Artificial intelligence-based advertisement delivery method and device, terminal equipment and medium

Abstract

The application relates to an advertisement putting method and device based on artificial intelligence, terminal equipment and medium. The method uses a first division method to determine a maximum sample quantity and a first optimal feature for data of N sample users, divides the N sample users into two sets by using the optimal feature when the maximum sample quantity is large, takes a set meeting a first preset condition as a first type leaf node, uses a second division method to determine a second optimal feature when the maximum sample quantity is small, divides the N sample users into two sets by using the optimal feature, takes a set meeting a second preset condition as a second type leaf node, returns to the initial division to obtain a decision tree when the condition is not met, inputs real-time label data of a user to be put into the decision tree, and determines the user falling into the first type leaf node, so that target users can be predicted and screened, and the target users are put into advertisements, the screening result is relatively accurate, and the precision of the advertisement putting is improved.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based advertising delivery method, apparatus, terminal equipment, and medium. Background Technology

[0002] Currently, with the rise of mobile internet, online advertising has become the mainstream method of advertising. As clients demand increasingly higher precision in their ad targeting, traditional indiscriminate advertising can no longer meet their needs. Therefore, to improve client satisfaction, it's essential to screen potential users. However, due to the complexity of current user data, manual screening is insufficient to effectively identify suitable target users. Thus, how to accurately and effectively screen target users to improve the precision of ad targeting has become a pressing issue. Summary of the Invention

[0003] In view of this, embodiments of this application provide an artificial intelligence-based advertising delivery method, apparatus, terminal device, and medium to address the problem of how to accurately and effectively screen target users in order to improve the accuracy of advertising delivery.

[0004] In a first aspect, embodiments of this application provide an artificial intelligence-based advertising delivery method, the advertising delivery method comprising:

[0005] Using the first partitioning method, the historical label feature values ​​of the corresponding label features in N sample users are divided into different sets based on the target feature value of each label feature. Combining the historical behavior feature values ​​of the corresponding sample users, the set with the largest positive sample accuracy and the largest sample size is determined as the maximum sample size. The label feature of this set when it is partitioned is the first optimal feature.

[0006] When the maximum sample size is greater than the sample size threshold, the N sample users are divided into two sets based on the first optimal feature. The set that meets the first preset condition is taken as the first type of leaf node. Otherwise, the set is returned and divided using the first partitioning method.

[0007] When the maximum sample size is not greater than the sample size threshold, the second partitioning method is used to divide the N sample users into different sets with the target feature value of each label feature. Combining the historical behavior feature value of the corresponding sample users, the label feature of the set with the largest accurate sample coefficient among all sets is determined as the second optimal feature.

[0008] The N sample users are divided into two sets using the second optimal feature. The set that satisfies the second preset condition is taken as the second type of leaf node. Otherwise, the set is returned to be divided using the first partitioning method. The above steps are repeated until all N sample users fall into the corresponding leaf node. All leaf nodes are connected to form a decision tree.

[0009] The real-time tag data of each tag among the users to be targeted is input into the decision tree to determine the users among the users to be targeted who fall into the first type of leaf node, and the advertisement is delivered to the users who fall into the first type of leaf node.

[0010] Secondly, embodiments of this application provide an artificial intelligence-based advertising delivery device, the advertising delivery device comprising:

[0011] The first partitioning module is used to divide the historical label feature values ​​of the corresponding label features in N sample users into different sets using the first partitioning method with the target feature value of each label feature. Combining the historical behavior feature values ​​of the corresponding sample users, the module determines the set with the largest positive sample accuracy and the largest sample size among all sets as the maximum sample size. The label feature of this set when it is partitioned is the first optimal feature.

[0012] The second partitioning module is used to partition the N sample users into two sets based on the first optimal feature when the maximum sample size is greater than the sample size threshold, and to take the set that meets the first preset condition as the first type of leaf node; otherwise, the set is returned to be partitioned using the first partitioning method.

[0013] The third partitioning module is used to divide the N sample users into different sets using the second partitioning method when the maximum sample size is not greater than the sample size threshold. Combined with the historical behavior feature values ​​of the corresponding sample users, the label feature of the set with the largest accurate sample coefficient among all sets is determined as the second optimal feature.

[0014] The fourth partitioning module is used to divide the N sample users into two sets using the second optimal feature. The set that satisfies the second preset condition is taken as the second type of leaf node. Otherwise, the set is returned to be partitioned using the first partitioning method. The above steps are repeated until all N sample users fall into the corresponding leaf nodes. All leaf nodes are connected to form a decision tree.

[0015] The advertising delivery module is used to input the real-time tag data of each tag among the users to be advertised into the decision tree, determine the users among the users to be advertised who fall into the first type of leaf node, and deliver the advertisement to the users who fall into the first type of leaf node.

[0016] Thirdly, embodiments of this application provide a terminal device, the terminal device including a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the advertising delivery method as described in the first aspect.

