, to generate a multi-objective game decision parameter;

[0049] a calculation module, configured to perform dynamic game deduction calculation based on the multi-objective game decision parameter, to generate a dynamic advertisement position adjustment scheme, for dynamic game decision of advertisement delivery.

[0050] In a third aspect, the present application provides a computing device comprising a processor and a memory, wherein the memory stores a computer program, and the processor is configured to execute the computer program to implement the method for dynamic game decision of advertisement placement based on multi-objective optimization according to any one of the first aspect.

[0051] In a fourth aspect, the present application provides a computer storage medium storing computer program instructions, wherein the computer program instructions are executed by a processor to implement the method for dynamic game decision of advertisement placement based on multi-objective optimization according to any one of the first aspect.

[0052] The present application captures the user visual focus position information in real time through an optical sensor, realizes millisecond-level data transmission to a multi-objective optimization model in combination with a preset low-delay wireless communication protocol, generates decision parameters by synchronously analyzing dynamic change characteristics and preset game constraint conditions, and finally outputs an advertisement position adjustment scheme through dynamic game deduction, thereby solving the strategy shock problem caused by data delay and target conflict in traditional advertisement decision in a real-time attention scenario, and realizing precise dynamic matching of user focus drift and advertisement display.

[0053] Further, the original signal sequence is collected through a visual sensor and time-sequentially segmented into discrete event units, a visual event feature vector is generated by extracting combined features, visual line drift noise is eliminated through a focus identifier and spatial clustering processing, and finally high-precision visual focus position information is generated by binding the user identity, thereby overcoming the defects of large single-frame recognition error and weak identity correlation in a complex visual environment, and providing reliable low-noise input data for the multi-objective optimization model.

[0054] These and other aspects of the present application will become more apparent from the following description of the embodiments. BRIEF DESCRIPTION OF DRAWINGS

[0055] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, hereinafter, a brief introduction will be given to the drawings needed to be used in the description of the embodiments or the prior art. Obviously, the drawings described below are some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor based on these drawings.

[0056] Figure 1 A flowchart of a method for dynamic game decision of advertisement placement based on multi-objective optimization provided by an embodiment of the present application;

[0057] Figure 2 A structural schematic diagram of a system for dynamic game decision of advertisement placement based on multi-objective optimization provided by an embodiment of the present application;

[0058] Figure 3A structural schematic diagram of a computing device provided for an embodiment of the present application. DETAILED DESCRIPTION

[0059] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings of the embodiments of the present application.

[0060] In some of the processes described in this specification and in the accompanying drawings, multiple operations are described in a specific order. However, it should be understood that unless otherwise specifically stated, these operations can be performed in any order, or in parallel, and that the sequence of operations can be changed. The sequence of operations, such as 101, 102, etc., is merely illustrative of the order in which the operations can be performed, and the numbering of the operations is merely for the purpose of distinguishing between different operations, and does not necessarily imply a sequence of execution. Additionally, the processes can include more or fewer operations than those specifically described, and the operations can be performed in a different order than those specifically described. It should be understood that the use of "first", "second", etc., to describe a process or object is merely for the purpose of distinguishing between different processes or objects, and does not necessarily imply a sequence of execution or a sequence of ordering.

[0061] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the accompanying drawings of the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work fall within the scope of the present application.

[0062] Figure 1 A flowchart of an advertisement delivery dynamic game decision method based on multi-objective optimization provided for an embodiment of the present application is shown in FIG. 1, and the method includes the following steps. Figure 1

[0063] ​In the real-time user attention change scene, online advertising faces the challenge of dynamic drift of user visual focus, and the existing technology relies on historical behavior data, which leads to decision lag and cannot capture the instantaneous changes of focus shift. At the same time, the adjustment of the advertising position needs to consider multiple targets such as exposure rate, interference degree and cost, and the traditional step-by-step optimization method causes strategy shock. In addition, the delay from focus collection to strategy execution is more than a long time, which misses the attention golden window. In view of these problems, the research and development idea of the present application is: through the optical sensor, the user visual focus position information is captured in real time, and the preset low-delay wireless communication protocol is combined to realize the millisecond-level data transmission to the multi-objective optimization model of the decision server; by synchronously analyzing the focus dynamic change characteristics and the preset game constraint condition, the multi-objective conflict problem is solved in a single calculation; finally, the advertising position adjustment scheme is generated through dynamic game deduction calculation, the real-time matching mechanism of focus shift and advertising display is established, and the closed-loop response from focus capture to strategy execution is ensured in a very short time. Based on this, the present application provides a multi-objective optimization-based dynamic game decision method for advertising, as shown in Figure 1 , comprising:

[0064] Step 101: obtaining user behavior key data, wherein the user behavior key data includes user visual focus position information.

[0065] In this step, the user behavior key data refers to a data set reflecting the user's real-time attention distribution collected by an optical sensor, including user visual focus position information and its dynamic characteristics; the user visual focus position information refers to a screen coordinate set obtained based on eye movement signal processing, used to represent the real-time landing point of the user's line of sight in the interface.

[0066] In the embodiment of the present application, first, the optical sensor is used to collect the original eye movement signal of the user in real time, second, the original signal is combined into a continuous visual signal sequence in chronological order, then the visual signal sequence is divided into discrete visual event units by fixed time window segmentation, next the speed vector and residence time characteristics of each visual event unit are extracted and combined into a visual event feature vector, and finally the feature vector is processed by a preset focus recognizer to generate a candidate focus coordinate set, and the user visual focus position information is output after removing noise points by a spatial clustering algorithm, completing the acquisition of user behavior key data.

