Cross-platform advertisement putting method based on AI multi-advertisement account data

Abstract

The invention relates to the technical field of advertising and putting, and particularly discloses a cross-platform advertisement putting method based on AI multi-advertisement account data, comprising the following steps: S1, preprocessing account data based on user and platform authorization to obtain an analysis data set; s2, constructing a multi-dimensional feature analysis model based on deep learning, and training and verifying the model by using the preprocessed analysis data set; s3, constructing a budget allocation coefficient calculation model, and determining a budget allocation proportion of each platform; s4, dynamically adjusting an advertisement putting strategy of each platform according to the budget distribution result and the real-time putting data; and S5, performing quantitative evaluation on the delivery effect, and when an iteration condition is satisfied, retraining the AI model and adjusting the budget allocation and delivery strategy. According to the method, the technical problem that core features of multiple platforms cannot be quantitatively analyzed is solved by accurately capturing user behavior features, flow fluctuation rules and advertisement competition situations of different platforms.

Description

Technical Field

[0001] This invention relates to the field of advertising and promotion technology, specifically to a cross-platform advertising delivery method based on AI-powered multi-ad account data. Background Technology

[0002] With the rapid development of the digital marketing industry, advertising channels are becoming increasingly diversified, with mainstream advertising platforms covering various types such as social media platforms, search engine platforms, and e-commerce marketing platforms. To expand their marketing reach and reach different user groups, companies typically open advertising accounts on multiple platforms for joint campaigns. However, current multi-platform advertising faces significant technical challenges: user profiles differ significantly across advertising platforms, and user behavior characteristics (such as click preferences, conversion paths, and dwell time) exhibit platform-specific patterns; simultaneously, traffic supply on each platform is affected by factors such as holidays, platform activities, and industry competition, resulting in complex and inconsistent fluctuations; furthermore, the competitive landscape, including advertising bidding mechanisms and competitor strategy adjustments, also varies significantly across platforms.

[0003] Current advertising methods largely rely on manual experience or simple data statistical models for budget allocation and strategy formulation, failing to accurately capture user behavior characteristics, traffic fluctuation patterns, and competitive landscape across different platforms. This leads to irrational budget allocation across platforms, with insufficient budgets on high-potential platforms and wasted budgets on inefficient platforms. Ultimately, this results in inconsistent advertising performance, with some platforms experiencing excessively high conversion costs and insufficient exposure, while others suffer from low-quality traffic and low user conversion rates.

[0004] In existing technologies, some advertising optimization methods only adjust the placement strategy within a single platform, lacking collaborative analysis of data from multiple platforms. Other cross-platform placement methods only consider basic data such as traffic volume and click-through rates, failing to delve into the intrinsic relationship between user behavior characteristics and platform competitive landscape, and lacking scientific quantitative models for dynamic optimization of budget allocation and placement strategies. Therefore, there is an urgent need for a cross-platform advertising placement method that can integrate data from multiple advertising accounts, accurately analyze the core characteristics of multiple platforms through AI technology, and achieve reasonable budget allocation and dynamic optimization of placement strategies. Summary of the Invention

[0005] The purpose of this invention is to provide a cross-platform advertising method based on AI multi-ad account data, which can accurately capture user behavior characteristics, traffic fluctuation patterns and advertising competition trends on different platforms, realize the rational allocation of advertising budget and dynamic optimization of advertising strategies, and improve the overall effect of cross-platform advertising.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A cross-platform ad delivery method based on AI-powered multi-ad account data includes the following steps:

[0008] S1. Multi-source data collection and preprocessing:

[0009] Based on user and platform authorization, account data from various advertising platforms is obtained, including user behavior data, traffic data, advertising competition data, and campaign performance data. The collected multi-source data is preprocessed, including data cleaning, data standardization, and data integration, to obtain an analysis dataset in a unified format.