[0017] Fourthly, embodiments of this application provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the advertising delivery method as described in the first aspect.

[0018] The beneficial effects of this application embodiment compared with the prior art are as follows: This application uses a first partitioning method to divide the historical label feature values ​​of the corresponding label features among N sample users into different sets based on the target feature value of each label feature. Combined with the historical behavior feature values ​​of the corresponding sample users, the set with the largest positive sample accuracy (greater than the accuracy threshold) and the largest sample size is determined as the maximum sample size. The label feature of this set at the time of partitioning is the first optimal feature. When the maximum sample size is greater than the sample size threshold, the N sample users are divided into two sets using the first optimal feature. The set that satisfies the first preset condition is designated as the first type of leaf node; otherwise, the set is returned to be partitioned using the first partitioning method. When the maximum sample size is not greater than the sample size threshold, a second partitioning method is used to divide the N sample users into different sets based on the target feature value of each label feature. Combined with the historical behavior feature values ​​of the corresponding sample users, the set is determined as the first type of leaf node. The label feature of the set with the largest accurate sample coefficient in the set is the second optimal feature. Using the second optimal feature, N sample users are divided into two sets. The set that meets the second preset condition is taken as the second type of leaf node. Otherwise, the set is returned to the first partitioning method for partitioning. The above steps are repeated until all N sample users fall into the corresponding leaf nodes. All leaf nodes are connected to form a decision tree. The real-time label data of each label of the users to be targeted is input into the decision tree to determine the users who fall into the first type of leaf node. Ads are placed to the users who fall into the first type of leaf node. The target type of users are determined by analyzing historical data. The real-time data of the users to be targeted is then analyzed to partition the users to be targeted. When the users to be targeted are the target type of users, ads are placed on them. This can intelligently filter target users and place ads. The filtering results are more accurate, thereby improving the accuracy of ad placement. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0020] Figure 1 This is a schematic diagram of an application environment for an artificial intelligence-based advertising delivery method provided in Embodiment 1 of this application;

[0021] Figure 2 This is a flowchart illustrating an artificial intelligence-based advertising delivery method provided in Embodiment 2 of this application;

[0022] Figure 3 This is a flowchart illustrating an artificial intelligence-based advertising delivery method provided in Embodiment 3 of this application;

[0023] Figure 4 This is a schematic diagram of the structure of an artificial intelligence-based advertising delivery device provided in Embodiment 4 of this application;

[0024] Figure 5 This is a schematic diagram of the structure of a terminal device provided in Embodiment 5 of this application. Detailed Implementation

[0025] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0026] It should be understood that, when used in this application specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or a collection thereof.

[0027] It should also be understood that the term “and / or” as used in this application specification and the appended claims means any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0028] As used in this application specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if detected [the described condition or event]" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once detected [the described condition or event]," or "in response to detection [the described condition or event]."

[0029] Furthermore, in the description of this application and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0030] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of this application include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0031] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0032] Foundational technologies for artificial intelligence generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies mainly encompass computer vision, robotics, biometrics, speech processing, natural language processing, and machine learning / deep learning.

[0033] It should be understood that the sequence number of each step in the following embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0034] To illustrate the technical solution of this application, specific embodiments are described below.

[0035] This application provides an AI-based advertising delivery method, applicable to the environment shown in Figure 1, where the client and server communicate. The client includes, but is not limited to, handheld computers, desktop computers, laptops, ultra-mobile personal computers (UMPCs), netbooks, cloud terminal devices, and personal digital assistants (PDAs). The server can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0036] See Figure 2 This is a flowchart illustrating an artificial intelligence-based advertising delivery method provided in Embodiment 2 of this application. The aforementioned advertising delivery method can be applied to... Figure 1 The server-side component connects to corresponding databases, platforms, clients, and other devices capable of providing user tag data to obtain user data. For example... Figure 2 As shown, this advertising delivery method may include the following steps:

[0037] Step S201: Using the first partitioning method, the historical label feature values ​​of the corresponding label features in the N sample users are divided into different sets based on the target feature value of each label feature. Combining the historical behavior feature values ​​of the corresponding sample users, the set with the largest positive sample accuracy and the largest sample size is determined as the maximum sample size. The label feature of this set when it is partitioned is the first optimal feature.

[0038] In this application, the sample users are those with historical tag feature values ​​and historical behavior feature values. Historical tag feature values ​​represent the user's profile characteristics, while historical behavior feature values ​​represent the user's actual behavior. For example, historical tag feature values ​​can refer to data corresponding to tag features such as the user's age, gender, and preferences; historical behavior feature values ​​can refer to data corresponding to behaviors such as the user's consumption and TV on / off times. The target feature value corresponding to a tag feature can refer to the ideal result corresponding to that tag feature. For example, for the age tag feature, the target feature value is 18 years old. For any tag feature, its historical tag feature value is compared with the corresponding target feature value to determine whether the historical tag feature value matches the target feature. A match means the historical tag feature value meets the target feature value; a mismatch means the historical tag feature value does not meet the target feature value. For example, if the historical tag feature value of the sample user's age tag is 18 years old, and the target feature value is also 18 years old, then the historical tag feature value matches the target feature value.

