Advertisement position flow prediction method and system

Abstract

The application discloses an advertisement position flow prediction method and system, and relates to the technical field of advertisement position flow prediction.The method comprises the following steps: acquiring multi-source data covering the surrounding physical environment, semantic information and external association of an advertisement position; using encryption technology to process the collected data; after the original data is encrypted in a local node, a cross-domain model aggregation is realized by using a federal learning algorithm; the data diversity is expanded by using a data enhancement algorithm, and an abnormal data point is identified by using an abnormality detection algorithm, and then a multi-dimensional feature structure is constructed, and feature conversion and dimension reduction processing are performed on specific data.The application provides strong support for decision-making by simulating flow changes under different strategies by using algorithms such as a data enhancement algorithm, a causal analysis framework quantization and a proximal policy optimization, reduces invalid exposure, and improves advertisement delivery effect.

Description

Technical Field

[0001] This invention relates to the field of advertising space traffic prediction technology, specifically to an advertising space traffic prediction method and system. Background Technology

[0002] In the wave of digital commerce, ad placement traffic prediction is crucial for precise ad targeting, but traditional methods face numerous challenges. Firstly, data collection and utilization are limited. Most methods rely solely on the historical traffic data of the ad placement itself, ignoring important factors such as the surrounding environment (pedestrian traffic, traffic conditions), and social media trends. This single data dimension leads to significant prediction bias. Furthermore, centralized data processing models, where data is uploaded to a central server, pose a risk of privacy breaches. Data holders, concerned about sensitive data leaks, are unwilling to share data, creating data silos that limit multi-source data fusion and fail to provide comprehensive data support for prediction models. Secondly, prediction models and decision optimization capabilities are insufficient. Traditional prediction models are often based on simple statistical analysis or linear regression, making it difficult to capture the complex spatiotemporal characteristics and dynamic changes in ad placement traffic. Prediction errors increase significantly when faced with unforeseen circumstances. Moreover, existing technologies lack effective strategies for optimizing ad placement resource allocation after traffic prediction, failing to help advertisers develop precise targeting plans based on prediction results, thus hindering efficient use of advertising resources and maximizing revenue. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for predicting ad placement traffic, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for predicting ad placement traffic, comprising the following steps:

[0005] Acquire multi-source data covering the physical environment surrounding the ad placement, semantic information, and external connections;

[0006] The collected data is processed using encryption technology. After encrypting the original data on the local node, a federated learning algorithm is used to achieve cross-domain model aggregation.

[0007] Data augmentation algorithms are used to expand data diversity, anomaly detection algorithms are used to identify anomalous data points, and then multidimensional feature structures are constructed and feature transformation and dimensionality reduction are performed on specific data.

[0008] Based on the causal analysis framework, the causal impact of various data factors on ad slot traffic is quantified, and the feature contribution is displayed with the help of visualization technology. The spatiotemporal characteristics and dynamic changes of ad slot traffic are captured by the hybrid model of fused neural networks, while an adaptive mechanism is set up to ensure the reliability of the prediction.

[0009] The ad placement scheduling strategy is trained using a near-end strategy optimization algorithm with the objective function of maximizing advertiser revenue. At the same time, a digital twin model is built to simulate traffic changes under different strategies, and a strategy rollback mechanism is triggered based on prediction errors.

[0010] Preferably, the acquisition of multi-source data covering the physical environment surrounding the ad placement, semantic information, and external connections includes:

[0011] Collect physical environment data such as pedestrian traffic and heat distribution around the advertising space;

[0012] Analyze social media data using natural language processing models to obtain semantic tagging information;

[0013] Obtain urban heat maps and traffic control information data.

[0014] Preferably, the step of processing the collected data using encryption technology, encrypting the original data at the local node, and then using a federated learning algorithm to achieve cross-domain model aggregation includes:

[0015] Deploy edge nodes locally on the advertiser's site and use homomorphic encryption technology to encrypt the original data;

[0016] After encrypting the original data on the local node, cross-domain model aggregation is achieved using the FedProx algorithm, where the local update formula is:

[0017] ;

[0018] The global aggregation formula is:

[0019] ;

[0020] In the formula, Indicates the first The local model in the ... Parameters during round training, For learning rate, It is the first The loss function of a local model. These are hyperparameters that control how closely the local model resembles the global model. It is the first Global model parameters during training rounds. It is the number of local models participating in federated learning. It is the first The weights of each local model;

[0021] Specifically, the FedProx algorithm adjusts the hyperparameters in the local model update formula when performing cross-domain model aggregation. This controls the degree of similarity between the local model and the global model.