[0010] The user behavior data includes user click count, browsing duration, conversion count, and user profile tags; the traffic data includes platform daily active users, ad impressions, peak traffic distribution, and traffic fluctuation coefficient; the ad competition data includes competitors' ad volume, bidding price, and ad display space occupancy; and the ad performance data includes ad click-through rate (CTR), conversion rate (CVR), cost per click (CPC), and return on investment (ROI).

[0011] Data preprocessing specifically includes:

[0012] Data cleaning: Remove invalid data with more than 30% missing values, use the mean imputation method to supplement a small number of missing values, and use the Z-score method to identify and remove outlier data;

[0013] Data standardization: Standardizing data with different dimensions. The standardization formula is as follows:

[0014] X std =(X-μ) / σ;

[0015] Among them, X std The data are standardized, where X is the original data, μ is the mean of the data, and σ is the standard deviation of the data.

[0016] Data integration: Based on the time dimension (in hours) and platform dimension, the preprocessed user behavior data, traffic data, advertising competition data and campaign performance data are correlated and integrated to form an analysis dataset.

[0017] S2, AI Model Building and Training:

[0018] A multi-dimensional feature analysis model based on deep learning was constructed, including a user behavior feature extraction module, a traffic fluctuation prediction module, and an advertising competition situation assessment module; the model was trained and validated using a preprocessed analysis dataset.

[0019] S2.1 User Behavior Feature Extraction Model:

[0020] By employing a convolutional neural network (CNN) combined with an attention mechanism, key features of user behavior across different platforms are extracted, and a user behavior fit index U is defined. i This is used to quantify the match between platform users and advertisements, and the calculation method is as follows:

[0021] U i =α(C i / I i )+β(T i / T)+γ(Con) i / V i );

[0022] Among them, U i Let C be the user behavior adaptation index for the i-th platform. i Let I be the number of clicks by users on the i-th platform. i For the ad impressions on the i-th platform, T i Let T be the average browsing time per user on the i-th platform, and T be the mean average browsing time per user across all platforms. i V represents the number of user conversions on the i-th platform. i Let be the number of users who clicked on the i-th platform; α, β, and γ are weight coefficients, and α+β+γ=1. The optimal value is determined by grid search (the values ​​are all in the range of 0-1).

[0023] S2.2 Flow Fluctuation Prediction Model:

[0024] A traffic prediction model is built based on a Long Short-Term Memory (LSTM) network. It takes historical traffic data (including daily active users, impressions, and traffic fluctuation coefficients) as input and outputs traffic predictions for the next 72 hours. A traffic stability index F is defined. i The calculation method is as follows:

[0025] ;

[0026] Among them, F i Let L be the traffic stability index of the i-th platform, where n is the duration of historical data (in hours). i,t For the traffic data of the i-th platform in the t-th hour, Let F be the average of the historical traffic data of the i-th platform; i The value ranges from 0 to 1, with values ​​closer to 1 indicating more stable traffic.

[0027] S2.3 Advertising Competition Assessment Model:

[0028] The Gradient Boosting Tree (GBDT) model is used to analyze advertising competition data, and the competition intensity index K is defined. i The calculation method is as follows:

[0029] K i =δ(Comp) i / Comp total )+ε(Bid) i / Bid)+ζ(1-Pos i / Pos max );

[0030] Among them, K i Let Comp be the competition intensity index for the i-th platform. i Let Comp be the number of competitors for the i-th platform. total Bid represents the total number of competitors across all platforms. i Let be the average bid price on the i-th platform, be the mean of the average bid prices across all platforms, and be the average bid price on all platforms. i Pos represents the average number of ad placements on the i-th platform (the earlier the placement, the smaller the value). max The maximum display position value across all platforms; δ, ε, and ζ are weighting coefficients, and δ+ε+ζ=1. The optimal value is determined by cross-validation (the values ​​range from 0 to 1).