[0039] For any given label feature, matching sample users are grouped into one set, and non-matching sample users are grouped into another set. Thus, the first partitioning method is used to partition the N sample users, resulting in 2N sets, where N is a positive integer.

[0040] For 2N sets, by combining the historical behavioral feature values ​​of sample users, the positive sample accuracy of each set can be obtained. The set corresponding to the simultaneous satisfaction of positive sample accuracy and sample size is the target set. The sample size of the target set is the maximum sample size. The label feature used when the target set is divided is the first optimal feature.

[0041] Optionally, the set with the largest positive sample accuracy (greater than the accuracy threshold) and the largest sample size among all sets is identified as the maximum sample size. The label features of this set when it is partitioned are the first optimal features, including:

[0042] Calculate the sample size of each set and the number of positive samples in the corresponding set whose historical behavioral feature values ​​are the target values;

[0043] The ratio of the positive sample size to the total sample size is used as the positive sample accuracy of the corresponding set.

[0044] Filter the set of positive samples whose accuracy is greater than the accuracy threshold, and determine the maximum sample size in the filtered set.

[0045] The label feature of the set with the largest sample size is determined as the first optimal feature.

[0046] The target value can represent the historical behavioral characteristics that meet the target. For example, the target value is to turn on the TV at least 3 times during the prime viewing period (18:00-24:00) within 3 days. Historical behavioral characteristics that meet this target value are considered positive samples.

[0047] The ratio of the number of positive samples to the total number of samples in a set is the positive sample accuracy of that set. Similarly, the ratio of the number of negative samples to the total number of samples is the negative sample accuracy of that set.

[0048] The accuracy threshold mentioned above can be 90% to ensure a high accuracy rate during segmentation, while also meeting a certain recall rate.

[0049] Alternatively, advertising methods may also include:

[0050] Obtain the historical label data and the historical behavior data of the corresponding sample users for each label feature of N sample users;

[0051] Based on the label conditions and behavior conditions corresponding to each label feature, the historical label data and historical behavior data are binarized to convert the historical label data into historical label feature values ​​and the historical behavior data into historical behavior feature values.

[0052] Historical label feature values ​​are determined to be 1 for historical label data that meet the label conditions, and 0 otherwise. Historical behavioral feature values ​​are determined to be 1 for historical behavioral data that meet the behavioral conditions, and 0 otherwise.

[0053] After acquiring the tag data and behavioral data, this data is first binarized for subsequent logical operations, converting it into 0 or 1 (i.e., feature values) based on the tag and behavioral conditions. For example, if the behavioral condition is to turn on the TV at least 3 times during prime viewing hours (18:00-24:00) within 3 days, then if a sample user's historical behavioral data meets this condition, the sample user's historical behavioral feature value is determined to be 1. Furthermore, after binarization, all feature values ​​are represented using 0 or 1. A match is achieved when the target feature value is the same as the historical tag feature value; a mismatch is achieved when the target feature value is different from the corresponding binary data of the historical tag feature value.

[0054] Step S202: When the maximum sample size is greater than the sample size threshold, the N sample users are divided into two sets using the first optimal feature. The set that satisfies the first preset condition is taken as the first type of leaf node. Otherwise, the set is returned to be divided using the first partitioning method.

[0055] In this application, the sample size threshold can be 500. When the maximum sample size exceeds this threshold, the historical label feature values ​​of the corresponding label features of the N sample users are compared again using the first optimal feature obtained above, thereby dividing the N sample users into two sets, such as a first set of users and a second set of users. Sample users whose historical label feature values ​​match the target feature value (i.e., satisfy the target feature value) are classified into the first set of users, and sample users whose historical label feature values ​​do not match the target feature value (i.e., do not satisfy the target feature value) are classified into the second set of users.

[0056] The first preset condition can be set according to actual needs. The purpose of the first preset condition is to filter the two sets mentioned above, find the set that meets the condition, and classify the samples in the set into the first type of leaf node. Sets that do not meet the condition are returned to the above step S201 for further processing until all samples are classified into the corresponding leaf node.

[0057] The first preset condition could be that the sample size in the set is greater than the sample size threshold, or that the accuracy of the positive samples in the set is greater than the accuracy threshold, etc.

[0058] Optionally, the set of nodes that satisfy the first preset condition is included as the first type of leaf nodes:

[0059] Calculate the sample size for each set and the positive sample accuracy for the corresponding set;

[0060] Select a set of positive samples whose accuracy is greater than the accuracy threshold and whose sample size is greater than the sample size threshold, and determine that the sample users corresponding to this set fall into the first type of leaf node.

