Training method of prediction model applied to page display and page display method

Abstract

This disclosure provides a training method, page display method, apparatus, device, and medium for a prediction model applied to page display, relating to the field of artificial intelligence technology, and particularly to the field of page display and content recommendation technology. The implementation scheme is as follows: based on the sample user characteristics of a sample user, the sample page characteristics of a sample page including multiple content blocks, and the sample content characteristics of each content block, first prediction information representing a first probability of a sample user performing a preset interactive behavior on a content block when browsing a sample page, and second prediction information representing the predicted statistical information of the sample user's behavior when browsing a sample page are output respectively; based on the first annotation information and the first prediction information corresponding to each content block, a first loss value is calculated; based on the second annotation information and the second prediction information corresponding to each content block, a second loss value is calculated; and based on the first loss value and the second loss value, the parameters of the prediction model are adjusted.

Description

Technical Field

[0001] This disclosure relates to the field of artificial intelligence technology, and more particularly to the field of page display and content recommendation technology. Specifically, it relates to a training method for a prediction model applied to page display, a page display method and apparatus, an electronic device, a computer-readable storage medium, and a computer program product. Background Technology

[0002] Artificial intelligence (AI) is the study of enabling computers to simulate certain human thought processes and intelligent behaviors (such as learning, reasoning, thinking, and planning). It encompasses both hardware and software technologies. AI hardware technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, and big data processing. AI software technologies mainly include computer vision, speech recognition, natural language processing, machine learning / deep learning, big data processing, and knowledge graph technologies.

[0003] In conventional technologies, recommendation systems typically make probabilistic predictions based on a user's single point of satisfaction interaction with a particular content resource. However, they do not take into account the comprehensive analysis of user behavioral characteristics at different granularities, and thus fail to accurately capture the user's true preferences and satisfaction levels, thereby limiting the effectiveness of content recommendation.

[0004] The methods described in this section are not necessarily methods that had been previously conceived or adopted. Unless otherwise specified, no method described in this section should be assumed to be prior art simply because it is included in this section. Similarly, unless otherwise specified, the issues mentioned in this section should not be considered to be accepted in any prior art. Summary of the Invention

[0005] This disclosure provides a method, apparatus, electronic device, computer-readable storage medium, and computer program product for training a prediction model applied to page display.

[0006] According to one aspect of this disclosure, a training method for a prediction model applied to page display is provided, wherein the prediction model includes a first prediction network and a second prediction network, and the method includes: based on sample user features of sample users, sample page features of a sample page including multiple content blocks, and sample content features of each content block in the multiple content blocks, outputting first prediction information using the first prediction network and outputting second prediction information using the second prediction network, wherein the first prediction information represents a first probability that the sample user performs a preset interactive behavior on the content block when browsing the sample page, and the second prediction information represents the predicted behavioral statistics of the sample user when browsing the sample page, wherein the behavioral statistics... The statistical information includes statistical information related to the number of times the sample user performs the preset interactive behavior on the sample page; a first loss value is calculated based on the first annotation information corresponding to each content block and the first prediction information corresponding to each content block, wherein the first annotation information corresponding to each content block represents whether the sample user performs the preset interactive behavior on the content block when browsing the sample page; a second loss value is calculated based on the second annotation information and the second prediction information corresponding to each content block, wherein the second annotation information represents the true value of the behavioral statistics information of the sample user when browsing the sample page; and the parameters of the prediction model are adjusted based on the first loss value and the second loss value.

[0007] According to one aspect of this disclosure, a page display method is provided, comprising: for each of a plurality of content blocks associated with a target page, determining first prediction information and second prediction information based on user characteristics of a target user, page characteristics of the target page, and content characteristics of the content block, wherein the first prediction information characterizes a first probability that the target user performs a preset interactive behavior on the content block when browsing the target page, and the second prediction information characterizes predicted behavioral statistics of the target user when browsing the target page including the content block, the behavioral statistics including statistical information related to the number of times the target user performs the preset interactive behavior on the target page; and determining a display strategy for the target page based on the first prediction information and the second prediction information.