[0031] Model Training and Validation: The analysis dataset was divided into training and validation sets in a 7:3 ratio. The training set was used for model parameter optimization, and the validation set was used for model performance evaluation. Mean absolute error (MAE) and coefficient of determination (R²) were used. 2 As an evaluation indicator, when MAE ≤ 5% and R 2 When the value is ≥0.85, the model meets the accuracy requirements.

[0032] S3. Dynamic allocation of advertising budget:

[0033] User behavior fit index U based on AI model output i Flow stability index F i and the intensity of competition index K i A budget allocation coefficient calculation model was constructed to determine the budget allocation ratio for each platform. The specific calculation method is as follows:

[0034] S3.1 Platform Overall Score Calculation:

[0035] S i =U i ×F i ×(1−K i );

[0036] Among them, S i U represents the overall score of the i-th platform. i F is the user behavior adaptation index. i K is the traffic stability index. iThis is a competition intensity index; the higher the overall score, the greater the platform's advertising potential.

[0037] S3.2 Budget Allocation Ratio Calculation:

[0038] ;

[0039] Among them, B i The budget allocation ratio for the i-th platform, m is the total number of advertising platforms, and θ is the ROI adjustment factor (ranging from 0.1 to 0.3). i ′ represents the standardized return on investment for the i-th platform (obtained through the data standardization method in step S1); and satisfies .

[0040] S4. Dynamic optimization of delivery strategy:

[0041] Based on budget allocation results and real-time delivery data, we dynamically adjust advertising strategies for each platform, including optimizing delivery times, adjusting bidding prices, and adapting ad content.

[0042] S4.1 Delivery Time Optimization:

[0043] Based on the peak traffic periods output by the traffic fluctuation prediction module, and combined with the distribution of active time periods in user behavior data, the optimal delivery time for each platform is determined; a time period adaptation coefficient H is defined. i,t The calculation method is as follows:

[0044] H i,t =(L i,t ′×C i,t ) / (L i,total ′×C i,total );

[0045] Among them, H i,t L represents the time period adaptation coefficient for the i-th platform at hour t. i,t Let ' be the predicted traffic of the i-th platform in the t-th hour, and C be the predicted traffic. i,t L represents the historical click count of the i-th platform in hour t. i,total Let ′ be the total predicted traffic for the i-th platform over the next 72 hours, and C be the total predicted traffic for the i-th platform. i,total Let H be the total number of clicks in the historical 72 hours of the i-th platform; select H i,t The top 30% of time slots are considered the optimal time slots for ad placement.

[0046] S4.2 Bidding Price Adjustment:

[0047] Based on the competition intensity index K i Based on real-time bidding data, the bidding price is dynamically adjusted using the following formula:

[0048] Bid i,new =Bid i,current (1+λ(K)) i - ));

[0049] Among them, Bid i,new The adjusted bid price, Bid i,current The current bidding price is λ, and the bidding adjustment coefficient is 0.05-0.2. The average competition intensity index across all platforms; when K i > When Ki < At that time, the bidding price should be appropriately reduced to control costs.

[0050] S4.3 Ad Content Adaptation:

[0051] Based on the user profile tags obtained from the user behavior feature extraction module, appropriate advertising content is matched for different platforms; the content adaptability C is defined. con,i The calculation method is as follows:

[0052] ;

[0053] Among them, C con,i Let p be the ad content fit for the i-th platform, p be the total number of user profile tags, and W be the ad content fit for the i-th platform. k For the weight of the k-th label, Match i,k The matching degree between the i-th platform advertisement content and the k-th user tag (the value ranges from 0 to 1 and is calculated by a text similarity algorithm).

[0054] S5. Performance Monitoring and Model Iteration:

[0055] Real-time monitoring of advertising performance data across various platforms, including click-through rate, conversion rate, and ROI; using a comprehensive performance evaluation index E. i The effectiveness of campaigns on each platform was quantitatively evaluated, and the calculation method is as follows:

[0056] E i =a×CTR i ′+b×CVR i ′+c×ROI i ′;

[0057] Among them, E i CTR is the comprehensive evaluation index of the advertising performance on the i-th platform. i ′、CVR i ROI i′ represents the standardized click-through rate, conversion rate, and return on investment, respectively; a, b, and c are weighting coefficients, and a+b+c=1, which are set according to business objectives (for example, if conversion is emphasized, b takes a higher value).

