Video advertisement playing analysis management platform and management method

Abstract

The invention discloses a video advertisement playing analysis management platform and management method, and belongs to the field of video advertisement playing analysis, and the platform comprises an information extraction module, an information integration module, a push strategy making module, a strategy execution module and a strategy optimization module. According to the video advertisement playing analysis management platform and management method, the maturity of a user account is judged, accurate pushing is achieved for low-maturity users based on region, age and gender matching group preference data, and a trigger detection mechanism is set to avoid data lag; for high-maturity users, a recommendation database is simplified through similarity analysis, and a strategy generation process is simplified; a strategy is dynamically corrected by combining a prediction model and an actual effect difference value, and the putting efficiency is optimized; according to the scheme, the problems of insufficient new user data, strategy lagging, model staticizing and the like are solved, and the advertisement putting accuracy and the resource utilization rate are improved.

Description

Technical Field

[0001] The present invention belongs to the field of video advertisement playback analysis, and in particular provides a video advertisement playback analysis management platform and a management method. Background Art

[0002] In the field of video advertising, existing technologies face many challenges: due to insufficient accumulation of behavioral data for new users, advertising recommendations lack accuracy and are prone to blind push; traditional recommendation platforms mostly use static models, which are difficult to respond to changes in user behavior in a timely manner, have data lag problems, and frequent calculations easily consume system resources; for mature users, existing strategies often have low efficiency in generating delivery strategies due to database redundancy and extensive screening mechanisms, making it difficult to accurately match users' dynamic preferences; in addition, the prediction model of delivery effects lacks a dynamic correction mechanism, and the accumulated deviation between actual effects and predictions can easily lead to resource waste and reduced delivery efficiency. In response to the above problems, the present invention proposes a video advertising playback analysis and management platform and method based on user account maturity. Summary of the Invention

[0003] The present invention aims to address at least one of the technical problems existing in the prior art. To this end, the present invention proposes a video advertising playback analysis and management platform and method, enabling relevant personnel to conveniently manage video advertising, avoiding issues such as blind push notifications due to insufficient new user data, policy updates lagging behind user behavior changes, crude preference matching for mature users, and accumulated deviations in delivery effect predictions.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] A video advertising playback analysis and management platform and management method, including an information extraction module, an information integration module, a push strategy formulation module, a strategy execution module, and a strategy optimization module;

[0006] The information collection module is used to collect basic data on advertising materials, playback, user interaction and users;

[0007] The information integration module is used to integrate and analyze the advertising preferences of users in different regions, ages, and genders;

[0008] The strategy formulation module receives preference-related indicators and user account age data, establishes a recommendation database, matches different advertisements based on the recommendation database and performs creative combinations, generates multiple delivery strategies, establishes a prediction model to estimate delivery effects, and selects the strategy with the best delivery effect;

[0009] The strategy execution module is used to execute the delivery strategy;

[0010] The strategy optimization module analyzes strategy performance patterns, anomalies, and influencing factors based on playback and user interaction data, calculates the error between actual and predicted delivery effects, and automatically corrects the prediction model and resets the delivery strategy when the error exceeds a threshold.

[0011] By calculating the error between actual and predicted delivery effects and automatically correcting the prediction model and resetting the delivery strategy when the threshold is exceeded, dynamic iteration and precise calibration of the advertising delivery strategy can be achieved.

[0012] Furthermore, the specific processing steps for collecting advertising materials, playback, user interaction and basic user data are as follows:

[0013] Collect relevant information of various video advertising materials uploaded to the platform, including advertising duration, resolution, format, and theme content; obtain playback data of video ads on various playback channels, including number of plays, playback time, and playback position; record user interaction with video ads, including clicks, pauses, and shares; collect basic user data, including user age, user gender, user region, and user account length.

[0014] Furthermore, the information integration module is used to integrate and analyze the advertising preferences of users of different regions, ages, and genders. The specific processing steps are as follows:

[0015] Clean and standardize the data. After processing, encode the categorical data including region, gender, and age and convert them into numerical form.

[0016] With user accounts as the core associated data, basic user data is associated with ad creative related information, playback data, and user interaction behavior data to build a comprehensive data set;

[0017] The comprehensive dataset is grouped by user region, age, and gender. For each group, metrics related to ad preferences are calculated. These metrics include the total number of clicks, shares, and plays on ads of different topics by users of different regions, ages, and genders, as well as the average interaction time between users of different groups and ads.