[0022] Preferably, the step of using data augmentation algorithms to expand data diversity, using anomaly detection algorithms to identify anomalous data points, and then constructing a multidimensional feature structure and performing feature transformation and dimensionality reduction on specific data includes:

[0023] The StyleGAN3 algorithm is used to generate virtual traffic samples. By adjusting the latent vectors in the generator, virtual traffic samples with different weather conditions and time periods are generated for data augmentation.

[0024] Use the IsolationForest algorithm to detect abnormal traffic points;

[0025] We construct a spatiotemporal feature cube and perform feature engineering on social media tag data using TF-IDF transformation and PCA dimensionality reduction.

[0026] Preferably, the causal analysis framework quantifies the causal impact of various data factors on ad slot traffic, and uses visualization technology to display feature contributions. A hybrid neural network model is used to capture the spatiotemporal characteristics and dynamic changes of ad slot traffic, while an adaptive mechanism is set to ensure prediction reliability, including:

[0027] Traffic forecasting is performed based on the CausalML framework and a hybrid forecasting model. The CausalML framework is used to quantify the causal effects of various factors on traffic, and the contribution of features is visualized through SHAP values.

[0028] The hybrid prediction model includes a spatiotemporal Transformer and a dynamic Bayesian network to capture the spatiotemporal characteristics and dynamic changes of ad placements, and sets an adaptive gating mechanism to switch to an expert system when the prediction confidence is lower than a set threshold.

[0029] The causal effect of each factor on the flow rate is quantified using the difference-in-differences formula:

[0030] ;

[0031] In the formula, Indicates the average treatment effect. and These represent the flow observations for the treatment group and the control group, respectively. It is an indicator variable that indicates whether or not the action has been taken. =1 indicates that it has been processed. =0 indicates that it has not been processed.

[0032] Preferably, in the step of training the ad slot scheduling strategy using a near-end strategy optimization algorithm with the objective function of maximizing advertiser revenue:

[0033] The reward function formula for the near-end policy optimization algorithm is as follows:

[0034] ;

[0035] In the formula, Indicates at time The reward value, , , These are weighting coefficients. It is a moment Click-through rate It is a moment The cost per thousand impressions It is a moment Diversity metrics for ad display;

[0036] The near-end strategy optimization algorithm is as follows:

[0037] ;

[0038] In the formula, Indicating in strategy The expectations below It is a discount factor. It is the time step. It is a moment The actions taken It is a moment The state.

[0039] Preferably, when the near-end strategy optimization algorithm trains the ad slot scheduling strategy, it adjusts the strategy parameters according to the reward function, where the weight coefficients in the reward function are... , , The adjustments will be made dynamically based on the advertiser's business priorities and objectives.

[0040] The present invention also provides an ad placement traffic prediction system, comprising:

[0041] The data collection and privacy protection module is used to acquire multi-source data covering the physical environment, semantic information and external connections around the ad space, and to process the collected data using encryption technology. After encrypting the original data on the local node, the federated learning algorithm is used to achieve cross-domain model aggregation.

[0042] The data preprocessing module is used to expand the diversity of data by employing data augmentation algorithms, identify anomalous data points by using anomaly detection algorithms, and then construct multidimensional feature structures and perform feature transformation and dimensionality reduction on specific data.

[0043] The predictive model design module is used to quantify the causal impact of various data factors on ad slot traffic based on a causal analysis framework, and to display the feature contribution with the help of visualization technology. It uses a hybrid model of fused neural networks to capture the spatiotemporal characteristics and dynamic changes of ad slot traffic, while setting an adaptive mechanism to ensure the reliability of the prediction.

[0044] The dynamic decision output module is used to train ad placement scheduling strategies using a near-end strategy optimization algorithm, with the objective function of maximizing advertiser revenue. At the same time, it constructs a digital twin model to simulate traffic changes under different strategies and triggers a strategy rollback mechanism based on prediction errors.

[0045] The present invention also provides an electronic device, which is a physical device, comprising:

[0046] The processor and the memory are communicatively connected.

[0047] The memory is used to store at least one executable instruction executed by the processor, which executes the executable instruction to implement the ad placement traffic prediction method as described above.