[0058] When a certain platform's E i If the value is below 80% of the average for three consecutive days, step S2 is restarted to iteratively train the AI ​​model, update the model parameters and the calculation results of various indices, and then adjust the budget allocation ratio and the deployment strategy to form a closed-loop optimization mechanism.

[0059] The beneficial effects of this invention are as follows: By constructing an AI multi-dimensional feature analysis model, this invention accurately captures user behavior characteristics, traffic fluctuation patterns, and advertising competition trends across different platforms, solving the technical challenge of traditional methods being unable to quantify and analyze the core features of multiple platforms. This provides a scientific basis for budget allocation and strategy optimization. It offers a series of customizable quantitative calculation methods, including user behavior adaptability index, traffic stability index, and competition intensity index, enabling quantifiable analysis of multi-platform features, avoiding the limitations of subjective experience judgment, and improving the accuracy of advertising decisions. It establishes a dynamic budget allocation model and a closed-loop optimization mechanism for advertising strategies, which can adjust budget ratios, advertising periods, bidding prices, and advertising content based on real-time data and model iteration results, ensuring reasonable budget allocation and significantly improving the overall effect and return on investment of cross-platform advertising. The method has strong versatility and operability, applicable to various advertising platforms (social media, search engines, e-commerce platforms, etc.), and can flexibly adjust model parameters and weight coefficients according to the advertising needs of different industries and products, showing broad application prospects. Attached Figure Description

[0060] The invention will now be further described with reference to the accompanying drawings.

[0061] Figure 1 This is a flowchart illustrating a cross-platform advertising delivery method based on AI-powered multi-ad account data, according to the present invention. Detailed Implementation

[0062] 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.

[0063] Please see Figure 1 As shown, this invention is a cross-platform advertising delivery method based on AI-powered multi-ad account data, comprising the following steps:

[0064] S1. Multi-source data collection and preprocessing:

[0065] Based on user and platform authorization, account data from various advertising platforms is obtained, including user behavior data, traffic data, advertising competition data, and campaign performance data. The collected multi-source data is preprocessed, including data cleaning, data standardization, and data integration, to obtain an analysis dataset in a unified format.

[0066] The user behavior data includes user click count, browsing duration, conversion count, and user profile tags; the traffic data includes platform daily active users, ad impressions, peak traffic distribution, and traffic fluctuation coefficient; the ad competition data includes competitors' ad volume, bidding price, and ad display space occupancy; and the ad performance data includes ad click-through rate (CTR), conversion rate (CVR), cost per click (CPC), and return on investment (ROI).

[0067] Data preprocessing specifically includes:

[0068] Data cleaning: Remove invalid data with more than 30% missing values, use the mean imputation method to supplement a small number of missing values, and use the Z-score method to identify and remove outlier data;

[0069] Data standardization: Standardizing data with different dimensions. The standardization formula is as follows:

[0070] X std =(X-μ) / σ;

[0071] Among them, X std The data are standardized, where X is the original data, μ is the mean of the data, and σ is the standard deviation of the data.

[0072] Data integration: Based on the time dimension (in hours) and platform dimension, the preprocessed user behavior data, traffic data, advertising competition data and campaign performance data are correlated and integrated to form an analysis dataset.

[0073] S2, AI Model Building and Training:

[0074] A multi-dimensional feature analysis model based on deep learning was constructed, including a user behavior feature extraction module, a traffic fluctuation prediction module, and an advertising competition situation assessment module; the model was trained and validated using a preprocessed analysis dataset.

