Display information pushing method and device, electronic equipment and storage medium
By using feature fusion and value prediction models of scene and non-scene data in the display information delivery system, the problems of different scenarios and generalization capabilities of different models are solved, and more accurate and efficient information push is achieved.
Patent Information
- Application Number
- CN202510433702.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-08
- Publication Date
- 2025-08-08
AI Technical Summary
In the existing technology, in the display information delivery system, it is difficult to take into account the differences between different display scenarios and the generalization ability of models, resulting in insufficient accuracy and efficiency of information push.
By obtaining the scene data and non-scene data of the target user, using the pre-trained value prediction model for feature fusion and value prediction, obtain the displayed value score, and push information to the user when the conditions are met. The model includes feature fusion networks and value prediction networks, which can be comprehensively analyzed under the collaboration of multitasking sharing and multiple value prediction subnets.
It improves the accuracy of information push and the generalization ability of the model, and can predict user value more accurately and push information in different display scenarios, reducing the need for separate modeling.
Smart Images
Figure CN120448625A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to a method, device, electronic device, and storage medium for display information push. Background Art
[0002] In display information delivery systems, providers typically distribute content across multiple channels. However, these channels vary significantly, and user behavior, such as clicks, pull-ups, and playbacks, also varies significantly across different scenarios. Before placing content, providers assess user value based on historical pull-ups, playbacks, and revenue. Traditional approaches model each display scenario independently, or create a shared model across multiple scenarios. Independent modeling ignores commonalities across scenarios, resulting in a large number of models and high maintenance costs. Hybrid modeling, on the other hand, ignores differences between scenarios, resulting in poor model generalization. Therefore, how to fully account for user differences across scenarios while enhancing model generalization is a pressing issue. Summary of the Invention
[0003] In view of this, in order to solve some or all of the above technical problems, the embodiments of the present application provide a method, device, electronic device and computer-readable storage medium for pushing display information.
[0004] In a first aspect, an embodiment of the present application provides a method for pushing display information, the method comprising: obtaining scene data and non-scene data corresponding to a target user, wherein the scene data represents data recorded for operations performed by the target user on display information of a preset type in a target display scene, and the non-scene data represents data recorded for the target user's behavior outside the target display scene; using a pre-trained value prediction model, performing display value prediction on the scene data and the non-scene data to obtain a display value score for the target user, wherein the display value score represents the probability of the target user performing a preset operation after viewing the target display information; if the display value score meets the preset display conditions, sending the target display information to the terminal used by the target user.
[0005] In one possible implementation, the value prediction model includes a feature fusion network and a value prediction network; using a pre-trained value prediction model, display value prediction is performed on scene data and non-scene data to obtain a display value score for the target user, including: using the feature fusion network to perform feature fusion on scene data and non-scene data to obtain fused features; using the value prediction network to perform display value prediction on the fused features to obtain a display value score for the target user.
[0006] In one possible implementation, the feature fusion network includes a scene feature extraction subnetwork and a non-scene feature extraction subnetwork; the feature fusion network is used to perform feature fusion on scene data and non-scene data to obtain fused features, including: using the scene feature extraction subnetwork to extract scene features from scene data to obtain scene features; using the non-scene feature extraction subnetwork to extract scene features from non-scene data to obtain non-scene features; and fusing the scene features with the non-scene features to obtain fused features.
[0007] In one possible implementation, a scene feature extraction subnetwork is used to extract scene features from scene data to obtain scene features, including: merging non-scene data and scene data to obtain merged data; and inputting the merged data into the scene feature extraction subnetwork to obtain scene features.
[0008] In one possible implementation, the value prediction network includes a multi-task shared subnetwork and at least two value prediction subnetworks; using the value prediction network, performing display value prediction on the fused features to obtain a display value score for the target user, including: using the multi-task shared subnetwork to perform feature extraction on the fused features to obtain multi-task basic features; using each value prediction subnetwork in the at least two value prediction subnetworks to perform value prediction on the multi-task basic features to obtain at least two display value prediction values; and fusing the at least two display value prediction values to obtain a display value score.
[0009] In one possible implementation, each value prediction subnetwork in at least two value prediction subnetworks is used to perform value prediction on the multi-task basic features to obtain at least two display value prediction values, including: obtaining scene features corresponding to the scene data; inputting the multi-task basic features and the scene features into each value prediction subnetwork in at least two value prediction subnetworks to perform value prediction on the multi-task basic features to obtain at least two display value prediction values.
