Rearrangement method applied to advertisement recommendation system

Abstract

The invention relates to the technical field of computer science and information engineering, and discloses a rearrangement method applied to an advertisement recommendation system, and the method comprises the steps: building a data collection module, a data preprocessing module, a data feature extraction module, a data algorithm module, a data evaluation module and a data optimization module; advertisement, user and context data collection is carried out in the data collection module, the data preprocessing module cleans and formats the collected data, the data feature extraction module carries out feature extraction on the formatted data, and the data algorithm module calculates the feature extraction data. And the data evaluation module performs system directivity evaluation on a calculation result, directivity evaluation optimization of the data evaluation module is realized through the data optimization module, and the data optimization module performs user interest direction adjustment, total benefit value improvement and comprehensive data favorable comment feedback synchronous adjustment according to the data, so that continuous optimization of the system is ensured.

Description

Technical Field

[0001] The present invention relates to the field of computer science and information engineering technology, and in particular to a rearrangement method applied to an advertisement recommendation system. Background Art

[0002] In traditional advertising recommendation systems, due to the lack of sufficient historical data, it is often difficult for the re-ranking algorithm to accurately assess the interests and preferences of new users or new ads, resulting in unsatisfactory recommendation results. At the same time, the traditional system has limited ability to regulate the flow of different materials during the re-ranking stage, making it difficult to meet specific business needs, increasing the difficulty of flow regulation. How to achieve effective flow regulation while ensuring accurate assessment of user interests and preferences has become a complex problem that needs to be solved urgently. Summary of the invention

[0003] 1. Technical issues to be solved

[0004] In view of the shortcomings of the prior art, the present invention provides a re-arrangement method applied to an advertising recommendation system, which has the advantage of ensuring accurate assessment of user interests and preferences while achieving effective traffic regulation, thereby solving the problem that traditional systems are unable to achieve effective traffic regulation while ensuring accurate assessment of user interests and preferences.

[0005] (II) Technical solution

[0006] To achieve the above object, the present invention provides the following technical solution: a rearrangement method applied to an advertisement recommendation system, comprising the following steps:

[0007] Step 1: Establish a data collection module, a data preprocessing module, a data feature extraction module, a data algorithm module, a data evaluation module and a data optimization module;

[0008] Step 2: Collect advertisement, user and context data in the data collection module;

[0009] Step 3: The data preprocessing module cleans and formats the collected data;

[0010] Step 4: The data feature extraction module extracts features from the formatted data;

[0011] Step 5: The data algorithm module calculates the feature extraction data;

[0012] Step 6: The data evaluation module conducts a system directional evaluation on the calculation results;

[0013] Step 7: Implement directional evaluation optimization of the data evaluation module through the data optimization module.

[0014] Preferably, the data collection module includes a user behavior data unit, an advertising attribute data unit and a context data unit. The user behavior data unit collects user behavior data through click, browse, purchase and search history; the advertising attribute data unit collects advertising attribute data through the category, content, format and target audience information of online advertisements; the context data unit collects context data through time, location, device type and network status. The user behavior data unit, advertising attribute data unit and context data unit organize the internal data of the unit and transmit them to the data preprocessing module through the network.

[0015] Preferably, the data preprocessing module cleans the internal data of the user behavior data unit, the advertising attribute data unit and the context data unit by removing duplicate, erroneous and incomplete data records, and converts the internal data of the unit into a format suitable for analysis through standardization and normalization methods, and then selects feature data useful for model training from the format data to perform feature data extraction, converts the user behavior data, advertising attribute data and context data into user behavior feature data, advertising attribute feature data and context feature data, and numbers them, and transmits the extracted feature data to the data algorithm module through the network for calculation.

[0016] Preferably, the data algorithm module includes an interest evaluation unit, a traffic control unit and a comprehensive data positive feedback synchronization unit. The interest evaluation unit calculates the user interest evaluation index Tx based on the user behavior characteristic data. The traffic control unit calculates the total benefit Qy of advertising traffic regulation based on the user behavior characteristic data and the advertising attribute characteristic data. The comprehensive data positive feedback synchronization unit calculates the comprehensive data positive feedback synchronization score Rv based on the user interest evaluation index Tx and the contextual characteristic data. The interest evaluation unit, the traffic control unit and the comprehensive data positive feedback synchronization unit transmit the calculation results to the data evaluation module through the network.

