Advertisement transaction system and advertisement transaction method

Abstract

The invention relates to the technical field of advertisement analysis, in particular to an advertisement transaction system and an advertisement transaction method.The advertisement transaction system comprises a data collection module, a cheating analysis module and an intelligent contract module, and the data collection module obtains user behavior data from a block chain; the cheating analysis module is used for identifying user cheating behaviors, performing feature construction on user behavior data, and analyzing a feature construction result by using a support vector machine algorithm to determine a user cheating behavior result; the intelligent contract module updates a corresponding node in the block chain according to a user cheating behavior result; and based on the node data in the updated block chain and a feature construction result, calculating a user reputation score for advertisement transaction.

Description

Technical Field

[0001] The present invention relates to the technical field of advertisement analysis, and in particular to an advertisement transaction system and an advertisement transaction method. Background Art

[0002] In traditional advertising trading systems, cheating in advertising transactions seriously affects advertisers' budgets and advertising effects. Common cheating methods include robot clicks, artificial volume increase, malicious scripts, etc. Specifically, traditional advertising trading systems are difficult to deal with complex cheating methods, with low recognition rates and high false positive rates. As time goes by, cheating methods are constantly updated and evolved, and traditional methods are difficult to respond in a timely manner. Traditional recognition methods often only rely on single-dimensional data and cannot fully analyze user behavior.

[0003] In addition, there is a lack of effective trust mechanisms in the advertising trading market, making it difficult for highly reputable advertisers to stand out and for publishers with low reputations to be identified and restricted. When calculating user reputation scores, traditional systems often fail to fully reflect the weight of the user's cheating behavior, thereby reducing the credibility of the user's reputation score and affecting the effectiveness of advertising delivery strategies. Summary of the invention

[0004] The purpose of the present invention is to provide an advertisement trading system and an advertisement trading method to solve the problems raised in the above background technology.

[0005] To achieve the above objectives, one of the objectives of the present invention is to provide an advertising trading system, including a data collection module, a cheating analysis module and a smart contract module, wherein:

[0006] The data collection module obtains user behavior data from the blockchain;

[0007] The cheating analysis module is used to identify user cheating behavior by constructing features of user behavior data and analyzing the results of feature construction using a support vector machine algorithm to determine the results of user cheating behavior;

[0008] The smart contract module updates the corresponding node in the blockchain according to the result of the user's cheating behavior; and calculates the user's reputation score based on the node data in the updated blockchain and the result of the feature construction for advertising transactions.

[0009] As a further improvement of the present technical solution, the data collection module sends a data request to the blockchain to obtain the latest user behavior data, and specifies the query time range when requesting, wherein the query time range is determined based on the data volume of the previous query result, the system load, and the data processing time.

[0010] As a further improvement of the present technical solution, the user behavior data in the data collection module includes the number of clicks, the number of impressions, the geographic location, the type and quantity of devices, the IP address, the type and quantity of browsers and operating systems.

[0011] As a further improvement of the technical solution, the cheating analysis module includes a feature construction unit, and the process of the feature construction unit constructing features of user behavior data specifically includes:

[0012] Construct features for user behavior data, and obtain the click rate of each user by dividing the number of clicks in the user behavior data by the number of impressions; and obtain the impression rate by dividing the number of impressions by the total number of impressions.

[0013] Count the number of clicks and impressions in different time windows in the user behavior data. Divide the number of clicks by the length of the time window to get the click frequency. Do the same with the number of impressions to get the impression frequency.

[0014] Count the number of different geographical locations where ads are clicked and displayed in user behavior data, calculate the number of clicks and displays by each user from different geographical locations, and form a geographical diversity feature;

[0015] Count the types and quantities of different devices used in user behavior data, calculate the number of clicks and displays by each user using different devices, and form device diversity characteristics;

[0016] Count the number of different IP addresses used when clicking and displaying ads in user behavior data, calculate the number of clicks and displays from different IP addresses for each user, and form IP address diversity characteristics;

[0017] Count the types and numbers of different browsers and operating systems used in user behavior data, calculate the number of clicks and displays by each user using different browsers and operating systems, and form user agent diversity characteristics;

[0018] The results of feature construction are click-through rate, display rate, click frequency, display frequency, geographic diversity feature, device diversity feature, IP address diversity feature and user agent diversity feature.

[0019] As a further improvement of the technical solution, the cheating analysis module includes a cheating result determination unit, and the process of the cheating result determination unit determining the result of the user's cheating behavior specifically includes:

[0020] Collect known normal user behavior data and cheating user behavior data; add binary labels to each piece of data to form a data set;

[0021] The support vector machine algorithm model is trained using the data set. The model is to find an optimal hyperplane to separate the data points of normal users and cheating users. The optimal hyperplane is the hyperplane that can maximize the interval between data points of different categories. The optimal hyperplane is found by minimizing the norm of the weight vector and maximizing the interval from the data point to the hyperplane.