[0048] The present invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the ad placement traffic prediction method described above.

[0049] Compared with the prior art, the beneficial effects of the present invention are:

[0050] By integrating multi-source information through multimodal data acquisition, a rich data foundation is provided for accurate prediction. Federated edge computing, combined with homomorphic encryption technology, promotes multi-source data collaboration while protecting data privacy, reducing data transmission volume, and improving data processing efficiency and security.

[0051] Data augmentation and anomaly detection techniques improve data quality and diversity, enhance the model's adaptability to complex situations, and feature engineering builds a comprehensive and efficient feature system to effectively extract and utilize data features.

[0052] The causal analysis-based approach delves into the causal relationships between various factors and traffic flow, and combines a hybrid prediction model to capture complex spatiotemporal characteristics and dynamic change patterns, significantly improving prediction accuracy. The adaptive gating mechanism ensures that reliable prediction results can still be provided under complex conditions, enhancing the stability and reliability of the system.

[0053] Near-end strategy optimization optimizes ad placement scheduling with the goal of maximizing advertiser revenue. It fully considers multiple key business indicators to achieve reasonable allocation and efficient utilization of advertising resources. The digital twin model provides an effective means of strategy verification and optimization. By simulating traffic changes under different strategies, it provides strong support for decision-making, reduces invalid exposure, and improves the effectiveness of ad placement. Attached Figure Description

[0054] Figure 1 The main flowchart of an ad placement traffic prediction method provided in an embodiment of the present invention;

[0055] Figure 2 This is a schematic diagram of the structure of an ad placement traffic prediction system provided in an embodiment of the present invention;

[0056] Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0057] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0058] The method in this embodiment is executed by a terminal, which can be a mobile phone, tablet computer, PDA, laptop or desktop computer, etc. Of course, it can also be other devices with similar functions, and this embodiment does not limit them.

[0059] Please see Figure 1 This invention provides a method and system for predicting ad placement traffic, wherein the method is applied to, including:

[0060] Step 100: Obtain multi-source data covering the physical environment surrounding the ad placement, semantic information, and external connections.

[0061] Step 100 includes:

[0062] Data on the physical environment around the advertising space, such as pedestrian traffic and heat distribution, is collected using LiDAR and thermal imaging cameras.

[0063] The ERNIE-3.0 natural language processing model is used to analyze social media data, such as short videos, to obtain semantic tag information.

[0064] Access city heat map API and traffic control information data.

[0065] Specifically, LiDAR and thermal imaging cameras are deployed to collect real-time pedestrian traffic around the advertising space (accuracy ±2 people / ㎡) and heat distribution (resolution 0.1℃). At the same time, the ERNIE-3.0 model is used to analyze scene tags related to the advertising space in short video platforms such as Douyin and Bilibili to obtain semantic information. In addition, the city heat map API and traffic control information are connected to collect various data that affect the traffic of the advertising space in an all-round way.

[0066] Step 200: The collected data is processed using encryption technology. After the original data is encrypted on the local node, the federated learning algorithm is used to achieve cross-domain model aggregation.

[0067] Specifically, edge nodes are deployed locally by the advertiser, and homomorphic encryption technology is used to encrypt the original data;

[0068] After encrypting the original data on the local node, cross-domain model aggregation is achieved using the FedProx algorithm, where the local update formula is:

[0069] ;

[0070] The global aggregation formula is:

[0071] ;

[0072] In the formula, Indicates the first The local model in the ... Parameters during round training, For learning rate, It is the first The loss function of a local model. These are hyperparameters that control how closely the local model resembles the global model. It is the first Global model parameters during training rounds. It is the number of local models participating in federated learning. It is the first The weights of each local model;

[0073] Specifically, the FedProx algorithm adjusts the hyperparameters in the local model update formula when performing cross-domain model aggregation. This controls the degree of similarity between the local model and the global model.

[0074] Specifically, edge nodes are deployed locally by advertisers, and homomorphic encryption (HE) technology is used to encrypt the original data to ensure data security during transmission and processing. Cross-domain model aggregation is achieved through the FedProx algorithm, which effectively reduces communication overhead while protecting data privacy, and reduces the amount of training data per round by 60%.

[0075] Step 300: Data augmentation algorithms are used to expand data diversity, anomaly detection algorithms are used to identify anomalous data points, and then a multidimensional feature structure is constructed and feature transformation and dimensionality reduction are performed on specific data.