[0017] Preferably, the user behavior characteristic data includes a set of items with the most clicks by users, a set of items with the most clicks by users and a set of items with the second highest number of clicks by users, and the data are numbered. The set of items with the most clicks by users, the set of items with the most clicks by users and the set of items with the second highest number of clicks by users are numbered S and S, respectively. n 、r i , P m The advertising attribute characteristic data includes the unit benefit value and the recommended advertising platform cost brought by the advertising flow of the item with the most user clicks, and the data is numbered. The unit benefit value and the recommended advertising platform cost brought by the advertising flow are numbered as g i, Q, the context feature data includes statistical advertising traffic regulation factors and weights corresponding to adjusting user interest advertising, and the data are numbered, and the advertising traffic regulation factors and weights corresponding to adjusting user interest advertising are numbered k and α respectively.

[0018] Preferably, the interest evaluation unit calculates the user interest evaluation index Tx according to the user behavior feature data, and the calculation formula is:

[0019]

[0020] In the formula, Tx represents the user interest evaluation index, S n 、r i , P m They represent the set of items with the most clicks by users, the set of items rated by users with the most clicks, and the set of items ranked second in terms of clicks by users, respectively. n ∩P m ) represents the set of items that are similar to the set of items with the most clicks and the set of items ranked second in terms of clicks. n represents the number of items with the most clicks, m represents the number of items ranked second in terms of clicks, and w i Represents the similarity weight between the set of items with the most clicks by users and the set of items with the second highest number of clicks by users.

[0021] Preferably, the traffic control unit calculates the total benefit Qy of advertising traffic regulation based on the user behavior feature data and the advertising attribute feature data, and the calculation formula is:

[0022]

[0023] In the formula, Qy represents the total benefit of advertising traffic regulation, g i represents the unit benefit value brought by the advertising traffic of the item with the most clicks by the i-th user, Q represents the cost of the recommendation advertising platform, n represents the number of items with the most clicks by users, and m represents the number of items ranked second in terms of clicks by users. Indicates the probability of the item ad with the most clicks being clicked.

[0024] Preferably, the comprehensive data positive feedback synchronization unit calculates the comprehensive data positive feedback synchronization score Rv according to the user interest evaluation index Tx and the context feature data, and the calculation formula is:

[0025] Rv=α*Tx+(1-α)*k

[0026] In the formula, Rv represents the synchronous score of comprehensive data positive feedback, Tx represents the user interest evaluation index, k and α represent the weights corresponding to the advertising traffic regulation factor and the adjustment of user interest advertising, respectively.

[0027] Preferably, the data evaluation module evaluates the user's interest direction according to the user interest evaluation index Tx, evaluates the total benefit value of the advertising recommendation content according to the total benefit of advertising traffic regulation Qy, and evaluates whether the system can achieve effective traffic regulation while accurately evaluating user interests and preferences according to the comprehensive data positive feedback synchronization score Rv. The data evaluation module is connected to the data optimization module through a network.

[0028] Preferably, the data optimization module adjusts the data evaluated by the data evaluation module in the direction of user interest, improves the total benefit value, and synchronously adjusts the comprehensive data positive feedback.

[0029] Compared with the prior art, the present invention provides a rearrangement method applied to an advertisement recommendation system, which has the following beneficial effects:

[0030] 1. The present invention calculates the user interest evaluation index Tx through user behavior feature data, establishes an effective feedback mechanism, allows users to provide feedback on recommendation results, and uses the user behavior feature feedback data to further improve the accuracy of system evaluation. Since user interests may change over time, it is necessary to regularly update the user behavior feature data and the user interest evaluation index Tx to determine the user interest evaluation model, so as to ensure the accuracy and timeliness of the system evaluation results.

[0031] 2. The present invention uses a calculation formula for the total benefit Qy of advertising traffic regulation to deeply analyze user behavioral feature data, and uses user clicks, browsing history, and purchase records to understand user interests, thereby performing a detailed analysis of the attribute characteristics of the type of advertisement, target audience, and delivery period to determine the relevance of the advertisement, and then calculates the unit benefit value brought by each advertisement traffic according to the provided total benefit calculation formula for advertising traffic regulation, thereby calculating the total benefit Qy of advertising traffic regulation, and setting an expected benefit value to formulate an advertising traffic regulation strategy for the system. When the data evaluation module evaluates that the total benefit Qy of advertising traffic regulation is lower than the expected benefit value, the data optimization module implements the formulated regulation strategy by adjusting the display frequency of the advertisement, changing the delivery location or time of the advertisement, or conducting targeted promotion for a specific user group, and evaluates the effect through real-time monitoring and data analysis, thereby adjusting the system strategy in a timely manner to optimize traffic distribution and improve advertising benefits. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0033] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0034] See also Figure 1 , a rearrangement method applied to an advertisement recommendation system, comprising the following steps:

[0035] Step 1: Establish a data collection module, a data preprocessing module, a data feature extraction module, a data algorithm module, a data evaluation module and a data optimization module;

[0036] Step 2: Collect advertisement, user and context data in the data collection module;

[0037] Step 3: The data preprocessing module cleans and formats the collected data;

[0038] Step 4: The data feature extraction module extracts features from the formatted data;

[0039] Step 5: The data algorithm module calculates the feature extraction data;

[0040] Step 6: The data evaluation module conducts a system directional evaluation on the calculation results;

[0041] Step 7: Implement directional evaluation optimization of the data evaluation module through the data optimization module.