[0018] Calculate the completed metrics to find out the most popular ad duration, resolution, format, and theme preferences in each group and use this as the analysis result.

[0019] Furthermore, the strategy formulation module receives preference-related indicators and user account age data, establishes a recommendation database, matches different advertisements based on the recommendation database and performs creative combinations, generates multiple delivery strategies, and establishes a prediction model to estimate delivery effects. The process of selecting the strategy with the best delivery effect is as follows:

[0020] Determine the maturity of the user account. If the user has used the account for less than 7 days and there is insufficient data on the user's preference-related indicators, the account is considered to have low maturity. Otherwise, the account is considered to have high maturity.

[0021] For low-maturity accounts, we match the corresponding groups from the comprehensive data set based on the user's region, age, and gender information, extract the analysis results of the group, and establish a recommendation database, storing the analysis results in the recommendation database;

[0022] Confirm the total playable duration of the ad and set a ratio to break down the total duration. Based on the recommended database content and the individual broken down durations, match the required ads in the ad library and combine the ads to generate multiple delivery strategies.

[0023] Establish a prediction model to estimate the effect of the delivery and select the best delivery strategy;

[0024] For low-maturity accounts, a trigger detection mechanism is set up, including: the user has accumulated 5 valid interactions; the 7th, 15th, and 30th days of user registration; the user's single-day interaction behavior exceeds the historical average by 300%;

[0025] After the detection is triggered, if the user account maturity is determined to be high, the user preferences are analyzed and a similarity analysis is performed between the user preference data and the data in the recommendation database. Data with a similarity of less than 0.15 is eliminated. Data with a similarity between 0.15 and 0.4 is soft-demoted to a secondary cache pool to retain the right of emergency recall. Data with a similarity greater than 0.4 is retained to reduce the database and further accurately understand user preferences. Finally, based on the content of the recommendation database and the duration of the ads, the ad library is matched with ads that meet the requirements, and the ads are combined to generate multiple delivery strategies.

[0026] When the user account maturity is first determined to be high, the user's region, age, gender information and personal preference data are directly combined to establish a recommendation database. Finally, based on the content of the recommendation database and the length of the advertisement, the advertisements that meet the requirements are matched in the advertisement library, and the advertisements are combined to generate multiple delivery strategies.

[0027] Furthermore, the specific process for determining whether the preference-related indicator data is insufficient is as follows:

[0028] First, verify the integrity of data types, requiring users to play at least three times and trigger at least one actual interaction, including clicks, pauses, and shares. Second, verify key behavior coverage, confirming that users are exposed to at least two types of advertising themes and their behaviors are distributed on at least three different dates, and that interactions occur in at least two network scenarios, either WiFi or data. Finally, calculate the behavior density threshold using the following formula:

[0029]

[0030] Among them, E is the calculated data adequacy index, N is the industry's minimum effective sample size, which is 5, and t is the number of days between the most recent and first behaviors, with an upper limit of 7. When E is less than 0.8, it means that there is insufficient data.

[0031] Furthermore, the process of establishing a prediction model to estimate the delivery effect and selecting the best delivery effect strategy is as follows:

[0032] Evaluate the impact of ad images, compare similarity with historically high-click materials, track topic popularity trends, overlay real-time search indexes, and combine historical click fluctuations of target demographic groups to ultimately calculate the expected click probability.

[0033] Analyze the rhythm density in the strategy, that is, the number of seconds between key shots, combined with the number of plot turning points, and simultaneously establish a curve of the average attention span of the target group. Match it with the curve of the average attention span of the target group to calculate the probability of the ad being played in its entirety;

[0034] Verify whether the conversion trigger design aligns with the group's behavioral habits. Specifically, analyze whether the positioning button position matches the group's click hotspot coordinates, and whether the intensity of the time-limited urgency is within the optimal range. Output the conversion rate value.

[0035] A prediction model is established to perform comprehensive scoring, which is used as the estimated delivery effect. The specific formula is:

[0036] G=C·40%+V·30%+R·25%+(1-T)·5;

[0037] Where G is the estimated delivery effect after calculation, C is the expected click probability, V is the probability of complete playback, R is the conversion rate value, and T is the fatigue factor, which ranges from [0 to 1]. A value of 0 indicates that the strategy elements are 100% innovative, and a value of 1 indicates that all elements completely repeat recent content.