[0010] In one possible embodiment, at least two value prediction subnetworks include a usage time prediction subnetwork and at least one revenue prediction subnetwork; using each of the at least two value prediction subnetworks, value prediction is performed on multi-task basic features to obtain at least two display value prediction values, including: using each revenue prediction subnetwork in at least one revenue prediction subnetwork to perform revenue value prediction on multi-task basic features respectively, to obtain at least one revenue prediction value as the display value prediction value corresponding to at least one revenue prediction subnetwork, wherein the revenue prediction value represents the size of the revenue obtained from the target user by using the target display information within a first preset time period in the future; using the usage time prediction network, usage time prediction is performed on the multi-task basic features to obtain the predicted usage time of the target user, wherein the predicted usage time represents the time the target user uses the object indicated by the target display information within a second preset time period in the future.
[0011] In a second aspect, an embodiment of the present application provides a display information push device, which includes: an acquisition module for acquiring scene data and non-scene data corresponding to a target user, wherein the scene data represents data recorded for the operations performed by the target user on a preset type of display information in a target display scene, and the non-scene data represents data recorded for the target user's behavior outside the target display scene; a prediction module for using a pre-trained value prediction model to perform display value prediction on the scene data and non-scene data to obtain a display value score of the target user, wherein the display value score represents the probability of the target user performing a preset operation after viewing the target display information; a sending module for sending the target display information to the terminal used by the target user if the display value score meets the preset display conditions.
[0012] In a third aspect, an embodiment of the present application provides an electronic device, comprising: a memory for storing a computer program; a processor for executing the computer program stored in the memory, and when the computer program is executed, it implements the method of any embodiment of the display information push method of the first aspect of the present application.
[0013] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the method of any embodiment of the display information push method of the first aspect described above is implemented.
[0014] In a fifth aspect, an embodiment of the present application provides a computer program, which includes a computer-readable code. When the computer-readable code runs on a device, the processor in the device implements a method as described in any embodiment of the display information push method of the first aspect above.
[0015] The display information push method, device, electronic device and computer-readable storage medium provided in the embodiments of the present application obtain scene data and non-scene data corresponding to the target user, wherein the scene data represents the data recorded for the target user's operations on the preset type of display information in the target display scene, and the non-scene data represents the data recorded for the target user's behavior outside the target display scene; using a pre-trained value prediction model, the scene data and non-scene data are predicted for display value, and the display value score of the target user is obtained. When the display value score meets the preset display conditions, the target display information is pushed to the terminal used by the target user. The embodiments of the present application realize the joint analysis of scene data and non-scene data using the same model, fully extracting the relevant features of the target user in the target display scene from the scene data, and the model extracts richer features from data with more dimensions, thereby improving the accuracy of predicting the user's display value and pushing display information to the target user more accurately. In addition, the value prediction model provided in the embodiments of the present application can also analyze scene data in different display scenes, without the need to model a specific display scene separately, thereby improving the generalization ability of the model while taking into account the differences between different display scenes. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0017] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.
[0018] One or more embodiments are exemplarily illustrated by pictures in the corresponding drawings. These exemplifications do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings are represented as similar elements. Unless otherwise stated, the figures in the drawings do not constitute proportional limitations.
[0019] Figure 1 A flowchart of a method for pushing display information provided in an embodiment of the present application;
[0020] Figure 2 A schematic diagram of the structure of the value prediction model provided in the embodiment of the present application;
[0021] Figure 3 A flowchart of another method for pushing display information provided in an embodiment of the present application;
[0022] Figure 4 A schematic diagram of the structure of another value prediction model provided in an embodiment of the present application;
[0023] Figure 5 A flowchart of another method for pushing display information provided in an embodiment of the present application;
[0024] Figure 6 A schematic diagram of the structure of another value prediction model provided in an embodiment of the present application;
[0025] Figure 7 A schematic diagram of the structure of another value prediction model provided in an embodiment of the present application;
[0026] Figure 8 A flowchart of another method for pushing display information provided in an embodiment of the present application;
[0027] Figure 9 A schematic diagram of the structure of another value prediction model provided in an embodiment of the present application;
[0028] Figure 10 A schematic diagram of the structure of another value prediction model provided in an embodiment of the present application;
[0029] Figure 11 A schematic diagram of the structure of a display information push device provided in an embodiment of the present application;
[0030] Figure 12 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0031] Various exemplary embodiments of the present application will now be described in detail with reference to the accompanying drawings. It is apparent that the described embodiments are only a portion of the embodiments of the present application, rather than all of the embodiments. It should be noted that, unless otherwise specifically stated, the relative arrangement of components and steps, numerical expressions, and numerical values described in these embodiments do not limit the scope of the present application.
[0032] Those skilled in the art will understand that the terms "first" and "second" in the embodiments of the present application are only used to distinguish between different steps, devices, modules and other objects, and neither represent any specific technical meaning nor indicate the logical order between them.
[0033] It should also be understood that in this embodiment, “a plurality of” may refer to two or more than two, and “at least one” may refer to one, two or more than two.