[0061] The accuracy of positive samples in the set is still calculated based on historical behavioral feature values. For example, the accuracy of positive samples in the first type of user set and the second type of user set is calculated based on historical behavioral features. If the accuracy of positive samples in the first type of user set is greater than the accuracy threshold and the sample size of the first type of user set is greater than the sample size threshold, then the first type of user set is determined to meet the first preset condition, that is, the sample users in the first type of user set are assigned to the first type of leaf node. If the accuracy of positive samples in the second type of user set is not greater than the accuracy threshold or the sample size of the second type of user set is not greater than the sample size threshold, then the second type of user set is determined not to meet the first preset condition, that is, the sample users in the second type of user set are returned to step S201 for processing again.

[0062] Step S203: When the maximum sample size is not greater than the sample size threshold, the second partitioning method is used to divide the N sample users into different sets based on the target feature value of each label feature. Combined with the historical behavior feature values ​​of the corresponding sample users, the label feature of the set with the largest accurate sample coefficient among all sets is determined as the second optimal feature.

[0063] In this application, when the maximum sample size is not greater than the sample size threshold, it can be seen that the partitioning result of the first partitioning method cannot meet the requirements. Therefore, another partitioning method (i.e., the second partitioning method) is used to partition the N sample users. Similarly, the target feature value of each label feature is matched with the historical label feature value.

[0064] For any given label feature, matching sample users are grouped into one set, and non-matching sample users are grouped into another set. Thus, the first partitioning method is used to partition N sample users, resulting in 2N sets.

[0065] For 2N sets, the accurate sample coefficient of each set can be calculated by combining the historical behavior feature values ​​of the sample users. The label feature used when the set with the largest accurate sample coefficient is divided is taken as the second optimal feature. The accurate sample coefficient is used to reflect that the set has a high positive sample accuracy and a high sample size, that is, a high division accuracy. Specifically, it can be calculated based on the positive sample accuracy, sample size and total sample size.

[0066] The set corresponding to the largest accurate sample coefficient is taken as the target set, and the label feature used when dividing the target set is the second optimal feature.

[0067] Optionally, the label feature used to determine the set with the largest accurate sample coefficients among all sets as the second optimal feature includes:

[0068] Calculate the sample size of each set and the number of positive samples in the corresponding set whose historical behavioral feature values ​​are the target values;

[0069] The ratio of the positive sample size to the total sample size is used as the positive sample accuracy of the corresponding set.

[0070] Based on the total sample size of N sample users, the sample size of each set, and the positive sample accuracy, determine the accurate sample coefficient of the corresponding set;

[0071] The label feature of the set with the largest accurate sample coefficient among all sets is determined as the second optimal feature.

[0072] The target value can represent the historical behavioral characteristics that meet the target. For example, the target value is to turn on the TV at least 3 times during the prime viewing period (18:00-24:00) within 3 days. Historical behavioral characteristics that meet this target value are considered positive samples.

[0073] The ratio of the number of positive samples to the total number of samples in a set is the positive sample accuracy of that set. Similarly, the ratio of the number of negative samples to the total number of samples is the negative sample accuracy of that set.

[0074] Optionally, based on the total sample size of the N sample users, the sample size corresponding to each set, and the positive sample accuracy, the accurate sample coefficients for the corresponding set are determined, including:

[0075] Calculate the ratio of the sample size of each set to the total sample size to obtain the ratio result for the corresponding set;

[0076] Multiply the positive sample accuracy of each set by the preset weight to obtain the multiplication result for the corresponding set;

[0077] Add the ratio results of each set to the product results, and determine the sum as the accurate sample coefficient of the corresponding set.

[0078] The accurate sample coefficients can be calculated based on the above process, but this application is not limited to using the above calculation process, as long as it can characterize the partitioning accuracy of different sets.

[0079] Step S204: Use the second optimal feature to divide the N sample users into two sets. The set that satisfies the second preset condition is taken as the second type of leaf node. Otherwise, the set is returned to be divided using the first partitioning method. Repeat the above steps until all N sample users fall into the corresponding leaf nodes. All leaf nodes are connected to form a decision tree.

[0080] In this application, the historical label feature values ​​of the corresponding label features of N sample users are compared again using the second optimal feature, thereby dividing the N sample users into two sets, such as a third set of users and a fourth set of users. Sample users whose historical label feature values ​​match the target feature value (i.e., satisfy the target feature value) are classified into the third set of users, and sample users whose historical label feature values ​​do not match the target feature value (i.e., do not satisfy the target feature value) are classified into the fourth set of users.

[0081] The second preset condition can be set according to actual needs. The purpose of this second preset condition is to filter the two sets mentioned above, find the set that meets the condition, and assign the samples in the set to the second type of leaf node. Sets that do not meet the condition are returned to step S201 for further processing until all samples are assigned to the corresponding leaf node. For example, the second preset condition can refer to a situation where the number of samples in the set is not greater than a sample size threshold.

[0082] In this application, N sample users are respectively placed into corresponding leaf nodes. The above partitioning process connects the leaf nodes to form a decision tree. The root node of the decision tree is the dataset. Each leaf node in the decision tree corresponds to its own decision condition. Using the decision condition, it can be determined whether the dataset of the root node will fall into the corresponding leaf node. The decision condition can refer to the first optimal feature or the second optimal feature mentioned above.