[0075] S2.1 User Behavior Feature Extraction Model:

[0076] By employing a convolutional neural network (CNN) combined with an attention mechanism, key features of user behavior across different platforms are extracted, and a user behavior fit index U is defined. iThis is used to quantify the match between platform users and advertisements, and the calculation method is as follows:

[0077] U i =α(C i / I i )+β(T i / T)+γ(Con) i / V i );

[0078] Among them, U i Let C be the user behavior adaptation index for the i-th platform. i Let I be the number of clicks by users on the i-th platform. i For the ad impressions on the i-th platform, T i Let T be the average browsing time per user on the i-th platform, and T be the mean average browsing time per user across all platforms. i V represents the number of user conversions on the i-th platform. i Let be the number of users who clicked on the i-th platform; α, β, and γ are weight coefficients, and α+β+γ=1. The optimal value is determined by grid search (the values ​​are all in the range of 0-1).

[0079] S2.2 Flow Fluctuation Prediction Model:

[0080] A traffic prediction model is built based on a Long Short-Term Memory (LSTM) network. It takes historical traffic data (including daily active users, impressions, and traffic fluctuation coefficients) as input and outputs traffic predictions for the next 72 hours. A traffic stability index F is defined. i The calculation method is as follows:

[0081] ;

[0082] Among them, F i Let L be the traffic stability index of the i-th platform, where n is the duration of historical data (in hours). i,t For the traffic data of the i-th platform in the t-th hour, Let F be the average of the historical traffic data of the i-th platform; i The value ranges from 0 to 1, with values ​​closer to 1 indicating more stable traffic.

[0083] S2.3 Advertising Competition Assessment Model:

[0084] The Gradient Boosting Tree (GBDT) model is used to analyze advertising competition data, and the competition intensity index K is defined. i The calculation method is as follows:

[0085] K i =δ(Comp) i / Comp total )+ε(Bid)i / Bid)+ζ(1-Pos i / Pos max );

[0086] Among them, K i Let Comp be the competition intensity index for the i-th platform. i Let Comp be the number of competitors for the i-th platform. total Bid represents the total number of competitors across all platforms. i Let be the average bid price on the i-th platform, be the mean of the average bid prices across all platforms, and be the average bid price on all platforms. i Pos represents the average number of ad placements on the i-th platform (the earlier the placement, the smaller the value). max The maximum display position value across all platforms; δ, ε, and ζ are weighting coefficients, and δ+ε+ζ=1. The optimal value is determined by cross-validation (the values ​​range from 0 to 1).

[0087] Model Training and Validation: The analysis dataset was divided into training and validation sets in a 7:3 ratio. The training set was used for model parameter optimization, and the validation set was used for model performance evaluation. Mean absolute error (MAE) and coefficient of determination (R²) were used. 2 As an evaluation indicator, when MAE ≤ 5% and R 2 When the value is ≥0.85, the model meets the accuracy requirements.

[0088] S3. Dynamic allocation of advertising budget:

[0089] User behavior fit index U based on AI model output i Flow stability index F i and the intensity of competition index K i A budget allocation coefficient calculation model was constructed to determine the budget allocation ratio for each platform. The specific calculation method is as follows:

[0090] S3.1 Platform Overall Score Calculation:

[0091] S i =U i ×F i ×(1−K i );

[0092] Among them, S i U represents the overall score of the i-th platform. i F is the user behavior adaptation index. i K is the traffic stability index. i This is a competition intensity index; the higher the overall score, the greater the platform's advertising potential.

[0093] S3.2 Budget Allocation Ratio Calculation:

[0094] ;

[0095] Among them, B i The budget allocation ratio for the i-th platform, m is the total number of advertising platforms, and θ is the ROI adjustment factor (ranging from 0.1 to 0.3). i ′ represents the standardized return on investment for the i-th platform (obtained through the data standardization method in step S1); and satisfies .

[0096] S4. Dynamic optimization of delivery strategy:

[0097] Based on budget allocation results and real-time delivery data, we dynamically adjust advertising strategies for each platform, including optimizing delivery times, adjusting bidding prices, and adapting ad content.