[0034] It should also be understood that any component, data or structure mentioned in the embodiments of the present application can generally be understood as one or more, unless explicitly limited or otherwise indicated in the context.
[0035] In addition, the term "and / or" in this application is simply a description of the association relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent three situations: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this application generally indicates that the related objects are in an "or" relationship.
[0036] It should also be understood that the description of each embodiment in this application focuses on the differences between the embodiments, and the same or similar aspects can be referenced with each other. For the sake of brevity, they will not be described one by one.
[0037] The following description of at least one exemplary embodiment is merely illustrative in nature and is in no way intended to limit the present disclosure, its application, or uses.
[0038] Technologies, circuits, and devices known to those of ordinary skill in the relevant art may not be discussed in detail, but where appropriate, the above-mentioned technologies, circuits, and devices should be considered part of the specification.
[0039] It should be noted that like reference numerals and letters refer to like items in the following figures, and therefore, once an item is defined in one figure, it need not be further discussed in subsequent figures.
[0040] It should be noted that, unless there is a conflict, the embodiments and features in the embodiments of this application can be combined with each other. To facilitate understanding of the embodiments of this application, the application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments. Obviously, the embodiments described are part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.
[0041] In order to solve the technical problem that the existing technology has poor targeting of information pushed to users, the present application provides a method for displaying information push, which can improve the generalization ability of the model while taking into account the differences in different display scenarios and improving the accuracy of information push.
[0042] Figure 1A flow chart of a method for pushing display information provided in an embodiment of the present application. This method can generally be applied to a server, by setting a value prediction model on the server, analyzing the user's scene data and non-scene data, and then pushing display information to the user terminal. In addition, the execution subject of this method can be hardware or software. When the above-mentioned execution subject is hardware, the execution subject can be one or more of the above-mentioned electronic devices. For example, a single electronic device can execute this method, or a plurality of electronic devices can cooperate with each other to execute this method. When the above-mentioned execution subject is software, this method can be implemented as multiple software or software modules, or as a single software or software module. No specific limitation is made here.
[0043] like Figure 1 As shown, the method specifically includes:
[0044] Step 101: Obtain scene data and non-scene data corresponding to the target user.
[0045] In some embodiments, the scenario data represents data recorded regarding the target user's actions on a preset type of display information within a target display scenario. The preset type of display information can include various types of information, such as advertising information, recommendation information, and item description information. The target display scenario is a specific information service platform, such as an application or website that provides audio and video playback services, or an application or website that provides document, image, or shopping services.
[0046] In a targeted display scenario, the server can push display information to the target user, who can then view the display information and, after viewing the display information, perform actions such as clicks, play, purchase, and registration. The server can record these actions. For example, the recorded scenario data may include the number of impressions, clicks, plays, and playback duration of the display information.
[0047] Non-scenario data refers to data recorded on the target user's behavior outside the target display scenario. Optionally, non-scenario data may include statistical data on the target user's operations on preset types of display information in other display scenarios. For example, data such as the total number of exposures, total number of clicks, and total number of plays of display information in other scenarios. In addition, non-scenario data may also include personal portrait data of the target user (such as the user's age, gender, preferences, etc.), as well as data recorded on the target user's behavior on the platform indicated by the display information (such as the video playback platform to which the display information is linked) (such as playback history data, collection data, etc.).
[0048] Step 102: Use the pre-trained value prediction model to predict the display value of the scene data and the non-scene data to obtain the display value score of the target user.
[0049] In some embodiments, the display value score output by the value prediction model represents the probability that a target user will perform a preset action after viewing the target display information. Specifically, the higher the display value score, the greater the probability that the user will perform the preset action, and the higher the potential revenue for the provider of the target display information.
[0050] The target display information may include various types of information, such as advertising information, recommendation information, item description information, etc. The preset operations include various types of pre-set operations, such as click operations, purchase operations after clicks, registration operations, etc.
[0051] The value prediction model is pre-trained using a machine learning algorithm. Specifically, a large number of training samples are collected in advance. Each training sample includes sample scenario data and sample non-scenario data obtained for the sample display information, as well as annotated revenue information and usage duration information from the sample users. During model training, the sample scenario data and sample non-scenario data serve as the model input, and the annotated revenue information and usage duration information serve as the model's expected output. By adjusting the model parameters, the error between the model's actual output and the expected output is minimized, resulting in a trained value prediction model.
[0052] When performing value prediction, the trained model can integrate scenario data and non-scenario data into the model. For example, the value prediction model can include a scenario feature extraction network that extracts scenario features from the scenario data. Using the scenario features as weights, the model multiplies the non-scenario data to obtain fused features, which are then used to predict the display value. The value prediction model can output predicted revenue information and predicted user usage time corresponding to the annotated information. Based on a preset value score calculation strategy, the model calculates the predicted revenue information and predicted user usage time information to obtain a display value score.