[0038] The strategy with the comprehensive score is used as the strategy with the best delivery effect and the strategy is output.

[0039] Through a comprehensive scoring formula, the indicators of each dimension are quantified into estimated effects. This process can not only predict the effect differences of different delivery strategies in advance, but also maximize the delivery effect and optimize the allocation of resources.

[0040] Furthermore, the strategy optimization module calculates the difference between the actual delivery effect and the predicted delivery effect based on the playback and user interaction data. When the difference exceeds a threshold, the prediction model is automatically corrected and the delivery strategy is reset. The processing process is as follows:

[0041] Based on playback data and user interaction behavior, the difference between the actual delivery effect and the predicted delivery effect is calculated, and a threshold is preset. When the difference exceeds the threshold, the key differences between the predicted effect and the actual delivery effect are compared, including the number of clicks, the number of complete plays, and the conversion rate, to locate the key differences.

[0042] The gradient attenuation formula is used to correct the weight of the key difference results. The specific formula is:

[0043] ΔW i =0.02·SHAP i ln(1+|Δ|);

[0044] Where ΔW i is the calculated weight adjustment, SHAP i is the attribution contribution of feature i, ln(1+|Δ|) is the logarithmic decay function;

[0045] Based on the recommendation database content and ad duration, ads that meet the requirements are matched in the ad library, and the ads are combined again to generate multiple delivery strategies. The updated prediction model is used to predict the effect and the strategy with the best delivery effect is selected for output.

[0046] A method for managing a video advertising playback analysis and management platform, the system using any one of the above video advertising playback analysis and management platforms, comprising the following steps:

[0047] S1. Collect advertising material data, playback data, user interaction data, and basic user data;

[0048] S2. Conduct integrated analysis of advertising preferences for users of different regions, ages, and genders;

[0049] S3. Establish a recommendation database based on known data and generate the best delivery strategy based on the recommendation database;

[0050] S4. Execute the delivery strategy and analyze the predicted effect of the delivery strategy and the actual delivery effect. When the difference exceeds the threshold, correct the prediction model and reset the delivery strategy.

[0051] Compared with the prior art, the present invention has the following beneficial effects:

[0052] This invention determines the maturity of user accounts and recommends advertisements to users with low maturity by matching the group preference data of corresponding groups from a comprehensive data set based on their region, age, and gender information. This method can make up for the lack of individual data by leveraging the common behavioral characteristics of users of the same dimension, and quickly build an initial recommendation strategy framework. This enables the platform to accurately identify potential preferences of users even when their data is limited, avoiding push blindness caused by data scarcity and improving the efficiency of initial matching between advertisements and users.

[0053] By setting up a trigger detection mechanism for low-maturity accounts, the present invention can proactively initiate detection when user behavior data accumulates to a critical point, avoiding the consumption of system resources due to frequent repeated calculations. At the same time, it ensures that user maturity judgment and recommendation strategies are updated in a timely manner when user behavior patterns change significantly or when data accumulates to a sufficient scale, effectively solving the problem of data lag. This allows the recommendation strategy to maintain sensitivity to changes in user behavior while optimizing computing efficiency, achieving the dual goals of dynamic response and resource conservation.

[0054] After the detection mechanism is triggered, if the user account is judged to be highly mature, the user's preference data is analyzed and similarity analysis is performed on the recommendation database, and data with low similarity is eliminated. This mechanism dynamically reduces the size of the recommendation database through refined screening, reducing the database content while skipping redundant data processing. This significantly simplifies the operation steps, making the delivery strategy generation more efficient and accurately matching the user's true preferences.

[0055] The present invention calculates the difference between the actual delivery effect and the predicted effect, compares key differences such as click volume, complete playback volume, conversion rate, etc. when the threshold is exceeded, and uses the gradient decay formula to correct the model weight after locating the problem. It re-matches advertisements and generates strategies based on the updated recommendation database. It can accurately capture changes in user behavior and market feedback, and promptly correct strategy deviations caused by the static nature of the model, avoid the decline in delivery efficiency caused by the accumulation of prediction errors, and enable the prediction model and delivery strategy to be continuously optimized according to the actual effect, thereby achieving the maximization of delivery effect and the optimal allocation of resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a block diagram of a video advertisement playback analysis and management platform of the present invention. DETAILED DESCRIPTION