[0022] The result of feature construction is input into the trained support vector machine algorithm model, and the label of user behavior, including normal or cheating, is output as the result of user cheating behavior.

[0023] As a further improvement of the technical solution, the process of the smart contract module updating the corresponding node in the blockchain specifically includes:

[0024] If a user is marked as cheating, a cheating mark is added to the corresponding node in the blockchain, and the user's latest cheating behavior results are updated to the corresponding node in the blockchain.

[0025] As a further improvement of the technical solution, the process of calculating the user's reputation score by the smart contract module specifically includes:

[0026] Based on the node data in the updated blockchain and the result of the user feature construction, the user reputation score is calculated. The user reputation score is calculated by assigning positive weights to the result of the user feature construction, assigning negative weights to the cheating marks in the blockchain, and performing weighted summation.

[0027] As a further improvement of the technical solution, the user reputation score is used for advertising transactions, and the corresponding advertising delivery strategy in the public database is selected according to the size of the user reputation score.

[0028] A second object of the present invention is to provide an advertising transaction method, comprising the following steps:

[0029] S1. Obtain user behavior data from the blockchain;

[0030] S2. Determine the result of the user's cheating behavior by constructing features of the user's behavior data and analyzing the results of the feature construction using a support vector machine algorithm;

[0031] S3. Update the corresponding node in the blockchain according to the result of the user's cheating behavior; and calculate the user's reputation score based on the node data in the updated blockchain and the result of feature construction for advertising transactions.

[0032] Compared with the prior art, the present invention has the following beneficial effects:

[0033] The advertising transaction system and advertising transaction method obtain the latest user behavior data from the blockchain, conduct a comprehensive analysis of user behavior through feature construction, and use a support vector machine algorithm to accurately identify cheating behavior, effectively solving the cheating identification problem in the traditional system; according to the identification result, cheating is marked in the blockchain, and the user reputation score is calculated, normal behavior features are given positive weights and cheating marks are given negative weights to ensure the accuracy and transparency of the score; this design not only improves the real-time and accuracy of cheating identification, but also establishes an effective trust mechanism, so that high-reputation advertisers stand out and low-reputation publishers are effectively restricted, thereby improving the effectiveness of advertising delivery strategies. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a schematic diagram of the overall method flow of the present invention;

[0035] Figure 2 It is a schematic diagram of the overall module of the present invention.

[0036] In the figure: 100, data collection module; 200, cheating analysis module; 201, feature construction unit; 202, cheating result determination unit; 300, smart contract module. DETAILED DESCRIPTION

[0037] The following will be combined with the accompanying drawings in 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.

[0038] Next, see Figure 1-Figure 2 One of the purposes of this embodiment is to provide an advertising transaction system, including a data collection module 100, a cheating analysis module 200 and a smart contract module 300.

[0039] The data collection module 100 obtains user behavior data from the blockchain, specifically including:

[0040] The advertising platform generates a piece of user behavior data each time a user clicks or displays an ad, and each piece of user behavior data contains multiple features;

[0041] The advertising platform uses the private key to digitally sign the generated records. The signing process ensures the integrity and non-tamperability of the data.

[0042] The signed record is uploaded to the blockchain, and each node in the blockchain verifies the signature and timestamp of the record to ensure the authenticity and source of the data;

[0043] The data collection module 100 periodically sends data requests to the blockchain to obtain the latest user behavior data. When requesting, the query time range, user ID, advertisement ID and other conditions are specified, for example: query all ad click and display records in the past hour; query all ad click records of a specific user ID; query all display records of a specific advertisement ID; the query time range is determined according to the data volume of the last query result, system load and data processing time, for example:

[0044] If the amount of data in the last query exceeds the maximum amount of data that the system can handle, shorten the query time range by 10-20 minutes;

[0045] If the amount of data in the last query is within the maximum amount of data that the system can handle, keep the current query time range unchanged;

[0046] If the amount of data in the last query is far less than the maximum amount of data that the system can handle, extend the query time range by 10-20 minutes;

[0047] If the system load is high, reduce the query frequency appropriately, for example, query once every 2 hours;

[0048] If the system load is low, increase the query frequency appropriately, for example, to query once every half an hour;

[0049] If the data processing time is long, shorten the query time range by 10-20 minutes;

[0050] If the data processing time is short, extend the query time range by 10-20 minutes.