[0053] Step 103: If the display value score meets the preset display conditions, target display information is sent to the terminal used by the target user.
[0054] In some embodiments, the aforementioned display conditions are pre-set, representing criteria for determining whether a target user can potentially generate revenue for the provider of the target display information. For example, a display value score threshold may be set; if the display value score is greater than or equal to the threshold, the target display information may be pushed to the target user. For another example, the display value scores of multiple users may be tallied and ranked in descending order of display value scores. If the target user is in the top X users, the target display information may be pushed to the target user.
[0055] The method for pushing display information provided by the embodiment of the present application obtains the scene data and non-scene data corresponding to the target user, wherein the scene data represents the data recorded for the target user's operations on the preset type of display information in the target display scene, and the non-scene data represents the data recorded for the target user's behavior outside the target display scene; using a pre-trained value prediction model, the scene data and the non-scene data are predicted for display value, and the display value score of the target user is obtained. When the display value score meets the preset display conditions, the target display information is pushed to the terminal used by the target user. The embodiment of the present application realizes the joint analysis of scene data and non-scene data using the same model, fully extracts the relevant features of the target user in the target display scene from the scene data, and the model extracts richer features from data of more dimensions, thereby improving the accuracy of predicting the user's display value and pushing display information to the target user more accurately. In addition, the value prediction model provided by the embodiment of the present application can also analyze scene data in different display scenes, without the need to model a specific display scene separately, thereby improving the generalization ability of the model while taking into account the differences between different display scenes.
[0056] In some optional implementations, such as Figure 2 As shown, the value prediction model includes a feature fusion network 201 and a value prediction network 202.
[0057] like Figure 3 As shown, step 102 includes:
[0058] Step 1021: Use a feature fusion network to fuse the scene data and the non-scene data to obtain fused features.
[0059] Among them, the feature fusion network can receive input scene data and non-scene data, extract features from the scene data and non-scene data, and fuse the extracted features (for example, by merging features or performing multiplication operations on features) to obtain fused features.
[0060] Step 1022: Use the value prediction network to predict the display value of the fused features to obtain the display value score of the target user.
[0061] The value prediction network can include a multi-layer neural network, which can be implemented based on architectures such as MMoE (Multi-gate Mixture-of-Experts) and PLE (Progressive Layered Extraction). The value prediction network extracts and calculates the input fusion features layer by layer, ultimately obtaining a predicted display value score through regression calculation.
[0062] This embodiment sets a fusion network and a value prediction network in the value prediction model to achieve the fusion of scene data and non-scene data before performing display value prediction, thereby more fully mining information that affects display value from scene data and improving the accuracy of value prediction.
[0063] In some optional implementations, such as Figure 4 As shown, the feature fusion network 201 includes a scene feature extraction subnetwork 2011 and a non-scene feature extraction subnetwork 2012.
[0064] like Figure 5 As shown, step 1021 includes:
[0065] Step 10211: Use the scene feature extraction sub-network to extract scene features from the scene data to obtain scene features.
[0066] like Figure 4 As shown, the scene feature extraction subnetwork receives input scene data and can include a multi-layer neural network, such as a Relu activation function layer and a sigmoid activation function layer. The scene feature extraction subnetwork extracts scene-related features from the scene data and fuses them with non-scene features as weights, enhancing the model's ability to differentiate between scene-related information.
[0067] In step 10212, the non-scene feature extraction sub-network is used to extract scene features from the non-scene data to obtain non-scene features.
[0068] like Figure 4 As shown, the non-scene feature extraction subnetwork receives input non-scene data. The non-scene feature extraction subnetwork can also include a multi-layer neural network to extract features related to the target user's behavior outside the target display scene from the non-scene data.
[0069] Step 10213: fuse the scene features with the non-scene features to obtain fused features.
[0070] The fusion of scene features and non-scene features can be achieved in various ways. For example, scene features and non-scene features can be directly merged to obtain fused features. Alternatively, scene features can be used as weights and multiplied with non-scene features to obtain fused features.
[0071] This embodiment sets a scene feature extraction subnetwork and a non-scene feature extraction subnetwork in the value prediction model, which can extract scene features and non-scene features from scene data and non-scene data, and fuse the two features. The obtained fused features contain differentiated information of the target display scene, so that the display value score of the target user can be determined more accurately in the target display scene.
[0072] In some optional implementations, step 10211 may be performed as follows:
[0073] First, the non-scene data and the scene data are merged to obtain merged data.
[0074] The non-scene data and the scene data can be directly merged to obtain the merged data; or the non-scene data and the scene data can be processed to a certain extent (such as filtering out invalid data, normalizing, etc.) and then merged.
[0075] Then, the merged data is input into the scene feature extraction sub-network to obtain the scene features.