[0076] Specifically, step 300 includes:

[0077] Step 310: Use the StyleGAN3 algorithm to generate virtual traffic samples. By adjusting the latent vectors in the generator, generate virtual traffic samples for different weather conditions and time periods to augment the data.

[0078] Step 320: Use the IsolationForest algorithm to detect abnormal traffic points;

[0079] Step 330: Construct a spatiotemporal feature cube and perform feature engineering on the social media tag data using TF-IDF transformation and PCA dimensionality reduction.

[0080] First, StyleGAN3 is used to generate virtual traffic samples for different weather and time periods, such as simulating traffic conditions on rainy nights or early mornings during holidays, which can expand the diversity of data. At the same time, the IsolationForest algorithm is used to detect abnormal traffic points and promptly identify abnormal situations such as traffic surges caused by sudden protests.

[0081] Secondly, a spatiotemporal feature cube containing information such as longitude, latitude, time, heat value, and semantic vectors is constructed to comprehensively describe the spatiotemporal state of the ad placement. For social media tags, they are first converted into sparse features using the TF-IDF algorithm, and then reduced to 64 dimensions using the PCA algorithm for subsequent model processing.

[0082] In reality, ad placement traffic is influenced by various factors, and real data may not cover all situations. Using StyleGAN3 to generate virtual traffic samples under different weather conditions and time periods can effectively expand the diversity of the data. StyleGAN3 is based on the principles of Generative Adversarial Networks (GANs), learning the distribution of real data through adversarial training between the generator and discriminator. When generating virtual traffic samples, researchers can adjust the latent vectors in the generator to generate traffic data for different weather conditions (such as sunny, rainy, and snowy days) and different time periods (such as weekday daytime, weekend nighttime, and holidays). These virtual samples supplement potentially missing information in the original data, enabling the model to learn more comprehensive traffic change patterns and improve its generalization ability in various scenarios. For example, when training the model to predict outdoor ad placement traffic, virtually generated traffic samples under heavy rain can help the model learn the impact of severe weather on traffic, thus enabling it to more accurately respond to similar situations in actual predictions.

[0083] Furthermore, outliers often exist in the actual collected data. These outliers can be caused by various reasons, such as equipment failure or special events. If left untreated, they can interfere with the training and prediction results of the model. The IsolationForest algorithm is based on the concept of an isolation forest. It isolates data points by constructing a binary tree. Normal data points are usually isolated at the bottom of the tree, while outliers are more likely to be isolated at higher levels. The algorithm determines whether a data point is an outlier by calculating the path length of each data point. In ad placement traffic data processing, by setting appropriate thresholds, the IsolationForest algorithm can quickly detect abnormal traffic points, such as a sudden surge in traffic to an ad placement during non-event periods or a sharp drop in traffic during normal periods. Timely detection and handling of these outliers can ensure data quality and improve the accuracy and reliability of model predictions.

[0084] Step 400: Based on the causal analysis framework, quantify the causal impact of each data factor on ad slot traffic, and use visualization technology to display the feature contribution. Use a hybrid model of fused neural networks to capture the spatiotemporal characteristics and dynamic changes of ad slot traffic, and set an adaptive mechanism to ensure the reliability of the prediction.

[0085] Specifically, traffic forecasting is performed based on the CausalML framework and a hybrid forecasting model. The CausalML framework is used to quantify the causal effects of various factors on traffic, and the contribution of features is visualized through SHAP values.

[0086] The hybrid prediction model includes a spatiotemporal Transformer and a dynamic Bayesian network to capture the spatiotemporal characteristics and dynamic changes of ad placements, and sets an adaptive gating mechanism to switch to an expert system when the prediction confidence is lower than a set threshold.

[0087] The causal effect of each factor on the flow rate is quantified using the difference-in-differences formula:

[0088] ;

[0089] In the formula, Indicates the average treatment effect. and These represent the flow observations for the treatment group and the control group, respectively. It is an indicator variable that indicates whether or not the action has been taken. =1 indicates that it has been processed. =0 indicates that it has not been processed.

[0090] Specifically, based on the CausalML framework, the causal effects of various factors on traffic are quantified in depth. For example, the impact of "the opening of a subway line" on the traffic of surrounding advertising spaces is analyzed, and it is concluded that it increases traffic by 18%. The contribution of features is visualized through SHAP values, which intuitively show the degree of influence of each factor on traffic prediction. For example, the influence coefficient of "promotional activities" on traffic is +0.25.