[0008] According to one aspect of this disclosure, a prediction model for page display is provided, comprising: a first prediction network for outputting first prediction information based on user characteristics of a target user, page characteristics of a target page, and content characteristics of a content block associated with the target page, wherein the first prediction information characterizes a first probability that the target user performs a preset interactive behavior on the content block when browsing the target page; and a second prediction network for outputting second prediction information based on the user characteristics, the page characteristics, and the content characteristics, wherein the second prediction information characterizes predicted behavioral statistics of the target user when browsing the target page including the content block, the behavioral statistics including statistical information related to the number of times the target user performs the preset interactive behavior on the target page.

[0009] According to one aspect of this disclosure, a training apparatus for a prediction model applied to page display is provided, wherein the prediction model includes a first prediction network and a second prediction network, and the apparatus includes: an output unit configured to output first prediction information using the first prediction network and second prediction information using the second prediction network based on sample user features of a sample user, sample page features of a sample page including multiple content blocks, and sample content features of each content block among the multiple content blocks; wherein the first prediction information represents a first probability that the sample user performs a preset interactive behavior on the content block when browsing the sample page, and the second prediction information represents predicted behavioral statistics of the sample user when browsing the sample page, the behavioral statistics including information related to the sample user's behavior when browsing the sample page. The system includes: statistical information related to the number of times a user performs the preset interactive behavior on the sample page; a first calculation unit configured to calculate a first loss value based on first annotation information corresponding to each content block and first prediction information corresponding to each content block, wherein the first annotation information corresponding to each content block characterizes whether the sample user performs the preset interactive behavior on the content block when browsing the sample page; a second calculation unit configured to calculate a second loss value based on second annotation information and second prediction information corresponding to each content block, wherein the second annotation information characterizes the true value of the behavioral statistics information of the sample user when browsing the sample page; and an adjustment unit configured to adjust the parameters of the prediction model based on the first loss value and the second loss value.

[0010] According to one aspect of this disclosure, a page display apparatus is provided, comprising: a first determining unit configured to, for each of a plurality of content blocks associated with a target page, determine first prediction information and second prediction information based on user characteristics of a target user, page characteristics of the target page, and content characteristics of the content block, wherein the first prediction information characterizes a first probability that the target user performs a preset interactive behavior on the content block when browsing the target page, and the second prediction information characterizes predicted behavioral statistics of the target user when browsing the target page including the content block, the behavioral statistics including statistical information related to the number of times the target user performs the preset interactive behavior on the target page; and a second determining unit configured to determine a display strategy for the target page based on the first prediction information and the second prediction information.

[0011] According to one aspect of this disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the training method or page display method described above for a prediction model applied to page display.

[0012] According to one aspect of this disclosure, a non-transitory computer-readable storage medium is provided storing computer instructions, wherein the computer instructions are used to cause the computer to execute the training method or page display method of the prediction model applied to page display described above.

[0013] According to one aspect of this disclosure, a computer program product is provided, comprising a computer program, wherein the computer program, when executed by a processor, is capable of implementing the training method or page display method described above for a prediction model applied to page display.

[0014] According to one or more embodiments of this disclosure, the page content that users prefer can be predicted more accurately, thereby improving the accuracy of page content display.

[0015] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this disclosure, nor is it intended to limit the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0016] The accompanying drawings exemplify embodiments and form part of the specification, serving together with the textual description to explain exemplary implementations of the embodiments. The illustrated embodiments are for illustrative purposes only and do not limit the scope of the claims. Throughout the drawings, the same reference numerals refer to similar but not necessarily identical elements.

[0017] Figure 1 A schematic diagram of an exemplary system in which various methods described herein may be implemented, according to exemplary embodiments of the present disclosure; Figure 2 A flowchart is shown of a training method for a prediction model applied to a page display according to an exemplary embodiment of the present disclosure; Figure 3 A flowchart of a page display method according to an exemplary embodiment of the present disclosure is shown; Figure 4 A schematic diagram of the structure of a prediction model according to an exemplary embodiment of the present disclosure is shown; Figure 5 A structural block diagram of a training apparatus for a prediction model applied to a page display according to an exemplary embodiment of the present disclosure is shown; Figure 6 A structural block diagram of a page display apparatus according to an exemplary embodiment of the present disclosure is shown; Figure 7 A structural block diagram of an exemplary electronic device that can be used to implement embodiments of the present disclosure is shown. Detailed Implementation

[0018] The exemplary embodiments of this disclosure are described below with reference to the accompanying drawings, including various details of the embodiments to aid understanding, and should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope of this disclosure. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0019] In this disclosure, unless otherwise stated, the use of terms such as "first," "second," etc., to describe various elements is not intended to limit the positional, temporal, or importance relationships of these elements; such terms are merely used to distinguish one element from another. In some examples, the first element and the second element may refer to the same instance of that element, while in other cases, based on the context, they may refer to different instances.