[0076] like Figure 6 As shown, the scene feature extraction sub-network receives the merged data and extracts scene features from the merged data.
[0077] This embodiment merges non-scene data and scene data, and then uses a scene feature extraction sub-network to extract scene features from the merged data. This allows the scene feature extraction sub-network to use non-scene data as prior information for scene feature extraction, and uses the prior information to influence the scene features to a certain extent, thereby helping to make the extracted scene features more closely match the actual scene and improving the accuracy of value prediction.
[0078] In some optional implementations, such as Figure 7 As shown, the value prediction network 202 includes a multi-task shared sub-network 2021 and at least two value prediction sub-networks 2022.
[0079] like Figure 8 As shown, step 1022 includes:
[0080] In step 10221, the multi-task shared sub-network is used to extract the fused features to obtain the multi-task basic features.
[0081] like Figure 7As shown, the architecture of the value prediction model can be implemented based on a multi-task model architecture such as MMoE. The multi-task shared sub-network can include multiple layers of deep neural networks. The fused features are calculated by the multi-task shared sub-network, allowing multiple networks performing specific prediction tasks to share the extracted multi-task basic features.
[0082] Step 10222: Use each value prediction subnetwork in the at least two value prediction subnetworks to perform value prediction on the multi-task basic features to obtain at least two display value prediction values.
[0083] like Figure 7 As shown, at least two value prediction subnetworks (including value prediction subnetwork 1 - value prediction subnetwork N, where N is a preset integer) respectively receive the input multi-task basic features and respectively perform corresponding prediction tasks (such as revenue prediction task, play duration prediction task, play frequency prediction task, etc.), and output display value prediction value 1 - display value prediction value N. When the value prediction model is implemented based on a multi-task model architecture such as MMoE, the at least two value prediction subnetworks are constructed in the form of a task tower.
[0084] Step 10223: Fusing at least two display value prediction values to obtain a display value score.
[0085] Specifically, a weight value corresponding to each display value prediction value can be set, and the weighted sum of each display value prediction value can be performed to obtain a display value score; or on the basis of the weighted sum, further calculation can be performed according to a preset numerical conversion strategy (such as normalization, etc.) to obtain a display value score.
[0086] This embodiment achieves multi-dimensional prediction of fusion features by setting a multi-task shared subnetwork and at least two value prediction subnetworks in the value prediction model. The multiple display value prediction values obtained can more comprehensively represent the potential value of the target user, thereby improving the accuracy of the display value score prediction.
[0087] In some optional implementations, step 10222 may be performed as follows:
[0088] First, obtain the scene features corresponding to the scene data.
[0089] Among them, scene features can be extracted from scene data. Figure 4 The scene feature extraction subnetwork shown extracts scene features from scene data.
[0090] Then, the multi-task basic features and the scene features are input into each value prediction sub-network of at least two value prediction sub-networks to perform value prediction on the multi-task basic features to obtain at least two display value prediction values.
[0091] like Figure 9 As shown, at least two value prediction subnetworks receive input multi-task basic features and scene features, respectively, and perform corresponding prediction tasks. Each value prediction subnetwork can introduce scene features, use them as weights for internal operations within the value prediction subnetwork, and multiply them with the features output by each layer in the value prediction subnetwork. The resulting data is then used to predict the display value.
[0092] This embodiment adds scene features to multi-task prediction, fully mines scene-related features from scene data, enhances the model's differentiated learning ability for scene-related information, and improves the accuracy of display value score prediction.
[0093] In some optional implementations, such as Figure 10 As shown, at least two value prediction subnetworks include a usage time prediction subnetwork and at least one revenue prediction subnetwork (including Figure 10 The revenue prediction subnetwork shown is the revenue prediction subnetwork M, where M is a preset integer).
[0094] Step 10222 may be performed as follows:
[0095] First, using each revenue prediction subnetwork in at least one revenue prediction subnetwork, revenue value prediction is performed on the multi-task basic features respectively to obtain at least one revenue prediction value as the display value prediction value corresponding to the at least one revenue prediction subnetwork.
[0096] The revenue forecast value represents the revenue that the advertiser will earn from the target user by using the target display information within a first preset time period (e.g., the next 7 days). For example, if the target display information is an advertisement, the revenue forecast value represents the estimated revenue that the advertiser will earn from the target user by placing the advertisement.
[0097] Each revenue prediction sub-network can perform a task to predict a specific type of revenue. For example, multiple revenue sub-networks can be used to perform tasks such as total revenue prediction, membership revenue prediction, purchase revenue prediction, and on-demand revenue prediction.
[0098] Then, the usage time prediction network is used to predict the usage time of the multi-task basic features to obtain the predicted usage time of the target user.
[0099] The predicted usage duration indicates the duration that the target user uses the object indicated by the target display information within a second preset duration in the future.