[0083] Optionally, the set of nodes satisfying the second preset condition can be designated as the second type of leaf nodes, including:

[0084] Calculate the sample size of each set, filter out sets with a sample size less than the sample size threshold, and determine the sample users corresponding to the set to fall into the second type of leaf node.

[0085] In this process, the sample size of the set is directly calculated. For example, if the sample size of the third type of user set is not greater than the sample size threshold, then the third type of user set is determined to meet the second preset condition, that is, the sample users in the third type of user set are assigned to the second type of leaf node. If the sample size of the fourth type of user set is greater than the sample size threshold, then the fourth type of user set is determined not to meet the second preset condition, that is, the sample users in the fourth type of user set are returned to step S201 for further processing.

[0086] Step S205: Input the real-time tag data of each tag among the users to be targeted into the decision tree, determine the users who fall into the first type of leaf node among the users to be targeted, and target the advertisement to the users who fall into the first type of leaf node.

[0087] In this application, the decision condition for the first type of leaf node is to obtain the target user. Therefore, the real-time tag data of the user to be targeted is input into the decision tree. If the user eventually falls into the first type of leaf node, it means that the user is the target user. If the user eventually falls into other leaf nodes, it means that the user is not the target user.

[0088] In this application, an advertisement can be placed on a target user. In this application, a target user can refer to a smart device, such as a smart TV or a mobile phone.

[0089] In a decision tree, there can be multiple first-class leaf nodes, each corresponding to a different decision condition. Furthermore, the user's data is judged based on different decision conditions, thereby determining the number of first-class leaf nodes that the user's data will eventually fall into. If the number is greater than the threshold, the user is determined to be the target user.

[0090] In this embodiment, a first partitioning method is used to divide the historical label feature values ​​of corresponding label features among N sample users into different sets based on the target feature value of each label feature. Combined with the historical behavioral feature values ​​of the corresponding sample users, the set with the largest positive sample accuracy (greater than the accuracy threshold) and the largest sample size is determined as the maximum sample size. The label feature of this set at the time of partitioning is the first optimal feature. When the maximum sample size is greater than the sample size threshold, the N sample users are divided into two sets using the first optimal feature. The set that meets the first preset condition is designated as the first type of leaf node; otherwise, the set is returned to be partitioned using the first partitioning method. When the maximum sample size is not greater than the sample size threshold, a second partitioning method is used to divide the N sample users into different sets based on the target feature value of each label feature. Combined with the historical behavioral feature values ​​of the corresponding sample users, the set with the largest positive sample accuracy (greater than the accuracy threshold) is determined as the maximum sample size. The label feature of the set with the largest number of users when it is partitioned is the second optimal feature. Using the second optimal feature, N sample users are divided into two sets. The set that meets the second preset condition is taken as the second type of leaf node. Otherwise, the set is returned to be partitioned using the first partitioning method. The above steps are repeated until all N sample users fall into the corresponding leaf nodes. All leaf nodes are connected to form a decision tree. The real-time label data of each label of the users to be targeted is input into the decision tree to determine the users who fall into the first type of leaf node. Ads are placed on the users who fall into the first type of leaf node. The target type of users are determined by analyzing historical data. The real-time data of the users to be targeted is then analyzed to partition the users to be targeted. When the users to be targeted are the target type of users, ads are placed on them. This method can intelligently filter out target users and place ads. The filtering results are more accurate, thereby improving the accuracy of ad placement.

[0091] See Figure 3 This is a flowchart illustrating an artificial intelligence-based advertising delivery method provided in Embodiment 3 of this application. Figure 3 As shown, this advertising delivery method may include the following steps:

[0092] Step S301: Using the first partitioning method, the historical label feature values ​​of the corresponding label features in the N sample users are divided into different sets based on the target feature value of each label feature. Combining the historical behavior feature values ​​of the corresponding sample users, the set with the largest positive sample accuracy and the largest sample size is determined as the maximum sample size. The label feature of this set when it is partitioned is the first optimal feature.

[0093] Step S302: When the maximum sample size is greater than the sample size threshold, the N sample users are divided into two sets using the first optimal feature. The set that satisfies the first preset condition is taken as the first type of leaf node. Otherwise, the set is returned and divided using the first partitioning method.

[0094] Step S303: When the maximum sample size is not greater than the sample size threshold, the second partitioning method is used to divide the N sample users into different sets based on the target feature value of each label feature. Combined with the historical behavior feature values ​​of the corresponding sample users, the label feature of the set with the largest accurate sample coefficient among all sets is determined as the second optimal feature.

[0095] Step S304: Use the second optimal feature to divide the N sample users into two sets. The set that satisfies the second preset condition is taken as the second type of leaf node. Otherwise, the set is returned to be divided using the first partitioning method. Repeat the above steps until all N sample users fall into the corresponding leaf nodes. All leaf nodes are connected to form a decision tree.