[0042] The data collection module includes a user behavior data unit, an advertising attribute data unit and a context data unit. The user behavior data unit collects user behavior data through click, browse, purchase and search history; the advertising attribute data unit collects advertising attribute data through the category, content, format and target audience information of online advertisements; the context data unit collects context data through time, location, device type and network status. The user behavior data unit, the advertising attribute data unit and the context data unit organize the internal data of the unit and transmit them to the data preprocessing module through the network.

[0043] The data preprocessing module cleans the internal data of user behavior data units, advertising attribute data units and context data units by removing duplicate, erroneous and incomplete data records, and converts the internal data of the units into a format suitable for analysis through standardization and normalization methods. It then selects feature data useful for model training from the format data to extract feature data, converts user behavior data, advertising attribute data and context data into user behavior feature data, advertising attribute feature data and context feature data, and numbers them, and transmits the extracted feature data to the data algorithm module through the network for calculation.

[0044] The data algorithm module includes an interest evaluation unit, a traffic control unit and a comprehensive data positive feedback synchronization unit. The interest evaluation unit calculates the user interest evaluation index Tx based on the user behavior feature data. The traffic control unit calculates the total benefit Qy of advertising traffic regulation based on the user behavior feature data and the advertising attribute feature data. The comprehensive data positive feedback synchronization unit calculates the comprehensive data positive feedback synchronization score Rv based on the user interest evaluation index Tx and the context feature data. The interest evaluation unit, the traffic control unit and the comprehensive data positive feedback synchronization unit transmit the calculation results to the data evaluation module through the network.

[0045] The user behavior feature data includes the set of items with the most clicks by users, the scores given by users to the items with the most clicks, and the set of items ranked second in terms of clicks by users, and the data are numbered. The set of items with the most clicks by users, the scores given by users to the items with the most clicks, and the set of items ranked second in terms of clicks by users are numbered S respectively. n 、r i , P m (It is necessary to ensure that the user behavior feature data used for calculation is accurate, complete and up-to-date, including the user's click volume, rating and browsing history data). The advertising attribute feature data includes the unit benefit value brought by the advertising traffic of the items with the most user clicks and the recommended advertising platform cost, and the data is numbered. The unit benefit value brought by the advertising traffic and the recommended advertising platform cost are numbered as g i , Q, the context feature data includes statistical advertising traffic regulation factors and weights corresponding to adjusting user interest advertising, and the data are numbered. The advertising traffic regulation factors and weights corresponding to adjusting user interest advertising are numbered k and α respectively.

[0046] The interest evaluation unit calculates the user interest evaluation index Tx according to the user behavior feature data, and the calculation formula is:

[0047]

[0048] In the formula, Tx represents the user interest evaluation index, Sn 、r i , P m They represent the set of items with the most clicks by users, the set of items rated by users with the most clicks, and the set of items ranked second in terms of clicks by users, respectively. n ∩P m ) represents the set of items that are similar to the set of items with the most clicks and the set of items ranked second in terms of clicks. n represents the number of items with the most clicks, m represents the number of items ranked second in terms of clicks, and w i Represents the similarity weight between the set of items with the most clicks by users and the set of items with the second highest number of clicks by users.

[0049] The advantages are: the user interest evaluation index Tx is calculated through user behavior feature data, an effective feedback mechanism is established, allowing users to provide feedback on recommendation results, and the user behavior feature feedback data is used to further improve the accuracy of system evaluation. Since user interests may change over time, it is necessary to regularly update user behavior feature data and user interest evaluation index Tx to determine the user interest evaluation model to ensure the accuracy and timeliness of the system evaluation results.

[0050] The traffic control unit calculates the total benefit Qy of advertising traffic regulation based on user behavior feature data and advertising attribute feature data. The calculation formula is:

[0051]

[0052] In the formula, Qy represents the total benefit of advertising traffic regulation, g i represents the unit benefit value brought by the advertising traffic of the item with the most clicks by the i-th user, Q represents the cost of the recommendation advertising platform, n represents the number of items with the most clicks by users, and m represents the number of items ranked second in terms of clicks by users. Indicates the probability of the item ad with the most clicks being clicked.