[0057] The technical solutions of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] like Figure 1 As shown, a video advertising playback analysis and management platform includes an information extraction module, an information integration module, a push strategy formulation module, a strategy execution module, and a strategy optimization module;

[0059] Information collection module, used to collect advertising materials, playback, user interaction and basic user data;

[0060] In this embodiment, the information collection module is used to collect advertising materials, playback, user interaction and basic user data. The specific processing steps are as follows:

[0061] Collect relevant information of various video advertising materials uploaded to the platform, including advertising duration, resolution, format, and theme content; obtain playback data of video ads on various playback channels, including number of plays, playback time, and playback position; record user interaction with video ads, including clicks, pauses, and shares; collect basic user data, including user age, user gender, user region, and user account length.

[0062] Information integration module, used to integrate and analyze the advertising preferences of users in different regions, ages, and genders;

[0063] In this embodiment, the information integration module is used to integrate and analyze the advertising preferences of users of different regions, ages, and genders. The specific processing steps are as follows:

[0064] Clean and standardize the data. After processing, encode the categorical data including region, gender, and age and convert them into numerical form.

[0065] With user accounts as the core associated data, basic user data is associated with ad creative related information, playback data, and user interaction behavior data to build a comprehensive data set;

[0066] The comprehensive dataset is grouped by user region, age, and gender. For each group, metrics related to ad preferences are calculated. These metrics include the total number of clicks, shares, and plays on ads of different topics by users of different regions, ages, and genders, as well as the average interaction time between users of different groups and ads.

[0067] Calculate the completed metrics to find out the most popular ad duration, resolution, format, and theme preferences in each group and use this as the analysis result.

[0068] The strategy formulation module receives preference-related indicators and user account age data, establishes a recommendation database, matches different ads based on the recommendation database, and creates creative combinations to generate multiple delivery strategies. It also establishes a prediction model to estimate delivery effects and selects the strategy with the best delivery effect.

[0069] In this embodiment, the strategy formulation module receives preference-related indicators and user account age data, establishes a recommendation database, matches different advertisements based on the recommendation database, and performs creative combinations to generate multiple delivery strategies. It also establishes a prediction model to estimate delivery effects. The process of selecting the strategy with the best delivery effect is as follows:

[0070] Determine the maturity of the user account. If the user has used the account for less than 7 days and there is insufficient data on the user's preference-related indicators, the account is considered to have low maturity. Otherwise, the account is considered to have high maturity.

[0071] For low-maturity accounts, we match the corresponding groups from the comprehensive data set based on the user's region, age, and gender information, extract the analysis results of the group, and establish a recommendation database, storing the analysis results in the recommendation database;

[0072] It should be noted that for low-maturity accounts, advertising recommendations are made based on group preference data corresponding to their region, age, and gender. This can make up for the lack of individual data by leveraging the common behavioral characteristics of users in the same dimension, and quickly build an initial recommendation strategy framework.

[0073] Confirm the total playable duration of the ad and set a ratio to break down the total duration. Based on the recommended database content and the individual broken down durations, match the required ads in the ad library and combine the ads to generate multiple delivery strategies.

[0074] Establish a prediction model to estimate the effect of the delivery and select the best delivery strategy;

[0075] For low-maturity accounts, a trigger detection mechanism is set up, including: the user has accumulated 5 valid interactions; the 7th, 15th, and 30th days of user registration; the user's single-day interaction behavior exceeds the historical average by 300%;

[0076] After the detection is triggered, if the user account maturity is determined to be high, the user preferences are analyzed and a similarity analysis is performed between the user preference data and the data in the recommendation database. Data with a similarity of less than 0.15 is eliminated. Data with a similarity between 0.15 and 0.4 is soft-demoted to a secondary cache pool to retain the right of emergency recall. Data with a similarity greater than 0.4 is retained to reduce the database and further accurately understand user preferences. Finally, based on the content of the recommendation database and the duration of the ads, the ad library is matched with ads that meet the requirements, and the ads are combined to generate multiple delivery strategies.

[0077] It should be noted that the preference data similarity analysis and database optimization mechanism implemented for high-maturity accounts after triggering detection can achieve dynamic lightweighting and precise iteration of the recommendation database through refined screening. This can not only improve the accuracy of user preference matching by reducing data granularity, but also take into account the flexibility and emergency response capabilities of the recommendation strategy through the secondary cache pool mechanism. At the same time, it can also simplify the operation steps and effectively improve the fit between the delivery scenario and the user's actual preferences.