[0091] Secondly, a spatiotemporal Transformer is used to capture the spatial dependencies between ad placements, such as calculating the cross-correlation coefficients of ad placement traffic within a business district. A dynamic Bayesian network is introduced to update the weights of external events on traffic in real time. For example, during the "college entrance examination" period, the traffic of ad placements around schools decreased by 22%. An adaptive gating mechanism is set up so that when the prediction confidence is lower than 70%, it automatically switches to a rule-based expert system to ensure the reliability of the prediction.

[0092] Step 500: Use the near-end strategy optimization algorithm to train the ad slot scheduling strategy with the objective function of maximizing advertiser revenue. At the same time, build a digital twin model to simulate traffic changes under different strategies and trigger the strategy rollback mechanism based on the prediction error.

[0093] The reward function formula for the near-end policy optimization algorithm is as follows:

[0094] ;

[0095] In the formula, Indicates at time The reward value, , , These are weighting coefficients. It is a moment Click-through rate It is a moment The cost per thousand impressions It is a moment Diversity metrics for ad display;

[0096] The near-end strategy optimization algorithm is as follows:

[0097] ;

[0098] This formula is used to update the policy parameters of a reinforcement learning model. In the formula, Indicating in strategy The expectations below It is a discount factor. It is the time step. It is a moment Actions taken (such as advertising strategies). It is a moment Status (such as ad space traffic, surrounding environment, etc.).

[0099] Among them, when training the ad slot scheduling strategy, the near-end strategy optimization algorithm adjusts the strategy parameters according to the reward function, and the weight coefficients in the reward function... , , The adjustments will be made dynamically based on the advertiser's business priorities and objectives.

[0100] The construction of the digital twin model to simulate traffic changes under different strategies includes:

[0101] Based on point cloud data collected by LiDAR and thermal data collected by thermal imaging cameras, a 3D model of the area surrounding the advertising space is constructed by combining geographic information data, and traffic changes are simulated by inputting different advertising placement strategy parameters.

[0102] Specifically, the Proximal Policy Optimization (PPO) algorithm is used to train the ad placement scheduling strategy. With maximizing advertiser revenue as the objective function, the algorithm comprehensively considers indicators such as cost per thousand impressions (CPM) and click-through rate (CTR) to output a prediction report with confidence intervals, such as "Friday evening peak traffic: 1200±80 people, it is recommended to increase the rotation frequency", providing advertisers with clear decision-making basis.

[0103] Secondly, a 3D digital twin model of the area surrounding the ad placement is constructed to simulate traffic changes under different strategies. When the prediction error exceeds 25% for two consecutive hours, a strategy rollback mechanism is automatically triggered to ensure the stability and effectiveness of the ad placement strategy.

[0104] In this embodiment, the multimodal data acquisition method integrates information from multiple sources, comprehensively covering various factors that affect ad placement traffic, and providing a rich data foundation for accurate prediction. Federated edge computing, combined with homomorphic encryption, achieves "usable but invisible" data, protecting data privacy while promoting multi-source data collaboration, reducing data transmission volume, and improving data processing efficiency and security. Data augmentation and anomaly detection technologies enhance data quality and diversity, improving the model's adaptability to complex situations. Carefully designed feature engineering constructs a comprehensive and efficient feature system, effectively extracting and utilizing data features to lay a solid foundation for accurate traffic prediction. Causal analysis-based methods delve into the causal relationships between various factors and traffic, combining hybrid prediction models to capture complex spatiotemporal characteristics and dynamic change patterns, significantly improving prediction accuracy. Adaptive gating mechanisms ensure reliable prediction results even under complex conditions, enhancing system stability and reliability. Reinforcement learning strategies optimize ad placement scheduling with the goal of maximizing advertiser revenue, fully considering multiple key business indicators to achieve reasonable allocation and efficient utilization of advertising resources. Digital twin models provide effective means for strategy verification and optimization, simulating traffic changes under different strategies to provide strong support for decision-making, reduce invalid exposure, and improve advertising effectiveness.