[0067] Step 102: based on the preset low-delay wireless communication protocol, the user visual focus position information is transmitted to the multi-objective optimization model in the preset decision server to generate multi-objective optimization model input data.

[0068] In this step, the preset low-latency wireless communication protocol refers to a preconfigured wireless data transmission rule for achieving millisecond-level information transmission between the user terminal and the decision server; the preset decision server refers to a pre-deployed remote computing device with a multi-objective optimization model for performing advertisement strategy generation; the multi-objective optimization model refers to an algorithm framework using the Pareto optimality principle for synchronously optimizing advertisement exposure rate, user interference degree, and delivery cost; and the multi-objective optimization model input data refers to reorganized structured vector data containing time-sequenced visual focus position information after protocol transmission.

[0069] In the embodiment of the present application, first, a dedicated data channel between the user terminal and the decision server is established through the preset low-latency wireless communication protocol; second, the user visual focus position information is split into data packets and attached with time stamp identifiers; then, the data packets are transmitted to the decision server in real time through the error correction mechanism built in the protocol; next, the data packets are reorganized according to the time stamp on the server side and verified for integrity; and finally, the reorganized focus position information is converted into a vector format that can be parsed by the multi-objective optimization model to generate multi-objective optimization model input data.

[0070] Step 103: Based on the dynamic change characteristics of the user visual focus position information, the multi-objective optimization model is used to analyze the multi-objective optimization model input data and the preset game constraint conditions to generate multi-objective game decision parameters.

[0071] In this step, the dynamic change characteristics refer to the movement speed and direction change amount extracted from the visual focus position information, which are used to quantify the attention drift trend; the preset game constraint conditions refer to a pre-defined set of advertisement delivery rules, including a position matching degree threshold, a cost weight upper limit, and an interference tolerance; the analysis operation refers to the processing process of performing multi-objective synchronous calculation and constraint condition verification in the multi-objective optimization model; and the multi-objective game decision parameters refer to the parameterized strategy set generated after model analysis, including advertisement position weight and target priority.

[0072] In the embodiment of the present application, first, the movement speed and direction change amount of the user visual focus position information are analyzed as dynamic change characteristics; second, the preset game constraint conditions include an advertisement display position matching degree threshold and a cost weight upper limit; then, the input data and game constraint conditions are synchronously loaded in the multi-objective optimization model; next, the Pareto frontier search mechanism is used to parallelly process the maximum exposure rate, minimum interference degree, and minimum cost minimization targets; and finally, the game decision parameters that satisfy the multi-objective equilibrium are output.

[0073] Step 104: Based on the multi-objective game decision parameters, dynamic game deduction calculation is performed to generate a dynamic advertisement position adjustment scheme for dynamic game decision of advertisement delivery.

[0074] In this step, the dynamic game deduction calculation operation refers to the sequential calculation process of constructing a game tree based on decision parameters, traversing strategy paths, and conflict resolution; the dynamic advertisement position adjustment scheme refers to the final execution instruction generated by the deduction calculation, including the coordinate adjustment value of the advertisement display area.

[0075] In the embodiment of the application, first, the dynamic game tree structure is initialized based on the multi-objective game decision parameters and the advertisement position node is generated, second, the candidate advertisement position strategy set is extracted by traversing the branches of the game tree, then the conflict detection is performed on the candidate strategy set and the resolution strategy sequence is generated by removing the strategies that violate the real-time rules, and finally, the optimal solution is selected from the sequence by the preset Pareto screening rule to generate the dynamic advertisement position adjustment scheme.

[0076] For example, first, the eye movement original signal is collected by the user terminal camera and combined into a visual signal sequence, and after being segmented into discrete event units, the speed features are extracted to generate a visual event feature vector, and the user visual focus position information is output by the focus recognizer and spatial clustering. Second, the focus position information is transmitted to the decision server through the preset 5G protocol, and is reorganized into a time sequence vector to generate input data for the multi-objective optimization model. Subsequently, the focus movement rate is parsed as a dynamic change feature on the server side, and the preset position matching degree threshold and cost constraint are loaded, and the game decision parameters of exposure rate-interference degree-cost balance are generated by parallel calculation through the multi-objective optimization model. Finally, based on the decision parameters, a dynamic game tree is constructed, the candidate strategy set is extracted by traversing the branches, and the conflict resolution is performed, and the advertisement position coordinate adjustment scheme is generated by the Pareto screening, which drives the real-time displacement of the webpage advertisement column to the user's visual focus area.

[0077] The embodiment of the application realizes millisecond-level data transmission by real-time capture of the dynamic drift features of the user's visual focus by an optical sensor and combination of a low-latency wireless communication protocol; utilizes a multi-objective optimization model to simultaneously process dynamic features and game constraint conditions, avoiding strategy oscillation caused by multi-objective conflict; generates an advertisement position adjustment scheme through dynamic game deduction, realizes real-time and accurate matching of user attention changes and advertisement display, and solves the placement failure problem caused by response delay and target fragmentation in traditional methods.

[0078] To solve the problem of large line-of-sight drift noise and weak identity association in a complex visual environment, this step generates high-precision focus positions through time sequence segmentation and spatial clustering, and binds user identities to construct behavior key data. The application provides a specific embodiment, step 101, obtaining user behavior key data, wherein the user behavior key data includes user visual focus position information, specifically including the following steps:

[0079] Step 111: Collecting user original visual signals through a preset visual sensor, and combining the user original visual signals to generate an original visual signal sequence.

[0080] In this step, the user original visual signal refers to the physiological feature data of eyeball directly collected by the visual sensor, including the pupil center coordinates, corneal reflection light spot position and blink frequency.