[0078] When the user account is first judged to be mature, the user's region, age, gender information and personal preference data are directly combined to establish a recommendation database. Finally, based on the content of the recommendation database and the duration of the advertisement, the advertisement library is matched with the required advertisements, and the advertisements are combined to generate multiple delivery strategies.

[0079] In this embodiment, the specific process of determining if preference-related indicator data is insufficient is as follows:

[0080] First, verify the integrity of data types, requiring users to play at least three times and trigger at least one actual interaction, including clicks, pauses, and shares. Second, verify key behavior coverage, confirming that users are exposed to at least two types of advertising themes and their behaviors are distributed on at least three different dates, and that interactions occur in at least two network scenarios, either WiFi or data. Finally, calculate the behavior density threshold using the following formula:

[0081]

[0082] Among them, E is the calculated data adequacy index, N is the industry's minimum effective sample size, which is 5, and t is the number of days between the most recent and first behaviors, with an upper limit of 7. When E is less than 0.8, it means that there is insufficient data.

[0083] In this embodiment, the process of establishing a prediction model to estimate the delivery effect and selecting the best delivery effect strategy is as follows:

[0084] Evaluate the impact of ad images, compare similarity with historically high-click materials, track topic popularity trends, overlay real-time search indexes, and combine historical click fluctuations of target demographic groups to ultimately calculate the expected click probability.

[0085] Analyze the rhythm density in the strategy, that is, the number of seconds between key shots, combined with the number of plot turning points, and simultaneously establish a curve of the average attention span of the target group. Match it with the curve of the average attention span of the target group to calculate the probability of the ad being played in its entirety;

[0086] Verify whether the conversion trigger design aligns with the group's behavioral habits. Specifically, analyze whether the positioning button position matches the group's click hotspot coordinates, and whether the intensity of the time-limited urgency is within the optimal range. Output the conversion rate value.

[0087] A prediction model is established to perform comprehensive scoring, which is used as the estimated delivery effect. The specific formula is:

[0088] G=C·40%+V·30%+R·25%+(1-T)·5;

[0089] Where G is the estimated delivery effect after calculation, C is the expected click probability, V is the probability of complete playback, R is the conversion rate value, and T is the fatigue factor, which ranges from [0 to 1]. A value of 0 indicates that the strategy elements are 100% innovative, and a value of 1 indicates that all elements completely repeat recent content.

[0090] It should be noted that since the fatigue factor takes values ​​in the range [0,1], 5% was changed to 5. The coefficient multiplied by 5 is to give 5% weight in the 100-point system;

[0091] The strategy with the comprehensive score is used as the strategy with the best delivery effect and the strategy is output.

[0092] Strategy execution module, used to execute delivery strategies;

[0093] The strategy optimization module analyzes strategy performance patterns, anomalies, and influencing factors based on playback and user interaction data, calculates the error between actual and predicted delivery effects, and automatically corrects the prediction model and resets the delivery strategy when the error exceeds a threshold.

[0094] In this embodiment, the strategy optimization module calculates the difference between the actual delivery effect and the predicted delivery effect based on the playback and user interaction data. When the difference exceeds the threshold, the prediction model is automatically corrected and the delivery strategy is reset. The processing process is as follows:

[0095] Based on playback data and user interaction behavior, the difference between the actual delivery effect and the predicted delivery effect is calculated, and a threshold is preset. When the difference exceeds the threshold, the key differences between the predicted effect and the actual delivery effect are compared, including the number of clicks, the number of complete plays, and the conversion rate, to locate the key differences.

[0096] The gradient attenuation formula is used to correct the weight of the key difference results. The specific formula is:

[0097] ΔW i =0.02·SHAP i ln(1+|Δ|);

[0098] Where ΔW i is the calculated weight adjustment, SHAP i is the attribution contribution of feature i, ln(1+|Δ|) is the logarithmic decay function;

[0099] Based on the recommendation database content and ad duration, ads that meet the requirements are matched in the ad library, and the ads are combined again to generate multiple delivery strategies. The updated prediction model is used to predict the effect and the strategy with the best delivery effect is selected for output.