[0105] Based on the above embodiments, such as Figure 2 As shown, the present invention also provides an ad placement traffic prediction system to support the ad placement traffic prediction method of the above embodiments. The ad placement traffic prediction system includes:

[0106] The data collection and privacy protection module 11 is used to acquire multi-source data covering the physical environment, semantic information and external associations around the ad space, and to process the collected data using encryption technology. After encrypting the original data at the local node, the federated learning algorithm is used to achieve cross-domain model aggregation.

[0107] The data preprocessing module 12 is used to expand the diversity of data by using data augmentation algorithms, identify abnormal data points by using anomaly detection algorithms, and then construct a multidimensional feature structure and perform feature transformation and dimensionality reduction on specific data.

[0108] The prediction model design module 13 is used to quantify the causal impact of various data factors on ad slot traffic based on the causal analysis framework, and to display the feature contribution with the help of visualization technology. It uses a hybrid model of fused neural networks to capture the spatiotemporal characteristics and dynamic changes of ad slot traffic, and sets an adaptive mechanism to ensure the reliability of prediction.

[0109] The dynamic decision output module 14 is used to train the ad slot scheduling strategy using the near-end strategy optimization algorithm, with the objective function of maximizing the advertiser's revenue. At the same time, it constructs a digital twin model to simulate traffic changes under different strategies and triggers a strategy rollback mechanism based on the prediction error.

[0110] In an optional embodiment, the data acquisition and privacy protection module 11 further includes:

[0111] The data acquisition unit is used to collect physical environment data such as pedestrian traffic and heat distribution around the advertising space; to parse social media data through a natural language processing model to obtain semantic tag information; and to acquire urban heat map and traffic control information data.

[0112] The data fusion unit is used to deploy edge nodes locally on the advertiser's site and encrypt the original data using homomorphic encryption technology. After encrypting the original data on the local node, cross-domain model aggregation is achieved through the FedProx algorithm.

[0113] In an optional embodiment, the data preprocessing module 12 further includes:

[0114] The data augmentation unit is used to generate virtual traffic samples using the StyleGAN3 algorithm. By adjusting the latent vectors in the generator, it generates virtual traffic samples for different weather and time periods for data augmentation, and uses the IsolationForest algorithm to detect abnormal traffic points.

[0115] The feature engineering processing unit is used to construct a spatiotemporal feature cube and perform feature engineering processing on social media tag data through TF-IDF transformation and PCA dimensionality reduction.

[0116] In an optional embodiment, the prediction model design module 13 further includes:

[0117] The spatiotemporal causal unit is used for traffic forecasting based on the CausalML framework and a hybrid prediction model. It uses the CausalML framework to quantify the causal effects of various factors on traffic and visualizes the feature contributions through SHAP values.

[0118] The hybrid prediction unit, comprising a spatiotemporal Transformer and a dynamic Bayesian network, is used to capture the spatiotemporal characteristics and dynamic changes of ad placements. It also features an adaptive gating mechanism that switches to the expert system when the prediction confidence level falls below a set threshold.

[0119] In an optional embodiment, the dynamic decision output module 14 further includes:

[0120] The digital twin verification unit is used to construct a 3D model of the area surrounding the advertising space based on point cloud data collected by LiDAR and thermal data collected by thermal imaging cameras, combined with geographic information data, and to simulate traffic changes by inputting different advertising placement strategy parameters.

[0121] In this embodiment, the multimodal data acquisition method integrates information from multiple sources, comprehensively covering various factors that affect ad placement traffic, and providing a rich data foundation for accurate prediction. Federated edge computing, combined with homomorphic encryption, achieves "usable but invisible" data, protecting data privacy while promoting multi-source data collaboration, reducing data transmission volume, and improving data processing efficiency and security. Data augmentation and anomaly detection technologies enhance data quality and diversity, improving the model's adaptability to complex situations. Carefully designed feature engineering constructs a comprehensive and efficient feature system, effectively extracting and utilizing data features to lay a solid foundation for accurate traffic prediction. Causal analysis-based methods delve into the causal relationships between various factors and traffic, combining hybrid prediction models to capture complex spatiotemporal characteristics and dynamic change patterns, significantly improving prediction accuracy. Adaptive gating mechanisms ensure reliable prediction results even under complex conditions, enhancing system stability and reliability. Reinforcement learning strategies optimize ad placement scheduling with the goal of maximizing advertiser revenue, fully considering multiple key business indicators to achieve reasonable allocation and efficient utilization of advertising resources. Digital twin models provide effective means for strategy verification and optimization, simulating traffic changes under different strategies to provide strong support for decision-making, reduce invalid exposure, and improve advertising effectiveness.