[0051] The data collection module 100 extracts all user behavior data related to ad clicks and displays from the blockchain, and the extracted user behavior data includes click counts, display counts, geographic locations, device types and quantities, IP addresses, browser and operating system types and quantities. For example, the data collection module 100 extracts records of all ad clicks and displays in the past hour and stores them in a local database or cache.

[0052] The feature construction unit 201 in the cheating analysis module 200 performs feature construction on the user behavior data, specifically including:

[0053] Construct features for user behavior data, and obtain the click rate of each user by dividing the number of clicks in the user behavior data by the number of impressions; and obtain the impression rate by dividing the number of impressions by the total number of impressions.

[0054] Count the number of clicks and impressions in different time windows (such as every hour, every day) in the user behavior data. Divide the number of clicks by the length of the time window (for example, the number of clicks per hour divided by 60 minutes) to get the click frequency. Do the same with the number of impressions to get the impression frequency.

[0055] Count the number of different geographical locations where ads are clicked and displayed in user behavior data, calculate the number of clicks and displays by each user from different geographical locations, and form a geographical diversity feature;

[0056] Count the types and quantities of different devices used in user behavior data, calculate the number of clicks and displays by each user using different devices, and form device diversity characteristics;

[0057] Count the number of different IP addresses used when clicking and displaying ads in user behavior data, calculate the number of clicks and displays from different IP addresses for each user, and form IP address diversity characteristics;

[0058] Count the types and numbers of different browsers and operating systems used in user behavior data, calculate the number of clicks and displays by each user using different browsers and operating systems, and form user agent diversity characteristics;

[0059] The results of feature construction are click-through rate, display rate, click frequency, display frequency, geographic diversity feature, device diversity feature, IP address diversity feature and user agent diversity feature.

[0060] The cheating result determination unit 202 in the cheating analysis module 200 uses a support vector machine algorithm to analyze the result of the feature construction and determine the result of the user's cheating behavior, specifically including:

[0061] Collect known normal user behavior data and cheating user behavior data; add a binary label (normal or cheating) to each piece of data to form a data set;

[0062] The support vector machine algorithm model is trained using the data set to find an optimal hyperplane that separates the data points of normal users and cheating users. The optimal hyperplane is the hyperplane that can maximize the interval between data points of different categories. The optimal hyperplane is found by solving an optimization problem. Specifically, it minimizes the norm of the weight vector while maximizing the interval from the data point to the hyperplane.

[0063] The result of feature construction is input into the trained support vector machine algorithm model, and the label of user behavior, including normal or cheating, is output as the result of user cheating behavior.

[0064] The smart contract module 300 updates the corresponding node in the blockchain according to the result of the user's cheating behavior; and calculates the user's reputation score based on the node data in the updated blockchain for advertising transactions, specifically including:

[0065] Receive the user cheating behavior result transmitted by the cheating analysis module 200, the result is a list including the user ID and the cheating label;

[0066] Each node in the blockchain stores user behavior data, including the number of clicks, impressions, times marked as cheating, geographic location diversity, device diversity, IP address diversity, and user agent diversity.

[0067] If a user is marked as cheating, a cheating mark is added to the corresponding node in the blockchain, and the user's latest cheating behavior results are updated to the corresponding node in the blockchain. For example, if user A is marked as cheating, the behavior data of node A is updated, and the number of times it is marked as cheating is increased;

[0068] Based on the node data in the updated blockchain and the result of the user feature construction, the user reputation score is calculated by assigning positive weights to the result of the user feature construction, assigning negative weights to the cheating marks in the blockchain, and performing weighted summation to calculate the user reputation score;

[0069] The calculated user reputation score is used for advertising transactions, and the corresponding advertising delivery strategy is selected through the public database. For example, for a high reputation score (for example, above 70 points): priority is given to advertising delivery to increase the chances of ad display and clicks; for a medium reputation score (for example, 50 to 70 points): appropriate reduction in advertising delivery for further observation and analysis; for a low reputation score (for example, below 50 points): restriction of ad display to avoid losses caused by cheating.

[0070] The second purpose of this embodiment is to provide an advertisement transaction method, including the following method steps:

[0071] S1. Obtain user behavior data from the blockchain;

[0072] S2. Determine the result of the user's cheating behavior by constructing features of the user's behavior data and analyzing the results of the feature construction using a support vector machine algorithm;

[0073] S3. Update the corresponding node in the blockchain according to the result of the user's cheating behavior; and calculate the user's reputation score based on the node data in the updated blockchain and the result of feature construction for advertising transactions.

[0074] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and descriptions are only preferred examples of the present invention and are not intended to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention may have various changes and improvements, which fall within the scope of the present invention. The scope of protection of the present invention is defined by the attached claims and their equivalents.