[0098] S4.1 Delivery Time Optimization:

[0099] Based on the peak traffic periods output by the traffic fluctuation prediction module, and combined with the distribution of active time periods in user behavior data, the optimal delivery time for each platform is determined; a time period adaptation coefficient H is defined. i,t The calculation method is as follows:

[0100] H i,t =(L i,t ′×C i,t ) / (L i,total ′×C i,total );

[0101] Among them, H i,t L represents the time period adaptation coefficient for the i-th platform at hour t. i,t Let ' be the predicted traffic of the i-th platform in the t-th hour, and C be the predicted traffic. i,t L represents the historical click count of the i-th platform in hour t. i,total Let ′ be the total predicted traffic for the i-th platform over the next 72 hours, and C be the total predicted traffic for the i-th platform. i,total Let H be the total number of clicks in the historical 72 hours of the i-th platform; select H i,t The top 30% of time slots are considered the optimal time slots for ad placement.

[0102] S4.2 Bidding Price Adjustment:

[0103] Based on the competition intensity index K i Based on real-time bidding data, the bidding price is dynamically adjusted using the following formula:

[0104] Bid i,new =Bid i,current (1+λ(K)) i - ));

[0105] Among them, Bid i,new The adjusted bid price, Bid i,current The current bidding price is λ, and the bidding adjustment coefficient is 0.05-0.2. The average competition intensity index across all platforms; when K i > When Ki < At that time, the bidding price should be appropriately reduced to control costs.

[0106] S4.3 Ad Content Adaptation:

[0107] Based on the user profile tags obtained from the user behavior feature extraction module, appropriate advertising content is matched for different platforms; the content adaptability C is defined. con,i The calculation method is as follows:

[0108] ;

[0109] Among them, C con,i Let p be the ad content fit for the i-th platform, p be the total number of user profile tags, and W be the ad content fit for the i-th platform. k For the weight of the k-th label, Match i,k The matching degree between the i-th platform advertisement content and the k-th user tag (the value ranges from 0 to 1 and is calculated by a text similarity algorithm).

[0110] S5. Performance Monitoring and Model Iteration:

[0111] Real-time monitoring of advertising performance data across various platforms, including click-through rate, conversion rate, and ROI; using a comprehensive performance evaluation index E. i The effectiveness of campaigns on each platform was quantitatively evaluated, and the calculation method is as follows:

[0112] E i =a×CTR i ′+b×CVR i ′+c×ROI i ′;

[0113] Among them, E i CTR is the comprehensive evaluation index of the advertising performance on the i-th platform. i ′、CVR i ROI i ′ represents the standardized click-through rate, conversion rate, and return on investment, respectively; a, b, and c are weighting coefficients, and a+b+c=1, which are set according to business objectives (for example, if conversion is emphasized, b takes a higher value).

[0114] When a certain platform's Ei If the value is below 80% of the average for three consecutive days, step S2 is restarted to iteratively train the AI ​​model, update the model parameters and the calculation results of various indices, and then adjust the budget allocation ratio and the deployment strategy to form a closed-loop optimization mechanism.

[0115] This invention constructs an AI-powered multi-dimensional feature analysis model to accurately capture user behavior characteristics, traffic fluctuation patterns, and advertising competition across different platforms. It solves the technical challenge of traditional methods failing to quantify the core characteristics of multiple platforms, providing a scientific basis for budget allocation and strategy optimization. It offers a series of customizable quantitative calculation methods, including user behavior fit index, traffic stability index, and competition intensity index, enabling quantifiable analysis of multi-platform characteristics, avoiding the limitations of subjective experience and improving the accuracy of advertising decisions. Furthermore, it establishes a dynamic budget allocation model and a closed-loop optimization mechanism for advertising strategies, adjusting budget proportions, advertising time periods, bidding prices, and ad content based on real-time data and model iteration results. This ensures reasonable budget allocation and significantly improves the overall effectiveness and ROI of cross-platform advertising.