[0122] Furthermore, the intelligent analysis of investment user operation habits cache data processing device can run the above-mentioned intelligent analysis of investment user operation habits cache data processing method. For specific implementation, please refer to the method embodiment, which will not be repeated here.

[0123] Based on the above embodiments, such as Figure 3 As shown, the present invention also provides an electronic device, the electronic device comprising:

[0124] The processor 22 includes at least one processor 22, at least one memory 21, a communication interface 23, and a communication bus 24, wherein the processor 22 is communicatively connected to the memory 21.

[0125] In this embodiment, the memory 21 can be implemented in any suitable manner, for example, the memory 21 can be a read-only memory, a hard disk drive, a solid-state drive, or a USB flash drive, etc.; the memory 21 is used to store at least one executable instruction executed by the processor;

[0126] In this embodiment, the processor 22 can be implemented in any suitable manner. For example, the processor 22 can take the form of a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) that can be executed by the (micro)processor, logic gates, switches, application-specific integrated circuits (ASICs), programmable logic controllers, and embedded microcontrollers, etc.; the processor is used to execute the executable instructions to implement the cached data processing method for intelligent analysis of investment user operating habits as described above.

[0127] Based on the above embodiments, the present invention also provides a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, it implements the above-described intelligent analysis of investment user operating habits cached data processing method.

[0128] Those skilled in the art will recognize that the modules and method steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0129] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, equipment, and modules described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0130] In the several embodiments provided in this application, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple modules or units may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices, or equipment, and may be electrical, mechanical, or other forms.

[0131] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0132] In addition, the functional modules in the various embodiments of the present invention can be integrated into one processing module, or each module can exist physically separately, or two or more modules can be integrated into one module.

[0133] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program instructions, such as USB flash drives, portable hard drives, read-only storage servers, random access storage servers, magnetic disks, or optical disks.

[0134] Furthermore, it should be noted that the combination of the various technical features in this case is not limited to the combination methods described in the claims of this case or the combination methods described in the specific embodiments. All technical features described in this case can be freely combined or combined in any way, unless they contradict each other.

[0135] It should be noted that the above examples are merely specific embodiments of the present invention, and the present invention is obviously not limited to the above embodiments, with many similar variations. All modifications that can be directly derived or conceived by those skilled in the art from the content disclosed in this invention should fall within the protection scope of this invention.

[0136] The above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. An advertisement inventory flow prediction method, characterized by, The method comprises the following steps: acquiring multi-source data covering the surrounding physical environment, semantic information and external association of the advertising position; processing the collected data using encryption technology, encrypting the original data at the local node, and implementing cross-domain model aggregation using a federated learning algorithm; expanding data diversity using a data enhancement algorithm and identifying abnormal data points using an anomaly detection algorithm; quantifying the causal effect of each data factor on the advertising position traffic based on a causal analysis framework, displaying feature contribution using visualization technology, capturing the spatiotemporal characteristics and dynamic changes of the advertising position traffic using a hybrid neural network model, and setting an adaptive mechanism to ensure prediction reliability; training an advertising position scheduling strategy using a proximal policy optimization algorithm, taking maximizing advertiser revenue as the objective function, constructing a digital twin model to simulate traffic changes under different strategies, and triggering a strategy rollback mechanism according to the prediction error; wherein the reward function formula of the proximal policy optimization algorithm is: ； In the formula, represents the reward value at time , , , is a weight coefficient, is the click rate at time , is the cost per thousand impressions at time , is the diversity index of advertisement display at time ; the proximal policy optimization algorithm is: ； This formula is used to update the policy parameters of a reinforcement learning model , where represents the expected under the policy is the discount factor, is the time step, is the action taken at time , and is the state at time . In the near-end policy optimization algorithm, the policy parameters are adjusted according to the reward function, and the weight coefficient in the reward function is 、 、 According to the business focus and target of the advertiser, dynamic adjustment is carried out.

2. The method of claim 1, wherein, the acquisition of multi-source data covering the surrounding physical environment, semantic information and external association of the advertising position comprises: collecting people flow and heat distribution data around the advertising position; analyzing social media data to obtain semantic tag information through a natural language processing model; obtaining city heat map and traffic control information data.