[0100] A method for managing a video advertising playback analysis and management platform, wherein the platform adopts any one of the above video advertising playback analysis and management platforms, comprising the following steps:

[0101] S1. Collect advertising material data, playback data, user interaction data, and basic user data;

[0102] S2. Conduct integrated analysis of advertising preferences for users of different regions, ages, and genders;

[0103] S3. Establish a recommendation database based on known data and generate the best delivery strategy based on the recommendation database;

[0104] S4. Execute the delivery strategy and analyze the predicted effect of the delivery strategy and the actual delivery effect. When the difference exceeds the threshold, correct the prediction model and reset the delivery strategy.

[0105] The above embodiments are only used to illustrate the technical method of the present invention and are not intended to limit the present invention. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A video advertising playback analysis and management platform, characterized by: It includes information extraction module, information integration module, push strategy formulation module, strategy execution module and strategy optimization module; The information collection module is used to collect basic data on advertising materials, playback, user interaction and users; The information integration module is used to integrate and analyze the advertising preferences of users in different regions, ages, and genders; The strategy formulation module receives preference-related indicators and user account age data, establishes a recommendation database, matches different advertisements based on the recommendation database and performs creative combinations, generates multiple delivery strategies, establishes a prediction model to estimate delivery effects, and selects the strategy with the best delivery effect; The strategy execution module is used to execute the delivery strategy; The strategy optimization module analyzes strategy performance patterns, anomalies, and influencing factors based on playback and user interaction data, calculates the error between actual and predicted delivery effects, and automatically corrects the prediction model and resets the delivery strategy when the error exceeds a threshold.

2. A video advertising playback analysis and management platform according to claim 1, characterized in that: The information collection module is used to collect advertising materials, playback, user interaction and basic user data. The specific processing steps are as follows: Collect relevant information of various video advertising materials uploaded to the platform, including advertising duration, resolution, format, and theme content; obtain playback data of video ads on various playback channels, including number of plays, playback time, and playback position; record user interaction with video ads, including clicks, pauses, and shares; collect basic user data, including user age, user gender, user region, and user account length.

3. The video advertising playback analysis and management platform according to claim 1, characterized in that: The information integration module is used to integrate and analyze the advertising preferences of users in different regions, ages, and genders. The specific processing steps are as follows: Clean and standardize the data. After processing, encode the categorical data including region, gender, and age and convert them into numerical form. With user accounts as the core associated data, basic user data is associated with ad creative related information, playback data, and user interaction behavior data to build a comprehensive data set; The comprehensive dataset is grouped by user region, age, and gender. For each group, metrics related to ad preferences are calculated. These metrics include the total number of clicks, shares, and plays on ads of different topics by users of different regions, ages, and genders, as well as the average interaction time between users of different groups and ads. Calculate the completed metrics to find out the most popular ad duration, resolution, format, and theme preferences in each group and use this as the analysis result.

4. A video advertising playback analysis and management platform according to claim 1, characterized in that: The strategy formulation module receives preference-related indicators and user account age data, establishes a recommendation database, matches different advertisements based on the recommendation database, and performs creative combinations to generate multiple delivery strategies. It also establishes a prediction model to estimate delivery effects, and selects the strategy with the best delivery effect. The process is as follows: Determine the maturity of the user account. If the user has used the account for less than 7 days and there is insufficient data on the user's preference-related indicators, the account is considered to have low maturity. Otherwise, it is considered to have high maturity. For low-maturity accounts, we match the corresponding groups from the comprehensive data set based on the user's region, age, and gender information, extract the analysis results of the group, and establish a recommendation database, storing the analysis results in the recommendation database; Confirm the total playable duration of the ad and set a ratio to break down the total duration. Based on the recommended database content and the individual broken down durations, match the required ads in the ad library and combine the ads to generate multiple delivery strategies. Establish a prediction model to estimate the effect of the delivery and select the best delivery strategy; For low-maturity accounts, a trigger detection mechanism is set up, including: the user has accumulated 5 valid interactions; the 7th, 15th, and 30th days of user registration; the user's single-day interaction behavior exceeds the historical average by 300%; After the detection is triggered, if the user account maturity is determined to be high, the user preferences are analyzed and a similarity analysis is performed between the user preference data and the data in the recommendation database. Data with a similarity of less than 0.15 is eliminated. Data with a similarity between 0.15 and 0.4 is soft-demoted to a secondary cache pool to retain the right of emergency recall. Data with a similarity greater than 0.4 is retained to reduce the database and further accurately understand user preferences. Finally, based on the content of the recommendation database and the duration of the ads, the ad library is matched with ads that meet the requirements, and the ads are combined to generate multiple delivery strategies. When the user account maturity is first determined to be high, the user's region, age, gender information and personal preference data are directly combined to establish a recommendation database. Finally, based on the content of the recommendation database and the length of the advertisement, the advertisements that meet the requirements are matched in the advertisement library, and the advertisements are combined to generate multiple delivery strategies.