[0081] In the embodiment of the present application, first, the preset visual sensor continuously collects the pupil center coordinates and reflection light spot position of the user's eyeball as the original visual signal, and second, the original visual signal of the continuous frame is arranged and combined in time sequence to generate the original visual signal sequence containing the time stamp.

[0082] Step 112: Time sequence segmentation is performed on the original visual signal sequence to generate discrete visual event units.

[0083] In this step, the time sequence segmentation operation refers to the processing process of cutting the continuous signal sequence according to the fixed time window, which is used to generate independent analysis units with clear start and end time; the discrete visual event unit refers to the signal segment unit formed after segmentation, which contains the complete motion trajectory feature of the eyeball in a specific time period.

[0084] In the embodiment of the present application, first, a fixed length segmentation window is set, second, the original visual signal sequence is cut into independent segments according to the window boundary, then the start and end time stamps of each segment are labeled, and finally the discrete visual event unit containing the motion trajectory segment is generated.

[0085] Step 113: Extract the features of each discrete visual event unit, and combine the features to generate a visual event feature vector.

[0086] In this step, the visual event feature vector refers to the data structure representing the core attributes of the visual event unit, including the average moving speed, direction angle variance and residence time.

[0087] In the embodiment of the present application, first, the average moving speed and direction angle of the eyeball in each discrete visual event unit are calculated, second, the focus residence time and displacement variance are counted, then the speed, direction and residence time features are normalized and combined according to the preset dimension, and finally the visual event feature vector in a unified format is generated.

[0088] Step 114: Input the visual event feature vector into the preset focus identifier to generate a candidate visual focus coordinate set.

[0089] In this step, the candidate visual focus coordinate set refers to the probabilistic screen coordinate set output by the focus identifier, reflecting the spatial distribution of the user's visual line possible landing point.

[0090] In the embodiment of the present application, first, the visual event feature vector is input into the preset focus recognizer, second, the spatial attention weight is extracted through the convolutional neural network layer built in the recognizer, then the weight is mapped to the screen coordinate probability distribution, and finally the coordinate point set with the probability value higher than the threshold is output as the candidate visual focus coordinate set.

[0091] Step 115: spatial clustering is performed on the candidate visual focus coordinate set to generate final visual focus position information.

[0092] In this step, the spatial clustering operation refers to the process of merging adjacent coordinate points and filtering outliers based on the density algorithm, which is used to determine the precise position of the focus; and the final visual focus position information refers to the two-dimensional screen coordinate data generated after clustering and denoising, which represents the actual landing point of the user's line of sight.

[0093] In the embodiment of the present application, first, the clustering radius threshold is set based on the candidate visual focus coordinate set, second, the density of each coordinate point is calculated and the density reachable points are merged to form a cluster group, then the isolated point noise is removed, and finally the cluster group center point coordinate is extracted to generate the final visual focus position information.

[0094] Step 116: binding the final visual focus position information with the preset user identity to generate user behavior key data.

[0095] In this step, the binding operation refers to the data processing process of permanently associating the visual focus position information with the user identity.

[0096] In the embodiment of the present application, first, the preset user identity code is read, second, the final visual focus position information is associated with the identity code in key-value pairs, then the associated data is encrypted and stored, and finally the user behavior key data with identity tag is generated.

[0097] The embodiment of the present application converts continuous visual signals into discrete event units that can be analyzed through time sequence segmentation, and generates standardized vectors through feature extraction; uses focus recognizer and spatial clustering technology to eliminate line of sight jitter noise and output high-precision focus position; and finally binds the user identity to construct personalized behavior data, providing low-error and strong-association visual attention input for advertisement decision-making.

[0098] To reduce the positioning deviation of advertisements caused by transmission delay, this step realizes lossless transmission of focus information through data compression and protocol packaging, and maps the advertisement scene coordinates to generate model input data. The present application provides a specific embodiment, step 102, based on the preset low-delay wireless communication protocol, the user visual focus position information is transmitted to the multi-objective optimization model in the preset decision server to generate multi-objective optimization model input data, specifically including the following steps:

[0099] Step 201: data compression is performed on the user visual focus position information to generate a compressed visual focus data packet.

[0100] In this step, the data compression operation refers to a process of reducing the storage space of visual focus position information by using an entropy encoding algorithm, including differential encoding and field reorganization; the compressed visual focus data packet refers to a binary data structure processed by the compression algorithm, containing basic coordinates and position offset information.

[0101] In the embodiment of the present application, first, the Huffman encoding algorithm is used to losslessly compress the coordinate sequence of the user visual focus position information, second, the timestamp redundant field is removed and the spatial position difference value is calculated, then the difference value is combined and encoded with the basic coordinates, and finally the compressed visual focus data packet in binary format is generated.

[0102] Step 202: encapsulate the compressed visual focus data packet through a preset low-latency wireless communication protocol to generate a protocol encapsulation data frame.

[0103] In this step, the encapsulation operation refers to the process of adding transmission control information according to the communication protocol specification, including data blocking, check code addition and frame header generation; the protocol encapsulation data frame refers to a transmission unit with a protocol header and check data added, which is used to ensure the integrity of low-latency wireless transmission.

[0104] In the embodiment of the present application, first, a data frame header structure is created through a preset low-latency wireless communication protocol, second, the compressed visual focus data packet is divided into data blocks of a length specified by the protocol, then a time synchronization marker and a cyclic redundancy check code are added for each data block, and finally a protocol encapsulation data frame conforming to the protocol specification is generated.

[0105] Step 203: transmit the protocol encapsulation data frame to a preset decision server to decompress the compressed visual focus data packet to generate original visual focus position information.