[0096] Steps S301 to S304 are the same as steps S201 to S204 above, and can be referred to the description of steps S201 to S204, which will not be repeated here.

[0097] Step S305: Input the real-time tag data of each tag among the users to be targeted into the decision tree, determine the users who fall into the second type of leaf node, and do not target the advertisement to the users who fall into the second type of leaf node.

[0098] Unlike the method for determining target users in Embodiment 2 above, in this application, if a user's data eventually falls into a second type of leaf node, it can be directly determined that the user is not a target user, that is, no advertisement will be delivered to the user.

[0099] In one implementation, the user's data is divided using a decision tree. If the user's data falls into both the first and second type of leaf nodes, the number of times the user's data falls into the first type of leaf node is compared with the number of times it falls into the second type of leaf node. The leaf node corresponding to the maximum value is taken as the leaf node where the user finally falls. For example, if the user's data falls into 3 first type leaf nodes and 2 second type leaf nodes, then the user belongs to the first type of leaf node.

[0100] This application embodiment can directly determine that the user to be advertised is not the target user, thereby determining that there is no need to advertise to them, thus improving processing efficiency.

[0101] Corresponding to the advertising delivery method in the above embodiments, Figure 4 This paper shows a structural block diagram of an AI-based advertising delivery device according to Embodiment 4 of this application. The above-described advertising delivery method can be applied to... Figure 1 The server in the diagram connects to the database to obtain user data. For ease of explanation, only the parts relevant to the embodiments of this application are shown.

[0102] See Figure 4 The advertising delivery device includes:

[0103] The first partitioning module 41 is used to use the first partitioning method to divide the historical label feature values ​​of the corresponding label features in N sample users into different sets based on the target feature value of each label feature. Combining the historical behavior feature values ​​of the corresponding sample users, the module determines the set with the largest positive sample accuracy and the largest sample size among all sets as the maximum sample size. The label feature of this set when it is partitioned is the first optimal feature.

[0104] The second partitioning module 42 is used to partition N sample users into two sets using the first optimal feature when the maximum sample size is greater than the sample size threshold, and to take the set that meets the first preset condition as the first type of leaf node; otherwise, the set is returned to be partitioned using the first partitioning method.

[0105] The third partitioning module 43 is used to divide N sample users into different sets using the second partitioning method when the maximum sample size is not greater than the sample size threshold. Combined with the historical behavior feature values ​​of the corresponding sample users, the label feature of the set with the largest accurate sample coefficient among all sets is determined as the second optimal feature.

[0106] The fourth partitioning module 44 is used to divide N sample users into two sets using the second optimal feature. The set that meets the second preset condition is taken as the second type of leaf node. Otherwise, the set is returned to be partitioned using the first partitioning method. The above steps are repeated until all N sample users fall into the corresponding leaf node. All leaf nodes are connected to form a decision tree.

[0107] The ad delivery module 45 is used to input the real-time tag data of each tag among the users to be advertised into the decision tree, determine the users who fall into the first type of leaf node among the users to be advertised, and deliver the ad to the users who fall into the first type of leaf node.

[0108] Optionally, the first partitioning module 41 mentioned above includes:

[0109] The first calculation unit is used to calculate the sample size of each set and the number of positive samples in the corresponding set whose historical behavioral feature values ​​are the target values;

[0110] The first ratio unit is used to take the ratio of the positive sample size to the sample size as the positive sample accuracy of the corresponding set.

[0111] The first screening unit is used to screen the set where the accuracy of positive samples is greater than the accuracy threshold, and to determine the maximum sample size in the screened set.

[0112] The first feature determination unit is used to determine the label feature of the set with the largest sample size as the first optimal feature when it is partitioned. Optionally, the third partitioning module 43 mentioned above includes:

[0113] The second calculation unit is used to calculate the sample size of each set and the number of positive samples in the corresponding set whose historical behavioral feature values ​​are the target values;

[0114] The second ratio unit is used to take the ratio of the positive sample size to the sample size as the positive sample accuracy of the corresponding set.

[0115] The coefficient determination unit is used to determine the accurate sample coefficients of the corresponding set based on the total sample size of N sample users, the sample size corresponding to each set, and the positive sample accuracy.

[0116] The second feature determination unit is used to determine the label feature of the set with the largest accurate sample coefficients among all sets as the second optimal feature. Optionally, the above coefficient determination unit includes:

[0117] The ratio calculation subunit is used to calculate the ratio of the sample size of each set to the total sample size, and obtain the ratio result of the corresponding set;

[0118] The multiplication calculation subunit is used to multiply the accuracy of the positive samples corresponding to each set by a preset weight to obtain the multiplication result of the corresponding set;

[0119] The coefficient determination subunit is used to add the ratio result of each set to the multiplication result, and determine the sum as the accurate sample coefficient of the corresponding set.