[0100] The second preset duration may be the same as or different from the first preset duration. The type of predicted usage duration may correspond to the actual type of the target display information. For example, if the target display information is information displaying a video playback website, the predicted usage duration may be the duration the target user accesses the video playback website and plays the video. For another example, if the target display information is information displaying a software download link, the predicted usage duration may be the duration the target user uses the software after downloading it.
[0101] The value prediction model provided in this embodiment can mark the actual benefits and actual usage time in real scenarios when it is pre-trained. By adjusting the model parameters, the error between the benefit prediction value and predicted usage time output by the model and the marked benefit value and usage time can be minimized.
[0102] like Figure 10 As shown, the output of each revenue forecast value and predicted usage time are integrated and calculated to obtain the display value score.
[0103] This embodiment sets up at least one revenue prediction subnetwork and a usage duration prediction subnetwork to achieve the use of the value prediction model to predict the revenue and usage duration brought to the target user. By using revenue and duration, the value generated after the target display information is delivered to the target user can be more accurately represented, thereby further improving the accuracy of the model's value prediction.
[0104] Figure 11 A schematic diagram of the structure of a display information push device provided in an embodiment of the present application. Specifically comprising: an acquisition module 1101, for acquiring scene data and non-scene data corresponding to a target user, wherein scene data represents data recorded for operations performed by the target user on display information of a preset type in a target display scene, and non-scene data represents data recorded for the target user's behavior outside the target display scene; a prediction module 1102, for using a pre-trained value prediction model to predict the display value of scene data and non-scene data to obtain a display value score for the target user, wherein the display value score represents the probability of the target user performing a preset operation after viewing the target display information; and a sending module 1103, for sending the target display information to the terminal used by the target user if the display value score meets the preset display conditions.
[0105] In some optional implementations, the value prediction model includes a feature fusion network and a value prediction network; the prediction module includes: a fusion unit, which uses the feature fusion network to perform feature fusion on scene data and non-scene data to obtain fusion features; and a prediction unit, which uses the value prediction network to perform display value prediction on the fusion features to obtain the display value score of the target user.
[0106] In some optional implementations, the feature fusion network includes a scene feature extraction subnetwork and a non-scene feature extraction subnetwork; the fusion unit includes: a first extraction subunit, used to use the scene feature extraction subnetwork to extract scene features from scene data to obtain scene features; a second extraction subunit, used to use the non-scene feature extraction subnetwork to extract scene features from non-scene data to obtain non-scene features; and a first fusion subunit, used to fuse scene features with non-scene features to obtain fused features.
[0107] In some optional implementations, the first extraction subunit is further used to: merge the non-scene data and the scene data to obtain merged data; and input the merged data into the scene feature extraction subnetwork to obtain scene features.
[0108] In some optional implementations, the value prediction network includes a multi-task shared subnetwork and at least two value prediction subnetworks; the prediction unit includes: a third extraction subunit, used to use the multi-task shared subnetwork to extract features from the fused features to obtain multi-task basic features; a prediction subunit, used to use each value prediction subnetwork in the at least two value prediction subnetworks to perform value prediction on the multi-task basic features to obtain at least two display value prediction values; and a second fusion subunit, used to fuse the at least two display value prediction values to obtain a display value score.
[0109] In some optional implementations, the prediction subunit is further used to: obtain scene features corresponding to the scene data; input the multi-task basic features and the scene features into each value prediction subnetwork in at least two value prediction subnetworks to perform value prediction on the multi-task basic features to obtain at least two display value prediction values.
[0110] In some optional implementations, at least two value prediction subnetworks include a usage time prediction subnetwork and at least one revenue prediction subnetwork; the prediction sub-unit is further used to: use each revenue prediction subnetwork in the at least one revenue prediction subnetwork to perform revenue value prediction on the multi-task basic features respectively, and obtain at least one revenue prediction value as the display value prediction value corresponding to the at least one revenue prediction subnetwork, wherein the revenue prediction value represents the size of the revenue obtained from the target user by using the target display information within a first preset time period in the future; use the usage time prediction network to predict the usage time of the multi-task basic features, and obtain the predicted usage time of the target user, wherein the predicted usage time represents the time the target user uses the object indicated by the target display information within a second preset time period in the future.
[0111] The display information pushing device provided in this embodiment can be as follows Figure 11The display information pushing device shown in can execute all the steps of the above display information pushing methods, thereby achieving the technical effects of the above display information pushing methods. For details, please refer to the above related descriptions. For the sake of brevity, they will not be repeated here.
[0112] Figure 12 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. Figure 12 The electronic device 1200 shown includes: at least one processor 1201, a memory 1202, at least one network interface 1204 and another user interface 1203. The various components in the electronic device 1200 are coupled together via a bus system 1205. It is understood that the bus system 1205 is used to achieve connection and communication between these components. In addition to including a data bus, the bus system 1205 also includes a power bus, a control bus, and a status signal bus. However, for the sake of clarity, the bus system 1205 is not described in detail. Figure 12 Various buses are labeled as bus system 1205.