[0106] In this step, the decompression operation refers to the process of reversing the compression algorithm at the receiving end, including data block reorganization and differential value decoding; the original visual focus position information refers to the recovered two-dimensional coordinate sequence of the screen, representing the original landing point distribution of the user's line of sight.

[0107] In the embodiment of the present application, first, the protocol encapsulation data frame is received through the network interface of the decision server, second, the frame header structure is parsed to verify data integrity, then the check code is stripped and the data blocks are reorganized, and finally the original coordinate sequence is restored by calling the corresponding decoder of the compression algorithm to generate the original visual focus position information.

[0108] Step 204: mapping the original visual focus position information with preset advertisement scene space coordinates to generate multi-objective optimization model input data.

[0109] In this step, the space mapping operation refers to the process of calculating the visual focus and the space distance of the advertisement position and establishing the weight correlation for quantifying the position matching degree.

[0110] In the embodiment of the application, first, a preset advertisement scene space coordinate set containing a page element position matrix is loaded, second, the spatial Euclidean distance of the original visual focus position information and the advertisement coordinates is calculated, then the shortest distance mapping relationship of the focus coordinates to the advertisement area is established, and finally the multi-objective optimization model input data with a space weight value is generated.

[0111] The embodiment of the application reduces the visual focus data transmission amount through lossless compression, and combines protocol packaging to ensure transmission reliability; after accurately restoring the original position information at the decision server end, the mapping relationship with the advertisement scene space coordinates is established, the model input data with a space weight is generated, and the problem of advertisement positioning deviation caused by transmission delay and coordinate distortion in the traditional method is solved.

[0112] To overcome the decision delay caused by the missing of trajectory features and the mismatch of parameter formats, this step generates feature data through trajectory merging, and outputs parameterized decision parameters through utility calculation and Pareto screening. The application provides a specific embodiment, step 103, based on the dynamic change characteristics of the user visual focus position information, the multi-objective optimization model is used to analyze the multi-objective optimization model input data and the preset game constraint condition, to generate multi-objective game decision parameters, specifically including the following steps:

[0113] Step 301: extracting the coordinate offset of the user visual focus position information at consecutive time stamps, connecting the coordinate offset to generate a focus movement trajectory.

[0114] In this step, the coordinate offset refers to the difference vector of the focus coordinates of adjacent time stamps, reflecting the direction and distance of the line of sight movement; the connection operation refers to the process of splicing the discrete offset at the beginning and the end in time sequence to form a continuous path trajectory; the focus movement trajectory refers to the path data structure composed of coordinate offset sequences, representing the trend of the line of sight movement.

[0115] In the embodiment of the application, first, the coordinate difference of the user visual focus position information at adjacent time stamps is obtained as the coordinate offset, second, the coordinate offset at consecutive time stamps is connected at the beginning and the end in time sequence to generate the focus movement trajectory representing the focus movement path.

[0116] Step 302: merging the focus movement trajectory with the input data of the multi-objective optimization model to generate trajectory feature data.

[0117] In this step, the input data refers to the basic input of the multi-objective optimization model, including ad position coordinates, timestamps, and historical click rates; the data merging operation refers to the process of aligning and superimposing the focus movement trajectory and the input data according to the timestamp; the trajectory feature data refers to the enhanced data set that fuses the trajectory points and the ad position coordinates, containing spatial correlation markers.

[0118] In the embodiment of the present application, first, the input data of the multi-objective optimization model includes ad position coordinates and timestamps, second, the focus movement trajectory is aligned and superimposed to the input data according to the timestamp, then the trajectory points and the ad position are merged to form a spatial correlation data set, and finally the trajectory feature data with trajectory markers is generated.

[0119] Step 303: input the trajectory feature data into the multi-objective optimization model to perform decision utility calculation, and generate decision utility output value.

[0120] In this step, the decision utility calculation operation refers to the process of parallel computing of ad benefit factors in the multi-objective optimization model; the decision utility output value refers to the comprehensive benefit evaluation scalar of the advertising strategy, and the larger the value is, the better the strategy is.

[0121] In the embodiment of the present application, first, the trajectory feature data is input into the utility calculation module of the multi-objective optimization model, second, the ad exposure rate gain, user interference degree penalty and delivery cost factor are calculated in parallel, then the values of each factor are weighted and summed, and finally the scalarized decision utility output value is generated.

[0122] Step 304: convert the preset game constraint condition into constraint boundary data, and filter the decision utility output value based on the constraint boundary data to generate a Pareto solution set.

[0123] In this step, the conversion operation refers to the process of converting the threshold value of the game constraint condition into a multi-dimensional mathematical boundary; the constraint boundary data refers to the mathematical expression set after the constraint condition is converted, which is used for solution filtering; the Pareto solution set refers to the decision utility solution set that simultaneously satisfies all constraint boundaries and is not dominated by each other.

[0124] In the embodiment of the present application, first, the preset game constraint condition includes a position matching degree threshold and a cost upper limit value, second, the threshold value is converted into multi-dimensional constraint boundary data, then the non-dominated solution of the decision utility output value that satisfies all boundary conditions is filtered, and finally the Pareto solution set is generated.

[0125] Step 305: select the solution with the highest weight allocation from the Pareto solution set based on the preset weight configuration scheme, and perform parameterized conversion on the solution with the highest weight allocation to generate a multi-objective game decision parameter.

[0126] In this step, the preset weight configuration scheme refers to the predefined advertising target priority rule, including exposure rate, interference degree, and cost weight value; the weight distribution highest solution refers to the optimal strategy vector with the maximum weighted sum value in the Pareto solution set; and the parameterized conversion refers to the processing process of mapping the solution vector into key-value pair parameters.