[0120] Optionally, the second partitioning module 42 mentioned above includes:

[0121] The third calculation unit is used to calculate the sample size of each set and the positive sample accuracy of the corresponding set;

[0122] The second filtering unit is used to filter out the set of positive samples whose accuracy is greater than the accuracy threshold and whose sample size is greater than the sample size threshold, and to determine the sample users corresponding to the set that fall into the first type of leaf node.

[0123] Optionally, the fourth partitioning module 44 mentioned above includes:

[0124] The third filtering unit is used to calculate the sample size of each set, filter out sets with a sample size less than the sample size threshold, and determine that the sample users corresponding to the set fall into the second type of leaf node.

[0125] Optionally, the aforementioned advertising delivery device also includes:

[0126] The data acquisition module is used to acquire historical label data and historical behavior data of N sample users for each label feature;

[0127] The binary processing module is used to perform binarization processing on historical label data and historical behavior data according to the label conditions and behavior conditions corresponding to each label feature, converting historical label data into historical label feature values ​​and historical behavior data into historical behavior feature values.

[0128] The feature value determination module is used to determine that the historical label feature value corresponding to the historical label data that meets the label conditions is 1, otherwise it is 0, and to determine that the historical behavior feature value corresponding to the historical behavior data that meets the behavior conditions is 1, otherwise it is 0.

[0129] It should be noted that the information interaction and execution process between the above modules are based on the same concept as the method embodiments of this application. For details on their specific functions and technical effects, please refer to the method embodiments section, which will not be repeated here.

[0130] Figure 5 This is a schematic diagram of the structure of a terminal device provided in Embodiment 5 of this application. Figure 5 As shown, the terminal device of this embodiment includes: at least one processor ( Figure 5 Only one is shown in the diagram), a memory, and a computer program stored in the memory and executable on at least one processor, which, when executed by the processor, implements the steps in any of the above-described advertising delivery method embodiments.

[0131] The terminal device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that... Figure 5 This is merely an example of a terminal device and does not constitute a limitation on the terminal device. A terminal device may include more or fewer components than shown in the figure, or a combination of certain components, or different components, such as network interfaces, displays, and input devices.

[0132] The processor referred to can be a CPU, but it can also be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor.

[0133] The memory includes readable storage media, internal memory, etc., wherein the internal memory can be the main memory of the terminal device, and the internal memory provides an environment for the operation of the operating system and computer-readable instructions stored in the readable storage media. The readable storage media can be the hard drive of the terminal device, or in some embodiments, it can be an external storage device of the terminal device, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital Card (SD), or a Flash Card. Furthermore, the memory can include both internal storage units and external storage devices of the terminal device. The memory is used to store the operating system, applications, bootloader, data, and other programs, such as program code for computer programs. The memory can also be used to temporarily store data that has been output or will be output.

[0134] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above device can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here. If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above method embodiments. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. A computer-readable medium can include at least: any entity or device capable of carrying computer program code, a recording medium, a computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0135] The implementation of all or part of the processes in the methods of the above embodiments can also be accomplished by a computer program product. When the computer program product is run on a terminal device, the terminal device executes the steps in the above method embodiments.

[0136] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0137] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0138] In the embodiments provided in this application, it should be understood that the disclosed apparatus / terminal devices and methods can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0139] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0140] The above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be included within the protection scope of this application.

Claims

1. An artificial intelligence-based advertising delivery method, characterized in that, The advertising delivery method includes: Using the first partitioning method, the historical label feature values ​​of the corresponding label features in N sample users are divided into different sets based on the target feature value of each label feature. Combining the historical behavior feature values ​​of the corresponding sample users, the set with the largest positive sample accuracy and the largest sample size is determined as the maximum sample size. The label feature of this set when it is partitioned is the first optimal feature. When the maximum sample size is greater than the sample size threshold, the N sample users are divided into two sets based on the first optimal feature. The set that meets the first preset condition is taken as the first type of leaf node. Otherwise, the set is returned and divided using the first partitioning method. When the maximum sample size is not greater than the sample size threshold, the second partitioning method is used to divide the N sample users into different sets with the target feature value of each label feature. Combining the historical behavior feature value of the corresponding sample users, the label feature of the set with the largest accurate sample coefficient among all sets is determined as the second optimal feature. The N sample users are divided into two sets using the second optimal feature. The set that satisfies the second preset condition is taken as the second type of leaf node. Otherwise, the set is returned to be divided using the first partitioning method. The above steps are repeated until all N sample users fall into the corresponding leaf node. All leaf nodes are connected to form a decision tree. Input the real-time tag data of each tag among the users to be targeted into the decision tree, determine the users among the users to be targeted who fall into the first type of leaf node, and target the advertisement to the users who fall into the first type of leaf node; The set with the largest positive sample accuracy (greater than the accuracy threshold) and the largest sample size among all sets is identified as the maximum sample size. The label features of this set when it is partitioned are the first optimal features, including: Calculate the sample size of each set and the number of positive samples in the corresponding set whose historical behavioral feature values ​​are the target values; The ratio of the positive sample size to the total sample size is used as the positive sample accuracy of the corresponding set. Filter the set of positive samples whose accuracy is greater than the accuracy threshold, and determine the maximum sample size in the filtered set. The label feature of the set with the largest sample size when it is divided is determined as the first optimal feature.