5. A video advertising playback analysis and management platform according to claim 4, characterized in that: The specific process for determining insufficient data for the preference-related indicators is as follows: First, verify the integrity of data types, requiring users to play at least three times and trigger at least one actual interaction, including clicks, pauses, and shares. Second, verify key behavior coverage, confirming that users are exposed to at least two types of advertising themes and their behaviors are distributed on at least three different dates, and that interactions occur in at least two network scenarios, either WiFi or data. Finally, calculate the behavior density threshold using the following formula: Among them, E is the calculated data adequacy index, N is the industry's minimum effective sample size, which is 5, and t is the number of days between the most recent and first behaviors, with an upper limit of 7. When E is less than 0.8, it means that there is insufficient data.

6. A video advertisement playback analysis and management platform according to claim 5, characterized in that: The process of establishing a prediction model to estimate the delivery effect and selecting the best delivery effect strategy is as follows: Evaluate the impact of ad images, compare similarity with historically high-click materials, track topic popularity trends, overlay real-time search indexes, and combine historical click fluctuations of target demographic groups to ultimately calculate the expected click probability. Analyze the rhythm density in the strategy, that is, the number of seconds between key shots, combined with the number of plot turning points, and simultaneously establish a curve of the average attention span of the target group. Match it with the curve of the average attention span of the target group to calculate the probability of the ad being played in its entirety; Verify whether the conversion trigger design aligns with the group's behavioral habits. Specifically, analyze whether the positioning button position matches the group's click hotspot coordinates, and whether the intensity of the time-limited urgency is within the optimal range. Output the conversion rate value. A prediction model is established to perform comprehensive scoring, which is used as the estimated delivery effect. The specific formula is: G=C·40%+V·30%+R·25%+(1-T)·5; Where G is the estimated delivery effect after calculation, C is the expected click probability, V is the probability of complete playback, R is the conversion rate value, and T is the fatigue factor, which ranges from [0 to 1]. A value of 0 indicates that the strategy elements are 100% innovative, and a value of 1 indicates that all elements completely repeat recent content. The strategy with the comprehensive score is used as the strategy with the best delivery effect and the strategy is output.

7. A video advertisement playback analysis and management platform according to claim 6, characterized in that: The strategy optimization module calculates the difference between the actual delivery effect and the predicted delivery effect based on the playback and user interaction data. When the difference exceeds the threshold, the prediction model is automatically corrected and the delivery strategy is reset. The processing process is as follows: Based on playback data and user interaction behavior, the difference between the actual delivery effect and the predicted delivery effect is calculated, and a threshold is preset. When the difference exceeds the threshold, the key differences between the predicted effect and the actual delivery effect are compared, including the number of clicks, the number of complete plays, and the conversion rate, to locate the key differences. The gradient attenuation formula is used to correct the weight of the key difference results. The specific formula is: ΔW i =0.02·SHAP i ·ln(1+∣Δ∣); Where ΔW i is the calculated weight adjustment, SHAP i is the attribution contribution of feature i, ln(1+|Δ|) is the logarithmic decay function; Based on the recommendation database content and ad duration, ads that meet the requirements are matched in the ad library, and the ads are combined again to generate multiple delivery strategies. The updated prediction model is used to predict the effect and the strategy with the best delivery effect is selected for output.

8. A management method for a video advertisement playback analysis management platform, characterized in that: The platform adopts a video advertisement playback analysis and management platform as described in any one of claims 1 to 7, including the following steps: S1. Collect advertising material data, playback data, user interaction data, and basic user data; S2. Conduct integrated analysis of advertising preferences for users of different regions, ages, and genders; S3. Establish a recommendation database based on known data and generate the best delivery strategy based on the recommendation database; S4. Execute the delivery strategy and analyze the predicted effect of the delivery strategy and the actual delivery effect. When the difference exceeds the threshold, correct the prediction model and reset the delivery strategy.

Images