[0127] In the embodiment of the application, firstly, the preset weight configuration scheme is called to determine the priority weight of exposure rate, interference degree, and cost; secondly, the weighted sum value of each solution in the Pareto solution set is calculated; then, the solution with the highest sum value is selected; and finally, the solution vector is mapped into a key-value pair format to generate multi-target game decision parameters.

[0128] The embodiment of the application constructs a focus movement trajectory by connecting coordinate offset, enhances the spatial correlation between the advertising position and the line-of-sight path, generates a Pareto optimal solution set by combining utility calculation and constraint boundary screening, and finally selects the best solution through weight configuration and converts it into executable decision parameters, thereby solving the strategy failure problem caused by the missing trajectory features and the mismatched parameter format in the traditional method.

[0129] To solve the strategy feasibility defects caused by constraint fragmentation processing, a dynamic filter is constructed by fusing the spatial conflict and frequency suppression boundary in this step, and the dominated solution is removed after hierarchical screening to generate a Pareto solution set. The application provides a specific embodiment, step 304, converting the preset game constraint condition into constraint boundary data, screening the decision utility output value based on the constraint boundary data to generate a Pareto solution set, specifically including the following steps:

[0130] Step 341: Extracting the candidate solution vector in the decision utility output value to generate a candidate solution set.

[0131] In this step, the candidate solution vector refers to the strategy parameter set output by the decision utility calculation, including the advertising position, frequency, and cost configuration value.

[0132] In the embodiment of the application, firstly, all strategy vectors meeting the basic constraint condition are extracted from the decision utility output value, secondly, the invalid solution with a target function value lower than the preset threshold is filtered out, and finally, a candidate solution set containing feasible strategies is generated.

[0133] Step 342: Analyzing the advertising position exclusion rule in the preset game constraint condition to generate a spatial conflict boundary.

[0134] In this step, the advertising position exclusion rule refers to the constraint condition that prohibits the simultaneous display of specific advertising positions, such as the mutual exclusion of the top of the home page and the sidebar; the analysis operation refers to the process of converting the text format exclusion rule into a mathematically computable boundary; and the spatial conflict boundary refers to a polygon geographic fence data set generated based on the exclusion rule, used to identify the advertising position conflict area.

[0135] In the embodiment of the present application, first, the exclusion rule of the preset game constraint condition that prohibits the adjacent display of the advertising positions is parsed, second, the minimum allowed spacing threshold between the advertising positions is calculated, then the exclusion area polygon with the center of the advertising position as the origin and the spacing threshold as the radius is generated, and finally the spatial conflict boundary data set is output.

[0136] Step 343: The exposure frequency threshold in the preset game constraint condition is extracted, and a frequency suppression boundary is generated.

[0137] In this step, the exposure frequency threshold refers to the maximum number of times that the preset unit time allows the same user to display an advertisement; the frequency suppression boundary refers to the mathematical condition set of the trigger suppression mechanism constructed according to the frequency threshold.

[0138] In the embodiment of the present application, first, the upper limit value of the single-user advertising exposure frequency in the preset game constraint condition is read, second, the frequency counting function in the time window is constructed, then the threshold boundary condition that triggers the frequency suppression is generated, and finally the frequency suppression boundary rule set is output.

[0139] Step 344: The spatial conflict boundary and the frequency suppression boundary are fused to generate a dynamic constraint filter.

[0140] In this step, the fusion operation refers to the process of integrating boundary rules of different dimensions into unified filtering logic; the dynamic constraint filter refers to a programmable logic module that performs real-time spatial and frequency joint detection.

[0141] In the embodiment of the present application, first, the spatial conflict boundary and the frequency suppression boundary are superimposed according to the logical AND relationship, second, the boundary priority order is established, then it is compiled into executable filtering code, and finally the dynamic constraint filter module is generated.

[0142] Step 345: The candidate solution set is hierarchically screened based on the dynamic constraint filter to generate an initial Pareto solution set.

[0143] In this step, the hierarchical screening operation refers to the filtering mechanism of applying the constraint condition layer by layer according to the priority order; the initial Pareto solution set refers to the intermediate result set that passes through the dynamic constraint filter but contains dominated solutions.

[0144] In the embodiment of the present application, first, the dynamic constraint filter is used to perform spatial conflict detection on the candidate solution set, second, the solutions that pass the detection are verified for frequency suppression, then the solutions that satisfy all boundary conditions are output in layers, and finally the initial Pareto solution set is generated.

[0145] Step 346: The dominated solutions in the initial Pareto solution set are removed to generate a Pareto solution set.

[0146] In this step, the dominated solution is the invalid strategy vector that is inferior to other solutions in all dimensions of the objective function.

[0147] In the embodiment of the application, the objective function values of the solutions in the initial Pareto solution set are compared first, weak solutions that are dominated by other solutions are identified second, all dominated solutions are removed third, and a pure non-dominated solution set is generated fourth.

[0148] The embodiment of the application generates a space conflict boundary by analyzing the ad position exclusion rules, constructs a frequency suppression boundary in combination with the exposure frequency threshold, fuses the two to form a dynamic constraint filter to realize hierarchical screening, and efficiently eliminates space conflicts and frequency over-limit strategies. Finally, the dominated solutions are removed to obtain a pure Pareto solution set, solving the quality defects of the feasibility and quality of the strategies caused by the constraint fragmentation processing of the traditional method.

[0149] To eliminate the decline in decision quality caused by policy conflicts and missing screening standards, the strategy set is extracted by game tree deduction in this step, and a dynamic adjustment scheme is generated through conflict resolution and Pareto screening. The application provides a specific embodiment, step 104, which performs dynamic game deduction calculation based on the multi-objective game decision parameters to generate a dynamic ad position adjustment scheme for dynamic game decision of ad placement, specifically including the following steps:

[0150] Step 401: Construct a dynamic game tree based on the multi-objective game decision parameters to obtain an initial game structure according to the dynamic game tree.