2. The advertising placement method according to claim 1, characterized in that, The label features used to determine the set with the largest accurate sample coefficients among all sets as the second optimal features include: Calculate the sample size of each set and the number of positive samples in the corresponding set whose historical behavioral feature values ​​are the target values; The ratio of the positive sample size to the total sample size is used as the positive sample accuracy of the corresponding set. Based on the total sample size of the N sample users, the sample size corresponding to each set, and the positive sample accuracy, determine the accurate sample coefficient of the corresponding set; The label feature of the set with the largest accurate sample coefficient among all sets is determined as the second optimal feature.

3. The advertising placement method according to claim 2, characterized in that, Based on the total sample size of the N sample users, the sample size corresponding to each set, and the positive sample accuracy, the accurate sample coefficients for the corresponding sets are determined as follows: Calculate the ratio of the sample size of each set to the total sample size to obtain the ratio result for the corresponding set; Multiply the positive sample accuracy of each set by the preset weight to obtain the multiplication result for the corresponding set; Add the ratio results of each set to the product results, and determine the sum as the accurate sample coefficient of the corresponding set.

4. The advertising placement method according to claim 1, characterized in that, The set that satisfies the first preset condition is designated as the first type of leaf node, including: Calculate the sample size for each set and the positive sample accuracy for the corresponding set; Select a set of positive samples whose accuracy is greater than the accuracy threshold and whose sample size is greater than the sample size threshold, and determine that the sample users corresponding to this set fall into the first type of leaf node.

5. The advertising placement method according to claim 1, characterized in that, The set that satisfies the second preset condition is designated as the second type of leaf node, including: Calculate the sample size of each set, filter out sets with a sample size less than the specified sample size threshold, and determine that the sample users corresponding to the set fall into the second type of leaf node.

6. The advertising placement method according to any one of claims 1 to 5, characterized in that, The advertising delivery method also includes: Obtain the historical label data and the historical behavior data of the corresponding sample users for each label feature of N sample users; Based on the label conditions and behavior conditions corresponding to each label feature, the historical label data and historical behavior data are binarized to convert the historical label data into historical label feature values ​​and the historical behavior data into historical behavior feature values. Historical label feature values ​​are determined to be 1 for historical label data that meet the label conditions, and 0 otherwise. Historical behavioral feature values ​​are determined to be 1 for historical behavioral data that meet the behavioral conditions, and 0 otherwise.

7. An artificial intelligence-based advertising delivery device, characterized in that, The advertising delivery device includes: The first partitioning module is used to divide the historical label feature values ​​of the corresponding label features in N sample users into different sets using the first partitioning method with the target feature value of each label feature. Combining the historical behavior feature values ​​of the corresponding sample users, the module determines the set with the largest positive sample accuracy and the largest sample size among all sets as the maximum sample size. The label feature of this set when it is partitioned is the first optimal feature. The second partitioning module is used to partition the N sample users into two sets based on the first optimal feature when the maximum sample size is greater than the sample size threshold, and to take the set that meets the first preset condition as the first type of leaf node; otherwise, the set is returned to be partitioned using the first partitioning method. The third partitioning module is used to divide the N sample users into different sets using the second partitioning method when the maximum sample size is not greater than the sample size threshold. Combined with the historical behavior feature values ​​of the corresponding sample users, the label feature of the set with the largest accurate sample coefficient among all sets is determined as the second optimal feature. The fourth partitioning module is used to divide the N sample users into two sets using the second optimal feature. The set that satisfies the second preset condition is taken as the second type of leaf node. Otherwise, the set is returned to be partitioned using the first partitioning method. The above steps are repeated until all N sample users fall into the corresponding leaf nodes. All leaf nodes are connected to form a decision tree. The advertising delivery module is used to input the real-time tag data of each tag among the users to be advertised into the decision tree, determine the users among the users to be advertised who fall into the first type of leaf node, and deliver the advertisement to the users who fall into the first type of leaf node. The first partitioning module includes: The first calculation unit is used to calculate the sample size of each set and the number of positive samples in the corresponding set whose historical behavioral feature values ​​are the target values; The first ratio unit is used to take the ratio of the positive sample size to the sample size as the positive sample accuracy of the corresponding set. The first screening unit is used to screen the set where the accuracy of positive samples is greater than the accuracy threshold, and to determine the maximum sample size in the screened set. The first feature determination unit is used to determine the label feature of the set with the largest sample size when it is divided as the first optimal feature.

8. A terminal device, characterized in that, The terminal device includes a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the advertising delivery method as described in any one of claims 1 to 6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the advertising delivery method as described in any one of claims 1 to 6.

Images