3. The method of claim 1, wherein, the processing of the collected data using encryption technology, the encryption of the original data at the local node, and the implementation of cross-domain model aggregation using a federated learning algorithm comprise: deploying an edge node locally at the advertiser, and encrypting the original data using homomorphic encryption technology; after encrypting the original data at the local node, implementing cross-domain model aggregation through a FedProx algorithm, wherein the local update formula is: ； the global aggregation formula is: ； wherein, denotes the parameters of the th local model at the th training round, is the learning rate, is the loss function of the th local model, is a hyperparameter that controls how close the local model is to the global model, denotes the parameters of the global model at the th training round, is the number of local models participating in federated learning, is the weight of the th local model; In the FedProx algorithm, when performing cross-domain model aggregation, the proximity of the local model to the global model is controlled by adjusting the hyperparameter in the local model update formula .

4. The method of claim 1, wherein, the expansion of data diversity using a data enhancement algorithm and the identification of abnormal data points using an anomaly detection algorithm comprise: generating virtual traffic samples using a StyleGAN3 algorithm, adjusting the latent vector in the generator to generate virtual traffic samples under different weather and time periods for data enhancement; detecting abnormal traffic points using an IsolationForest algorithm; constructing a spatiotemporal feature cube and performing feature engineering processing on social media tag data through TF-IDF transformation and PCA dimension reduction.

5. The method of claim 1, wherein, the quantification of the causal effect of each data factor on the advertising position traffic based on a causal analysis framework, the display of feature contribution using visualization technology, the capture of the spatiotemporal characteristics and dynamic changes of the advertising position traffic using a hybrid neural network model, and the setting of an adaptive mechanism to ensure prediction reliability comprise: performing traffic prediction based on a CausalML framework and a hybrid prediction model, quantifying the causal effect of each factor on traffic using the CausalML framework, and visualizing feature contribution through SHAP values; the hybrid prediction model includes a spatiotemporal Transformer and a dynamic Bayesian network, which are used to capture the spatiotemporal characteristics and dynamic changes of the advertising position, and set an adaptive gating mechanism to switch to an expert system when the prediction confidence is lower than a set threshold; wherein the quantification of the causal effect of each factor on traffic uses a double difference method formula: ； wherein represents the average treatment effect, and represent the flow observations for the treatment and control groups, respectively, is an indicator variable that indicates whether or not a subject was treated, = 1 indicates that a subject was treated, = 0 indicates that a subject was not treated.

6. An advertisement inventory flow prediction system, comprising: which comprises: The data collection and privacy protection module is used to acquire multi-source data covering the physical environment, semantic information and external connections around the ad space, and to process the collected data using encryption technology. After encrypting the original data on the local node, the federated learning algorithm is used to achieve cross-domain model aggregation. The data preprocessing module is used to enhance data diversity using data augmentation algorithms and to identify outlier data points using anomaly detection algorithms. The predictive model design module is used to quantify the causal impact of various data factors on ad slot traffic based on a causal analysis framework, and to display the feature contribution with the help of visualization technology. It uses a hybrid model of fused neural networks to capture the spatiotemporal characteristics and dynamic changes of ad slot traffic, while setting an adaptive mechanism to ensure the reliability of the prediction. The dynamic decision output module is used to train the ad placement scheduling strategy using the near-end strategy optimization algorithm, with the objective function of maximizing advertiser revenue. At the same time, it builds a digital twin model to simulate traffic changes under different strategies and triggers a strategy rollback mechanism based on prediction error. The reward function formula for the near-end policy optimization algorithm is as follows: ； In the formula, represents the reward value at time , , , is a weight coefficient, is the click rate at time , is the cost per thousand impressions at time , is the diversity index of advertisement display at time ; The near-end strategy optimization algorithm is as follows: ； This formula is used to update the policy parameters of a reinforcement learning model. In the formula, Indicating in strategy The expectations below It is a discount factor. It is the time step. It is a moment The actions taken It is a moment The state; In the near-end policy optimization algorithm, the policy parameters are adjusted according to the reward function, and the weight coefficient in the reward function is 、 、 adjusted dynamically according to the business focus and target of the advertiser.

7. An electronic device, comprising: The electronic device includes: The processor and the memory are communicatively connected. The memory is used to store at least one executable instruction executed by the processor, the processor being used to execute the executable instruction to implement the ad slot traffic prediction method as described in any one of claims 1 to 5.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the ad placement traffic prediction method as described in any one of claims 1 to 5.

Images