[0113] The user interface 1203 may include a display, a keyboard, or a pointing device (eg, a mouse, a trackball, a touchpad, or a touch screen).
[0114] It is understood that the memory 1202 in the embodiment of the present application can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories. Among them, the non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory can be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of RAM are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDRSDRAM), enhanced synchronous DRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct RAM bus random access memory (DRRAM). The memory 1202 described herein is intended to include, but is not limited to, these and any other suitable types of memory.
[0115] In some embodiments, the memory 1202 stores the following elements, executable units, or data structures, or a subset thereof, or an extended set thereof: an operating system 12021 and application programs 12022 .
[0116] Among them, the operating system 12021 includes various system programs, such as the framework layer, core library layer, and driver layer, which are used to implement various basic services and handle hardware-based tasks. Application 12022 includes various application programs, such as media players and browsers, which are used to implement various application services. The program that implements the method of the embodiment of the present application can be included in application 12022.
[0117] In this embodiment, by calling a program or instruction stored in the memory 1202, specifically, a program or instruction stored in the application 12022, the processor 1201 is configured to execute the method steps provided in each method embodiment, for example, including:
[0118] Obtain the scenario data and non-scenario data corresponding to the target user, wherein the scenario data represents the data recorded for the target user's operations on the preset type of display information in the target display scenario, and the non-scenario data represents the data recorded for the target user's behavior outside the target display scenario; use the pre-trained value prediction model to predict the display value of the scenario data and the non-scenario data to obtain the display value score of the target user, wherein the display value score represents the probability of the target user performing the preset operation after viewing the target display information; if the display value score meets the preset display conditions, send the target display information to the terminal used by the target user.
[0119] The methods disclosed in the above embodiments of the present application can be applied to or implemented by processor 1201. Processor 1201 may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in processor 1201 or by software instructions. The above processor 1201 may be a general-purpose processor, a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The various methods, steps, and logic block diagrams disclosed in the embodiments of the present application can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of the present application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software units in the decoding processor. The software units can be located in a storage medium mature in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory 1202 , and the processor 1201 reads the information in the memory 1202 and completes the steps of the above method in combination with its hardware.
[0120] It is understood that the embodiments described herein may be implemented using hardware, software, firmware, middleware, microcode, or a combination thereof. For hardware implementation, the processing unit may be implemented in one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field-programmable gate arrays (FPGAs), general-purpose processors, controllers, microcontrollers, microprocessors, or other electronic units or combinations thereof for performing the above-mentioned functions of the present application.
[0121] For software implementation, the techniques described above can be implemented by a unit that performs the functions described above. The software code can be stored in a memory and executed by a processor. The memory can be implemented in the processor or external to the processor.
[0122] The electronic device provided in this embodiment may be Figure 12 The electronic device shown in can execute all the steps of the above-mentioned display information push methods, thereby achieving the technical effects of the above-mentioned display information push methods. Please refer to the above related description for details. For the sake of brevity, it will not be repeated here.
[0123] The present application also provides a storage medium (computer-readable storage medium). The storage medium stores one or more programs. The storage medium may include volatile memory, such as random access memory; the memory may also include non-volatile memory, such as read-only memory, flash memory, hard disk, or solid-state drive; and the memory may also include a combination of the aforementioned types of memory.
[0124] When one or more programs in the storage medium can be executed by one or more processors, the above-mentioned display information push method executed on the electronic device side can be implemented.
[0125] The processor is configured to execute a program stored in the memory to implement the following steps of a method for pushing display information executed on the electronic device side:
[0126] Obtain the scenario data and non-scenario data corresponding to the target user, wherein the scenario data represents the data recorded for the target user's operations on the preset type of display information in the target display scenario, and the non-scenario data represents the data recorded for the target user's behavior outside the target display scenario; use the pre-trained value prediction model to predict the display value of the scenario data and the non-scenario data to obtain the display value score of the target user, wherein the display value score represents the probability of the target user performing the preset operation after viewing the target display information; if the display value score meets the preset display conditions, send the target display information to the terminal used by the target user.
[0127] Professionals should also be further aware that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the above description has generally described the components and steps of each example according to their functions. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians can use different circuits to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.
[0128] The steps of the circuits or algorithms described in conjunction with the embodiments disclosed herein may be implemented using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.
[0129] It should be understood that the terms used herein are for the purpose of describing specific example embodiments only and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used herein may also be meant to include plural forms. The terms "comprise", "include", "contain" and "have" are inclusive and therefore specify the presence of stated features, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The steps, processes, and operations described herein are not to be construed as necessarily requiring them to be performed in the specific order described or illustrated, unless the order of execution is clearly indicated. It should also be understood that additional or alternative steps may be used.