[0151] In this step, the construction operation refers to the process of creating game tree nodes and branch connections based on decision parameters, including node initialization and topology relationship generation. The dynamic game tree refers to a tree-shaped data structure reflecting the decision path of the ad position, including root nodes, decision branch nodes, and leaf node strategies. The initial game structure refers to the physical topology of the constructed game tree, representing the framework of all possible decision paths.

[0152] In the embodiment of the application, the game tree root node is initialized based on the multi-objective game decision parameters first, the ad position decision branch node is created and the parent-child relationship is connected second, the node utility value and constraint condition are set third, and the initial game structure containing the complete node topology is generated fourth.

[0153] Step 402: Traverse the decision path of the initial game structure to generate a candidate ad position strategy set.

[0154] In this step, the decision path refers to the complete decision sequence from the root node to the leaf node, corresponding to a specific ad position strategy chain. The candidate ad position strategy set refers to the full set of strategies extracted by traversing the decision path, including feasible and conflicting strategies.

[0155] In the embodiment of the application, firstly, the root node of the initial game structure is started, secondly, all decision branch paths are traversed in depth, then, the advertisement position strategy corresponding to the path node sequence is recorded, and finally, the candidate advertisement position strategy set is generated.

[0156] Step 403: Conflict resolution processing is performed on the candidate advertisement position strategy set to generate a strategy set after conflict resolution.

[0157] In this step, the conflict resolution processing refers to the operation process of detecting and eliminating spatial conflicts and frequency overruns between strategies, and the strategy set after conflict resolution refers to the effective strategy set retained after conflict detection and removal processing.

[0158] In the embodiment of the application, firstly, spatial position conflicts and frequency overrun strategies in the candidate advertisement position strategy set are detected, secondly, invalid strategies that violate the preset exclusion rules and frequency threshold are removed, then, the target weight distribution of the remaining strategies is optimized, and finally, a conflict-free strategy set after conflict resolution is generated.

[0159] Step 404: The strategy set after conflict resolution is screened based on a preset Pareto screening rule to generate a dynamic advertisement position adjustment scheme for dynamic game decision of advertisement placement.

[0160] In this step, the preset Pareto screening rule refers to a preconfigured non-dominated solution optimization standard, including a target weight distribution algorithm.

[0161] In the embodiment of the application, firstly, the preset Pareto screening rule is loaded to define the target priority, secondly, the weighted non-dominated value of each strategy in the strategy set after conflict resolution is calculated, then, the strategy with the highest non-dominated value is selected, and finally, a dynamic advertisement position adjustment scheme containing an advertisement position coordinate adjustment instruction is generated.

[0162] The embodiment of the application realizes the visualization of the advertisement position decision path through dynamic game tree construction, and extracts a complete strategy set through depth traversal; after removing spatially exclusive and frequency rule-violating strategies through conflict resolution, the optimal adjustment scheme is screened by applying the Pareto rule, thereby solving the decision quality defects caused by strategy conflicts and missing screening standards in traditional methods.

[0163] To solve the problem that static rules cannot adapt to dynamic changes in attention, in this step, real-time rules are generated by analyzing the focus features, and an equilibrium strategy set is output after rule reconstruction and synchronous resolution verification. The application provides a specific embodiment, step 403, conflict resolution processing is performed on the candidate advertisement position strategy set to generate a strategy set after conflict resolution, specifically including the following steps:

[0164] Step 431: The dynamic change characteristics of the user visual focus position information are analyzed to generate real-time conflict resolution rules.

[0165] In this step, the parsing operation refers to the process of extracting quantitative indicators from visual focus dynamic characteristics and converting them into rules, including rate threshold mapping and constraint coefficient association; the real-time conflict resolution rule refers to a set of constraint conditions generated based on focus dynamic characteristics, including interference tolerance threshold and frequency constraint relaxation coefficient.

[0166] In the embodiment of the present application, first, the rate of change of user visual focus position information and acceleration value are quantified, then the rate threshold is mapped to the interference tolerance rule of the advertisement position, the acceleration change is associated with the frequency constraint relaxation coefficient, and finally the real-time conflict resolution rule containing dynamic threshold adjustment mechanism is generated.

[0167] Step 432: input the real-time conflict resolution rule into the candidate advertisement position strategy set to generate a strategy set after rule reconstruction.

[0168] In this step, the rule-reconstructed strategy set refers to the new strategy set formed after the original strategy is covered by dynamic rules, and the evaluation function is adapted to real-time attention changes.

[0169] In the embodiment of the present application, first, the original constraint conditions in the candidate advertisement position strategy set are read, then the real-time conflict resolution rule is applied to the static threshold in the original constraint, then the strategy evaluation function is reconstructed, and finally the rule-reconstructed strategy set adapted to the focus dynamic characteristics is generated.

[0170] Step 433: perform synchronous conflict detection and elimination operation on the rule-reconstructed strategy set to generate a preliminary resolution strategy set.

[0171] In this step, the synchronous conflict detection and elimination operation refers to the joint processing mechanism of parallelly executing spatial position mutual exclusion verification and exposure frequency compliance check; the preliminary resolution strategy set refers to the intermediate strategy set retained after synchronous conflict detection, containing partial feasible solutions without verified balance.

[0172] In the embodiment of the present application, first, the rule-reconstructed strategy set is parallelly executed for spatial position conflict detection and exposure frequency check, then the strategy items violating dynamic rules are removed immediately, then the target weight of the retained strategy is optimized, and finally the conflict-free preliminary resolution strategy set is generated.