Claims

1. An advertising transaction system, characterized in that: The system comprises a data collection module (100), a cheating analysis module (200) and a smart contract module (300), wherein: The data collection module (100) obtains user behavior data from the blockchain; The cheating analysis module (200) is used to identify cheating behavior of users, by constructing features of user behavior data and analyzing the results of the feature construction using a support vector machine algorithm to determine the results of the user's cheating behavior; The smart contract module (300) updates the corresponding node in the blockchain according to the result of the user's cheating behavior; and calculates the user's reputation score based on the node data in the updated blockchain and the result of the feature construction for use in advertising transactions.

2. The advertising transaction system according to claim 1, characterized in that: The data collection module (100) sends a data request to the blockchain to obtain the latest user behavior data, and specifies the query time range when requesting, wherein the query time range is determined based on the data volume of the previous query result, the system load and the data processing time.

3. The advertising transaction system according to claim 2, characterized in that: The user behavior data in the data collection module (100) includes the number of clicks, the number of impressions, the geographic location, the type and quantity of devices, the IP address, the type and quantity of browsers and operating systems.

4. The advertising transaction system according to claim 3, characterized in that: The cheating analysis module (200) comprises a feature construction unit (201), and the process of the feature construction unit (201) constructing features for user behavior data specifically comprises: Construct features for user behavior data, and obtain the click rate of each user by dividing the number of clicks in the user behavior data by the number of impressions; and obtain the impression rate by dividing the number of impressions by the total number of impressions. Count the number of clicks and impressions in different time windows in the user behavior data. Divide the number of clicks by the length of the time window to get the click frequency. Do the same with the number of impressions to get the impression frequency. Count the number of different geographical locations where ads are clicked and displayed in user behavior data, calculate the number of clicks and displays by each user from different geographical locations, and form a geographical diversity feature; Count the types and quantities of different devices used in user behavior data, calculate the number of clicks and displays by each user using different devices, and form device diversity characteristics; Count the number of different IP addresses used when clicking and displaying ads in user behavior data, calculate the number of clicks and displays from different IP addresses for each user, and form IP address diversity characteristics; Count the types and numbers of different browsers and operating systems used in user behavior data, calculate the number of clicks and displays by each user using different browsers and operating systems, and form user agent diversity characteristics; The results of feature construction are click-through rate, display rate, click frequency, display frequency, geographic diversity feature, device diversity feature, IP address diversity feature and user agent diversity feature.

5. The advertising transaction system according to claim 4, characterized in that: The cheating analysis module (200) comprises a cheating result determination unit (202), and the process of the cheating result determination unit (202) determining the result of the user's cheating behavior specifically comprises: Collect known normal user behavior data and cheating user behavior data; add binary labels to each piece of data to form a data set; The support vector machine algorithm model is trained using the data set. The model is to find an optimal hyperplane to separate the data points of normal users and cheating users. The optimal hyperplane is the hyperplane that can maximize the interval between data points of different categories. The optimal hyperplane is found by minimizing the norm of the weight vector and maximizing the interval from the data point to the hyperplane. The result of feature construction is input into the trained support vector machine algorithm model, and the label of user behavior, including normal or cheating, is output as the result of user cheating behavior.

6. The advertising transaction system according to claim 5, characterized in that: The process of the smart contract module (300) updating the corresponding node in the blockchain specifically includes: If a user is marked as cheating, a cheating mark is added to the corresponding node in the blockchain, and the user's latest cheating behavior results are updated to the corresponding node in the blockchain.

7. The advertising transaction system according to claim 6, characterized in that: The process of calculating the user reputation score by the smart contract module (300) specifically includes: Based on the node data in the updated blockchain and the result of the user feature construction, the user reputation score is calculated. The user reputation score is calculated by assigning positive weights to the result of the user feature construction, assigning negative weights to the cheating marks in the blockchain, and performing weighted summation.

8. The advertisement trading system according to claim 7, characterized in that: The user reputation score is used for advertising transactions, and the corresponding advertising delivery strategy in the public database is selected according to the size of the user reputation score.

9. A method for using an advertisement transaction as claimed in any one of claims 1 to 8, characterized in that: The method comprises the following steps: S1. Obtain user behavior data from the blockchain; S2. Determine the result of the user's cheating behavior by constructing features of the user's behavior data and analyzing the results of the feature construction using a support vector machine algorithm; S3. Update the corresponding nodes in the blockchain according to the results of the user's cheating behavior; Based on the node data in the updated blockchain and the results of feature construction, the user reputation score is calculated for advertising transactions.

Images