[0130] The foregoing is merely a list of specific embodiments of the present application, intended to enable those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the broadest scope consistent with the principles and novel features of the present application.
Claims
1. A method for pushing display information, characterized in that: The method comprises: Obtaining scenario data and non-scenario data corresponding to the target user, wherein the scenario data represents data recorded for operations performed by the target user on preset types of display information in a target display scenario, and the non-scenario data represents data recorded for the target user's behavior outside the target display scenario; Using a pre-trained value prediction model, the display value of the scene data and the non-scene data is predicted to obtain the display value score of the target user, wherein the display value score represents the probability of the target user performing a preset operation after viewing the target display information; If the display value score meets the preset display conditions, the target display information is sent to the terminal used by the target user.
2. The method according to claim 1, characterized in that The value prediction model includes a feature fusion network and a value prediction network; The method of using a pre-trained value prediction model to predict the display value of the scene data and the non-scene data to obtain the display value score of the target user includes: Using the feature fusion network, performing feature fusion on the scene data and the non-scene data to obtain fused features; The value prediction network is used to predict the display value of the fusion feature to obtain the display value score of the target user.
3. The method according to claim 2, characterized in that The feature fusion network includes a scene feature extraction subnetwork and a non-scene feature extraction subnetwork; The step of utilizing the feature fusion network to fuse the scene data and the non-scene data to obtain fused features includes: Utilizing the scene feature extraction subnetwork, performing scene feature extraction on the scene data to obtain scene features; Using the non-scene feature extraction subnetwork, extracting scene features from the non-scene data to obtain non-scene features; The scene feature and the non-scene feature are fused to obtain the fused feature.
4. The method according to claim 3, characterized in that The step of extracting scene features from the scene data using the scene feature extraction subnetwork to obtain scene features includes: Merging the non-scene data and the scene data to obtain merged data; The merged data is input into the scene feature extraction sub-network to obtain the scene features.
5. The method according to claim 2, characterized in that The value prediction network includes a multi-task shared subnetwork and at least two value prediction subnetworks; The step of using the value prediction network to predict the display value of the fusion feature to obtain the display value score of the target user includes: Using the multi-task shared sub-network, extracting features from the fused features to obtain multi-task basic features; Using each value prediction subnetwork of the at least two value prediction subnetworks, perform value prediction on the multi-task basic features to obtain at least two display value prediction values; The at least two display value prediction values are fused to obtain the display value score.
6. The method according to claim 5, characterized in that The method of using each value prediction subnetwork of the at least two value prediction subnetworks to perform value prediction on the multi-task basic features to obtain at least two display value prediction values includes: Obtaining scene features corresponding to the scene data; The multi-task basic features and the scene features are input into each value prediction sub-network of the at least two value prediction sub-networks to perform value prediction on the multi-task basic features to obtain at least two display value prediction values.
7. The method according to claim 5 or 6, characterized in that The at least two value prediction subnetworks include a usage time prediction subnetwork and at least one revenue prediction subnetwork; The method of using each value prediction subnetwork of the at least two value prediction subnetworks to perform value prediction on the multi-task basic features to obtain at least two display value prediction values includes: Using each revenue prediction subnetwork in the at least one revenue prediction subnetwork, respectively predicting revenue values for the multi-task basic features, to obtain at least one revenue prediction value as a display value prediction value corresponding to the at least one revenue prediction subnetwork, wherein the revenue prediction value represents the amount of revenue obtained from the target user by using the target display information within a first preset time period in the future; The usage time prediction network is used to predict the usage time of the multi-task basic features to obtain the predicted usage time of the target user, wherein the predicted usage time represents the time the target user uses the object indicated by the target display information within a second preset time period in the future.
8. A display information push device, characterized in that: The device comprises: an acquisition module, configured to acquire scenario data and non-scenario data corresponding to a target user, wherein the scenario data represents data recorded regarding operations performed by the target user on display information of a preset type in a target display scenario, and the non-scenario data represents data recorded regarding the target user's behavior outside the target display scenario; A prediction module, configured to use a pre-trained value prediction model to predict the display value of the scenario data and the non-scenario data, and obtain a display value score for the target user, wherein the display value score represents the probability that the target user will perform a preset operation after viewing the target display information; The sending module is configured to send the target display information to a terminal used by the target user if the display value score meets a preset display condition.
9. An electronic device, characterized in that: include: memory for storing computer programs; The processor is configured to execute the computer program stored in the memory, and when the computer program is executed, the method for pushing display information according to any one of claims 1 to 7 is implemented.
10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the display information pushing method according to any one of claims 1 to 7 is implemented.