[0053] The advantages are: through the calculation formula of the total benefit Qy of advertising traffic regulation, the user's behavioral characteristic data is deeply analyzed, and the user's click volume, browsing history and purchase history are used to understand the user's interests, so as to conduct a detailed analysis of the attribute characteristics of the type of advertisement, target audience, and delivery time period to determine the relevance of the advertisement, and then according to the provided total benefit calculation formula of advertising traffic regulation, the unit benefit value brought by each advertising traffic is calculated, so as to calculate the total benefit Qy of advertising traffic regulation, and set the expected value of benefit, and formulate an advertising traffic regulation strategy for the system. When the data evaluation module evaluates that the total benefit Qy of advertising traffic regulation is lower than the expected value of benefit, the data optimization module implements the formulated regulation strategy by adjusting the display frequency of advertisements, changing the delivery location or time of advertisements, or conducting targeted promotion for specific user groups, and evaluates the effect through real-time monitoring and data analysis, so as to adjust the system strategy in time to optimize traffic distribution and improve advertising benefits.

[0054] The comprehensive data positive feedback synchronization unit calculates the comprehensive data positive feedback synchronization score Rv according to the user interest evaluation index Tx and context feature data, and its calculation formula is:

[0055] Rv=α*Tx+(1-α)*k

[0056] In the formula, Rv represents the synchronous score of comprehensive data positive feedback, Tx represents the user interest evaluation index, k and α represent the weights corresponding to the advertising traffic regulation factor and the adjustment of user interest advertising, respectively.

[0057] The advantages are: by calculating the comprehensive data positive feedback synchronization score Rv, using the given comprehensive data positive feedback synchronization score formula, combining the user interest evaluation index and context feature data, the comprehensive score of each advertisement is calculated. The weight factor in the formula needs to be adjusted according to actual business needs and data analysis results to ensure the accuracy of the score. According to the calculated comprehensive data positive feedback synchronization score Rv, it is helpful to implement real-time regulation of advertising traffic, and will give priority to displaying advertisements with high user interest scores to reduce or not display advertisements with low user interest scores, thereby achieving effective distribution of traffic. Through the comprehensive data positive feedback synchronization score Rv, the advertising recommendation system can achieve effective traffic regulation while accurately evaluating user interests and preferences.

[0058] The data evaluation module evaluates the user's interest direction based on the user interest evaluation index Tx (this index reflects the user's interest in different content, helps the system understand user preferences, and recommends more favorite ads to users). This helps to optimize advertising delivery strategies, increase advertising exposure and click-through rates, and achieve higher profits. The total benefit value of advertising recommendation content is evaluated based on the total benefit of advertising traffic regulation Qy. The system is evaluated based on the synchronous score Rv of comprehensive data positive feedback (this score is used to evaluate whether the system has achieved effective traffic regulation while accurately evaluating user interests and preferences). The data evaluation module is connected to the data optimization module through the network.

[0059] The data optimization module adjusts the user interest direction of the data evaluated by the data evaluation module, improves the total benefit value, and makes synchronous adjustments to the comprehensive data positive feedback (the data optimization module adjusts the user interest direction, improves the total benefit value, and makes synchronous adjustments to the comprehensive data positive feedback based on these data, ensuring continuous optimization of the system).

[0060] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A rearrangement method applied to an advertisement recommendation system, characterized in that: The following steps are involved: Step 1: Establish a data collection module, a data preprocessing module, a data feature extraction module, a data algorithm module, a data evaluation module and a data optimization module; Step 2: Collect advertisement, user and context data in the data collection module; Step 3: The data preprocessing module cleans and formats the collected data; Step 4: The data feature extraction module extracts features from the formatted data; Step 5: The data algorithm module calculates the feature extraction data; Step 6: The data evaluation module conducts a system directional evaluation on the calculation results; Step 7: Implement directional evaluation optimization of the data evaluation module through the data optimization module.

2. The rearrangement method applied to an advertisement recommendation system according to claim 1, characterized in that: The data collection module includes a user behavior data unit, an advertisement attribute data unit and a context data unit, wherein the user behavior data unit collects user behavior data through click, browse, purchase and search history; The advertising attribute data unit collects advertising attribute data through the category, content, format and target audience information of online advertisements; the context data unit collects context data through time, location, device type and network status; the user behavior data unit, advertising attribute data unit and context data unit organize the internal data of the unit and transmit them to the data preprocessing module through the network.