[0116] The foregoing has provided a detailed description of one embodiment of the present invention, but this description is merely a preferred embodiment and should not be construed as limiting the scope of the invention. All equivalent variations and modifications made within the scope of the present invention should still fall within the scope of the present invention.

Claims

1. A cross-platform advertising delivery method based on AI-powered multi-ad account data, characterized in that, Includes the following steps: S1. Based on user and platform authorization, obtain account data from various advertising platforms, including user behavior data, traffic data, advertising competition data, and campaign performance data; preprocess the collected multi-source data to obtain an analysis dataset in a unified format; S2. Construct a multi-dimensional feature analysis model based on deep learning, including a user behavior feature extraction model, a traffic fluctuation prediction model, and an advertising competition situation assessment model; train and validate the model using the preprocessed analysis dataset; S3. Based on the user behavior adaptability index, traffic stability index and competition intensity index output by the AI ​​model, construct a budget allocation coefficient calculation model to determine the budget allocation ratio of each platform; S4. Based on the budget allocation results and real-time delivery data, dynamically adjust the advertising delivery strategies of each platform, including optimizing delivery time, adjusting bidding prices, and adapting advertising content. S5 monitors advertising performance data on various platforms in real time, quantifies the performance, and retrains the AI ​​model and adjusts budget allocation and advertising strategies when iteration conditions are met.

2. The cross-platform advertising delivery method based on AI multi-ad account data according to claim 1, characterized in that, In step S2, the user behavior feature extraction model specifically includes the following components in constructing a deep learning-based multi-dimensional feature analysis model: By employing a convolutional neural network combined with an attention mechanism, key features of user behavior across different platforms are extracted, and a user behavior fit index U is defined. i This is used to quantify the match between platform users and advertisements, and the calculation method is as follows: U i =α(C i / I i )+β(T i / T)+γ(Con i / V i ); Among them, U i Let C be the user behavior adaptation index for the i-th platform. i Let I be the number of clicks by users on the i-th platform. i For the ad impressions on the i-th platform, T i Let T be the average browsing time per user on the i-th platform, and T be the mean average browsing time per user across all platforms. i V represents the number of user conversions on the i-th platform. i Let be the number of users who clicked on the i-th platform; α, β, and γ are weighting coefficients, and α+β+γ=1.

3. The cross-platform advertising delivery method based on AI multi-ad account data according to claim 2, characterized in that, In step S2, the traffic fluctuation prediction model specifically includes the following components in constructing a multi-dimensional feature analysis model based on deep learning: A traffic prediction model is constructed based on a Long Short-Term Memory (LSTM) network. It takes historical traffic data as input and outputs predicted traffic values ​​for the next 72 hours. A traffic stability index F is defined. i The calculation method is as follows: ； Among them, F i Let L be the traffic stability index of the i-th platform, n be the duration of historical data, and L be the length of the data. i,t For the traffic data of the i-th platform in the t-th hour, Let F be the average of the historical traffic data of the i-th platform; i The value ranges from 0 to 1, with values ​​closer to 1 indicating more stable traffic.

4. The cross-platform advertising delivery method based on AI multi-ad account data according to claim 3, characterized in that, In step S2, the advertising competition situation assessment model specifically includes the following components in constructing a multi-dimensional feature analysis model based on deep learning: The gradient boosting tree model is used to analyze advertising competition data, and the competition intensity index K is defined. i The calculation method is as follows: K i =δ(Comp i / Comp total )+ε(Bid i / Bid)+ζ(1-Pos i / Pos max ); Among them, K i Let Comp be the competition intensity index for the i-th platform. i Let Comp be the number of competitors for the i-th platform. total Bid represents the total number of competitors across all platforms. i Let be the average bid price on the i-th platform, be the mean of the average bid prices across all platforms, and be the average bid price on all platforms. i Pos represents the average ad placement value on the i-th platform. The earlier the ad placement, the smaller the value. max The maximum display position value across all platforms; δ, ε, and ζ are weighting coefficients, and δ+ε+ζ=1.