[0173] Step 434: verify the multi-objective balance of the preliminary resolution strategy set to generate a conflict-resolved strategy set.

[0174] In this step, multi-objective balance refers to the Pareto optimal state in which the strategy is not dominated by other solutions in the three-dimensional target of exposure rate, interference degree, and cost; the verification operation refers to the process of identifying and removing dominated solutions by comparing the values of objective functions.

[0175] In the embodiment of the present application, firstly, a three-dimensional coordinate system of exposure rate-interference degree-cost of the initial conflict resolution strategy set is established, secondly, whether each strategy meets the Pareto equilibrium condition is verified, then, the strategies with the existing dominance relationship in the target dimension are removed, and finally, the conflict resolution strategy set of multi-objective equilibrium is generated.

[0176] The embodiment of the present application generates real-time conflict resolution rules by analyzing the dynamic characteristics of the focus, dynamically reconstructs the strategy constraint conditions, efficiently eliminates the space and frequency conflicts after synchronous detection, verifies the multi-objective equilibrium, and outputs the Pareto optimal strategy set, thereby solving the problem of strategy mismatch caused by the fact that the traditional static rules cannot adapt to the dynamic changes of attention.

[0177] Figure 2 A structural diagram of an advertisement delivery dynamic game decision system based on multi-objective optimization is provided for the embodiment of the present application, as shown in Figure 2 The system comprises:

[0178] The acquisition module 21 is configured to acquire user behavior key data, wherein the user behavior key data comprises user visual focus position information.

[0179] The transmission module 22 is configured to transmit the user visual focus position information to a multi-objective optimization model in a preset decision server based on a preset low-delay wireless communication protocol, and generate multi-objective optimization model input data.

[0180] The analysis module 23 is configured to analyze the multi-objective optimization model input data and a preset game constraint condition based on the dynamic change characteristics of the user visual focus position information by using the multi-objective optimization model, and generate multi-objective game decision parameters.

[0181] The calculation module 24 is configured to perform dynamic game deduction calculation based on the multi-objective game decision parameters, generate a dynamic advertisement position adjustment scheme, and perform dynamic game decision for advertisement delivery.

[0182] Figure 2 The multi-objective optimization-based advertisement delivery dynamic game decision system can perform Figure 1 The implementation principle and technical effects of the multi-objective optimization-based advertisement delivery dynamic game decision method described in the embodiment shown in the above are not described again. The specific operation of each module and unit of the multi-objective optimization-based advertisement delivery dynamic game decision system described in the above embodiment has been described in detail in the embodiment related to the method, and will not be described in detail here.

[0183] In one possible design, Figure 2 The multi-objective optimization-based advertisement delivery dynamic game decision system described in the embodiment shown in the above can be implemented as a computing device, such asFigure 3 As shown, the computing device can include a storage component 31 and a processing component 32.

[0184] The storage component 31 stores one or more computer instructions, wherein the one or more computer instructions are called for execution by the processing component 32.

[0185] The processing component 32 is configured to obtain user behavior key data, wherein the user behavior key data includes user visual focus position information; based on a preset low-delay wireless communication protocol, transmit the user visual focus position information to a multi-objective optimization model in a preset decision server, to generate multi-objective optimization model input data; based on dynamic change characteristics of the user visual focus position information, use the multi-objective optimization model to analyze the multi-objective optimization model input data and a preset game constraint condition, to generate multi-objective game decision parameters; based on the multi-objective game decision parameters, perform dynamic game deduction calculation, to generate a dynamic advertisement position adjustment scheme, for dynamic game decision of advertisement placement.

[0186] The processing component 32 can include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component can also be one or more application specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors or other electronic components, configured to execute the above method.

[0187] The storage component 31 is configured to store various types of data to support the operation of the terminal. The storage component can be implemented by any type of volatile or non-volatile storage device or their combination, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0188] Of course, the computing device can also include other components, such as an input / output interface, a display component, a communication component, etc.

[0189] The input / output interface provides an interface between the processing component and peripheral interface modules, which can be output devices, input devices, etc.

[0190] The communication component is configured to facilitate wired or wireless communication between the computing device and other devices, etc.

[0191] The computing device can be a physical device or an elastic computing host provided by a cloud computing platform, and the computing device can be a cloud server, and the processing component and the storage component can be basic server resources rented or purchased from the cloud computing platform.

[0192] The embodiment of the present application further provides a computer storage medium storing a computer program, and the computer program can realize the above method when executed by a computer. Figure 1 The embodiment shown in the figure is a dynamic game decision method for advertisement delivery based on multi-target optimization.

[0193] Those skilled in the art can clearly understand the specific working process of the system, device and unit described above for the convenience and brevity of description, and the corresponding process in the foregoing method embodiments can be referred to, which will not be repeated here.

[0194] The device embodiments described above are only schematic, and the units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0195] Through the description of the foregoing embodiments, those skilled in the art can clearly understand that each embodiment can be realized by means of software and a necessary general hardware platform, and of course, it can also be realized by hardware. Based on such understanding, the foregoing technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the method described in each embodiment or some parts of the embodiment.

[0196] Finally, it should be noted that: the foregoing embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement to some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A dynamic game-theoretic decision-making method for advertising placement based on multi-objective optimization, characterized in that, include: Acquire key user behavior data, wherein the key user behavior data includes user visual focus position information; Based on a preset low-latency wireless communication protocol, the user's visual focus position information is transmitted to a multi-objective optimization model in a preset decision server to generate multi-objective optimization model input data. Based on the dynamic change characteristics of the user's visual focus position information, the multi-objective optimization model is used to analyze the input data of the multi-objective optimization model and the preset game constraints to generate multi-objective game decision parameters. Based on the multi-objective game decision parameters, dynamic game deduction calculations are performed to generate dynamic advertising position adjustment schemes for dynamic game decision-making in advertising placement.