3. The rearrangement method applied to an advertisement recommendation system according to claim 2, characterized in that: The data preprocessing module cleans the internal data of the user behavior data unit, the advertising attribute data unit and the context data unit by removing duplicate, erroneous and incomplete data records, and converts the internal data of the unit into a format suitable for analysis through standardization and normalization methods, and then selects feature data useful for model training from the format data to extract feature data, converts the user behavior data, advertising attribute data and context data into user behavior feature data, advertising attribute feature data and context feature data, and numbers them, and transmits the extracted feature data to the data algorithm module through the network for calculation.

4. The rearrangement method applied to an advertisement recommendation system according to claim 3, characterized in that: The data algorithm module includes an interest evaluation unit, a traffic control unit and a comprehensive data positive feedback synchronization unit. The interest evaluation unit calculates the user interest evaluation index Tx based on the user behavior feature data. The traffic control unit calculates the total benefit Qy of advertising traffic regulation based on the user behavior feature data and the advertising attribute feature data. The comprehensive data positive feedback synchronization unit calculates the comprehensive data positive feedback synchronization score Rv based on the user interest evaluation index Tx and the context feature data. The interest evaluation unit, the traffic control unit and the comprehensive data positive feedback synchronization unit transmit the calculation results to the data evaluation module through the network.

5. The rearrangement method applied to an advertisement recommendation system according to claim 3, characterized in that: The user behavior characteristic data includes a set of items with the most clicks by users, a set of items with the most clicks by users, and a set of items with the second highest number of clicks by users, and the data is numbered. The set of items with the most clicks by users, the set of items with the most clicks by users, and the set of items with the second highest number of clicks by users are numbered S respectively. n 、r i , P m The advertising attribute characteristic data includes statistics on the unit benefit value and the recommended advertising platform cost brought by the advertising traffic of the item with the most customer clicks, and data numbering. The unit benefit value and the recommended advertising platform cost brought by the advertising traffic are numbered as g i , Q, the context feature data includes statistical advertising traffic regulation factors and weights corresponding to adjusting user interest advertising, and the data are numbered, and the advertising traffic regulation factors and weights corresponding to adjusting user interest advertising are numbered k and α respectively.

6. The rearrangement method applied to an advertisement recommendation system according to claim 5, characterized in that: The interest evaluation unit calculates the user interest evaluation index Tx according to the user behavior feature data, and the calculation formula is: In the formula, Tx represents the user interest evaluation index, S n 、r i , P m They represent the set of items with the most clicks by users, the set of items rated by users with the most clicks, and the set of items ranked second in terms of clicks by users, respectively. n ∩P m ) represents the set of items that are similar to the set of items with the most clicks and the set of items ranked second in terms of clicks. n represents the number of items with the most clicks, m represents the number of items ranked second in terms of clicks, and w i Represents the similarity weight between the set of items with the most clicks by users and the set of items with the second highest number of clicks by users.

7. The rearrangement method applied to an advertisement recommendation system according to claim 5, characterized in that: The traffic control unit calculates the total benefit Qy of advertising traffic regulation based on the user behavior feature data and the advertising attribute feature data, and the calculation formula is: In the formula, Qy represents the total benefit of advertising traffic regulation, g i represents the unit benefit value brought by the advertising traffic of the item with the most clicks by the i-th user, Q represents the cost of the recommendation advertising platform, n represents the number of items with the most clicks by users, and m represents the number of items ranked second in terms of clicks by users. Indicates the probability of the item ad with the most clicks being clicked.

8. The rearrangement method applied to an advertisement recommendation system according to claim 5, characterized in that: The comprehensive data positive feedback synchronization unit calculates the comprehensive data positive feedback synchronization score Rv according to the user interest evaluation index Tx and the context feature data, and the calculation formula is: Rv=α*Tx+(1-α)*k In the formula, Rv represents the synchronous score of comprehensive data positive feedback, Tx represents the user interest evaluation index, k and α represent the weights corresponding to the advertising traffic regulation factor and the adjustment of user interest advertising, respectively.

9. The rearrangement method applied to an advertisement recommendation system according to claim 8, characterized in that: The data evaluation module evaluates the user's interest direction based on the user interest evaluation index Tx, evaluates the total benefit value of the advertising recommendation content based on the total benefit of advertising traffic regulation Qy, and evaluates whether the system can achieve effective traffic regulation while accurately evaluating user interests and preferences based on the comprehensive data positive feedback synchronization score Rv. The data evaluation module is connected to the data optimization module through the network.

10. The rearrangement method applied to an advertisement recommendation system according to claim 9, characterized in that: The data optimization module adjusts the data evaluated by the data evaluation module in the direction of user interest, improves the total benefit value, and synchronously adjusts the comprehensive data positive feedback.

Images