5. A cross-platform advertising delivery method based on AI multi-ad account data according to claim 4, characterized in that, In step S3, constructing a budget allocation coefficient calculation model and determining the budget allocation ratio for each platform specifically includes: User behavior fit index U based on AI model output i Flow stability index F i and the intensity of competition index K i A budget allocation coefficient calculation model was constructed to determine the budget allocation ratio for each platform. The specific calculation method is as follows: S3.1 Calculation of Platform Overall Score: S i =U i ×F i ×(1−K i ); Among them, S i U represents the overall score of the i-th platform. i F is the user behavior adaptation index. i K is the traffic stability index. i This is a competition intensity index; the higher the overall score, the greater the platform's advertising potential. S3.2 Calculation of Budget Allocation Ratio: ； Among them, B i The budget allocation percentage for the i-th platform, where m is the total number of advertising platforms, j represents the index of the advertising platform, and θ is the ROI adjustment factor, ranging from 0.1 to 0.

3. i ′ is the standardized return on investment for the i-th platform; and satisfies .

6. A cross-platform advertising delivery method based on AI multi-ad account data according to claim 5, characterized in that, Step S4 specifically includes: Based on the peak traffic periods output by the traffic fluctuation prediction module, and combined with the distribution of active time periods in user behavior data, the optimal delivery time for each platform is determined. Define the time period adaptation coefficient H i,t The calculation method is as follows: H i,t =(L) i,t ×C i,t ) / (L i,total ×C i,total ); Among them, H i,t L represents the time period adaptation coefficient for the i-th platform at hour t. i,t Let ' be the predicted traffic of the i-th platform in the t-th hour, and C be the predicted traffic. i,t L represents the historical click count of the i-th platform in hour t. i,total Let ′ be the total predicted traffic for the i-th platform over the next 72 hours, and C be the total predicted traffic for the i-th platform. i,total The total number of clicks on the i-th platform over the past 72 hours; Select H i,t The top 30% of time slots are considered the optimal time slots for ad placement.

7. A cross-platform advertising delivery method based on AI multi-ad account data according to claim 6, characterized in that, Step S4 further includes: Based on the competition intensity index K i Based on real-time bidding data, the bidding price is dynamically adjusted using the following formula: Bid i,new =Bid i,current (1+λ(K i - )); Among them, Bid i,new The adjusted bid price, Bid i,current The current bidding price is λ, which is the bidding adjustment coefficient, ranging from 0.05 to 0.

2. The average competition intensity index across all platforms; when K i > When Ki < At that time, lower the bidding price; Based on the user profile tags obtained from the user behavior feature extraction module, appropriate advertising content is matched for different platforms; Define content adaptability C con,i The calculation method is as follows: ； Among them, C con,i Let p be the ad content fit for the i-th platform, p be the total number of user profile tags, and W be the ad content fit for the i-th platform. k For the weight of the k-th label, Match i,k Let represent the matching degree between the i-th platform advertisement content and the k-th user tag, with a value ranging from 0 to 1.

8. A cross-platform advertising delivery method based on AI multi-ad account data according to claim 7, characterized in that, Step S5 specifically includes: The comprehensive evaluation index E for campaign effectiveness is adopted. i The effectiveness of campaigns on each platform was quantitatively evaluated, and the calculation method is as follows: AND i =a×CTR i ′+b×CVR i ′+c×ROI i '; Among them, E i CTR is the comprehensive evaluation index of the advertising performance on the i-th platform. i ′、CVR i ROI i ′ represent the standardized click-through rate, conversion rate, and return on investment, respectively; a, b, and c are weighting coefficients, and a+b+c=1; When a certain platform's E i If the value is below the preset threshold for three consecutive days, step S2 is restarted to iteratively train the AI ​​model, update the model parameters and the calculation results of various indices, and then adjust the budget allocation ratio and the deployment strategy to form a closed-loop optimization mechanism.

Images