2. The method according to claim 1, characterized in that, Obtain key user behavior data, wherein the key user behavior data includes user visual focus position information, including: The user's original visual signals are collected by a preset visual sensor, and the user's original visual signals are combined to generate an original visual signal sequence. The original visual signal sequence is temporally segmented to generate discrete visual event units; Features are extracted from each discrete visual event unit, and the features are combined to generate a visual event feature vector. The visual event feature vector is input into a preset focus recognizer to generate a set of candidate visual focus coordinates; Spatial clustering is performed on the candidate visual focus coordinate set to generate the final visual focus position information; The final visual focus position information is bound to a preset user identity identifier to generate key user behavior data.

3. The method according to claim 1, characterized in that, Based on a preset low-latency wireless communication protocol, the user's visual focus position information is transmitted to a multi-objective optimization model in a preset decision server to generate multi-objective optimization model input data, including: The user's visual focus position information is compressed to generate a compressed visual focus data packet; The compressed visual focus data packet is encapsulated using a preset low-latency wireless communication protocol to generate a protocol-encapsulated data frame. The protocol-encapsulated data frame is transmitted to a preset decision server to decompress the compressed visual focus data packet and generate the original visual focus position information; The original visual focus position information is spatially mapped to the preset advertising scene spatial coordinates to generate input data for a multi-objective optimization model.

4. The method according to claim 1, characterized in that, Based on the dynamic changes in the user's visual focus position information, the multi-objective optimization model is used to analyze the input data and preset game constraints to generate multi-objective game decision parameters, including: Extract the coordinate offset of the user's visual focus position information under consecutive timestamps, and connect the coordinate offsets to generate the focus movement trajectory; The focus movement trajectory is merged with the input data of the multi-objective optimization model to generate trajectory feature data; The trajectory feature data is input into the multi-objective optimization model to calculate decision utility and generate a decision utility output value. The preset game constraints are converted into constraint boundary data, and the decision utility output value is filtered based on the constraint boundary data to generate a Pareto solution set. Based on a preset weight configuration scheme, the solution with the highest weight allocation is selected from the Pareto solution set, and the solution with the highest weight allocation is parameterized to generate multi-objective game decision parameters.

5. The method according to claim 4, characterized in that, The preset game constraints are converted into constraint boundary data. Based on the constraint boundary data, the decision utility output values ​​are filtered to generate a Pareto solution set, including: Extract candidate solution vectors from the decision utility output values ​​to generate a candidate solution set; The ad space exclusion rules in the preset game constraints are analyzed to generate spatial conflict boundaries; Extract the exposure frequency threshold from the preset game constraint conditions to generate a frequency suppression boundary; The spatial conflict boundary and the frequency suppression boundary are fused to generate a dynamic constraint filter; Based on the dynamic constraint filter, the candidate solution set is hierarchically filtered to generate an initial Pareto solution set; The dominant solutions in the initial Pareto solution set are removed to generate the Pareto solution set.

6. The method according to claim 1, characterized in that, Based on the multi-objective game decision parameters, dynamic game simulation calculations are performed to generate dynamic ad placement adjustment schemes for dynamic game decision-making in ad placement, including: A dynamic game tree is constructed based on the multi-objective game decision parameters, and the initial game structure is obtained from the dynamic game tree. The decision paths of the initial game structure are traversed to generate a set of candidate ad placement strategies; The candidate ad placement strategy set is subjected to conflict resolution processing to generate a conflict-resolved strategy set; The set of strategies after conflict resolution is filtered based on a preset Pareto screening rule to generate a dynamic ad placement adjustment scheme for dynamic game decision-making in ad placement.

7. The method according to claim 6, characterized in that, The candidate ad placement strategy set is subjected to conflict resolution processing to generate a conflict-resolved strategy set, including: The dynamic change characteristics of the user's visual focus position information are analyzed to generate real-time conflict resolution rules; The real-time conflict resolution rules are input into the candidate ad position strategy set to generate a rule-reconstructed strategy set; Perform synchronous conflict detection and elimination operations on the reconstructed policy set to generate a preliminary resolution policy set; The multi-objective balance of the initial conflict resolution strategy set is verified to generate the conflict-resolved strategy set.

8. A dynamic game-theoretic decision-making system for advertising placement based on multi-objective optimization, characterized in that, include: The acquisition module is used to acquire key user behavior data, wherein the key user behavior data includes user visual focus position information; The transmission module is used to transmit the user's visual focus position information to the multi-objective optimization model in the preset decision server based on a preset low-latency wireless communication protocol, and generate input data for the multi-objective optimization model. The analysis module is used to analyze the input data of the multi-objective optimization model and the preset game constraints based on the dynamic change characteristics of the user's visual focus position information, and generate multi-objective game decision parameters. The calculation module is used to perform dynamic game deduction calculations based on the multi-objective game decision parameters, generate dynamic advertising position adjustment schemes, and make dynamic game decisions for advertising placement.

9. A computing device, characterized in that, It includes a processing component and a storage component; the storage component stores one or more computer instructions; the one or more computer instructions are invoked and executed by the processing component to implement the advertising placement dynamic game decision-making method based on multi-objective optimization as described in any one of claims 1 to 7.

10. A computer storage medium, characterized in that, The system contains a computer program that, when executed by a computer, implements a dynamic game-theoretic decision-making method for advertising placement based on multi-objective optimization as described in any one of claims 1 to 7.

Images