[0020] The terminology used in the description of the various examples described in this disclosure is for the purpose of describing particular examples only and is not intended to be limiting. Unless the context explicitly indicates otherwise, an element may be one or more unless the number of elements is specifically limited. Furthermore, the term "and / or" as used in this disclosure covers any one of the listed items and all possible combinations thereof.

[0021] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings.

[0022] Figure 1 A schematic diagram of an exemplary system 100 in which the various methods and apparatus described herein can be implemented according to embodiments of this disclosure is shown. Reference Figure 1 The system 100 includes one or more client devices 101, 102, 103, 104, 105 and 106, a server 120, and one or more communication networks 110 coupling the one or more client devices to the server 120. The client devices 101, 102, 103, 104, 105 and 106 can be configured to execute one or more applications.

[0023] In embodiments of this disclosure, server 120 may run one or more services or software applications that enable the execution of training methods for a predictive model applied to page display.

[0024] In some embodiments, server 120 may also provide other services or software applications, which may include non-virtual and virtual environments. In some embodiments, these services may be provided as web-based services or cloud services, such as to users of client devices 101, 102, 103, 104, 105 and / or 106 under a Software as a Service (SaaS) model.

[0025] exist Figure 1 In the configuration shown, server 120 may include one or more components that implement the functions performed by server 120. These components may include software components, hardware components, or combinations thereof that can be executed by one or more processors. Users operating client devices 101, 102, 103, 104, 105, and / or 106 can sequentially interact with server 120 using one or more client applications to utilize the services provided by these components. It should be understood that various different system configurations are possible and may differ from system 100. Therefore, Figure 1 This is an example of a system used to implement the various methods described herein, and is not intended to be limiting.

[0026] Users can use client devices 101, 102, 103, 104, 105, and / or 106 to send training sample data or feature data to be processed. The client devices can provide an interface that allows users to interact with the client devices. The client devices can also output information to the user through this interface. Although... Figure 1 Only six client devices are described, but those skilled in the art will understand that this disclosure can support any number of client devices.

[0027] Client devices 101, 102, 103, 104, 105, and / or 106 may include various categories of computer devices, such as portable handheld devices, general-purpose computers (such as personal computers and laptops), workstation computers, wearable devices, smart screen devices, self-service terminal devices, service robots, gaming systems, thin clients, various messaging devices, sensors, or other sensing devices. These computer devices can run various categories and versions of software applications and operating systems, such as Microsoft Windows, Apple iOS, UNIX-like operating systems, Linux or Linux-like operating systems (such as Google Chrome OS); or include various mobile operating systems, such as Microsoft Windows Mobile OS, iOS, Windows Phone, and Android. Portable handheld devices may include cellular phones, smartphones, tablets, personal digital assistants (PDAs), etc. Wearable devices may include head-mounted displays (such as smart glasses) and other devices. Gaming systems may include various handheld gaming devices, internet-enabled gaming devices, etc. Client devices are capable of executing various applications, such as various internet-related applications, communication applications (such as email applications), short message service (SMS) applications, and can use various communication protocols.

[0028] Network 110 can be any type of network well known to those skilled in the art, and can support data communication using any of a variety of available protocols (including, but not limited to, TCP / IP, SNA, IPX, etc.). By way of example only, one or more networks 110 can be a local area network (LAN), an Ethernet-based network, a token ring network, a wide area network (WAN), the Internet, a virtual network, a virtual private network (VPN), an intranet, an extranet, a blockchain network, a public switched telephone network (PSTN), an infrared network, a wireless network (e.g., Bluetooth, WIFI), and / or any combination of these and / or other networks.

[0029] Server 120 may include one or more general-purpose computers, special-purpose server computers (e.g., PC (personal computer) servers, UNIX servers, mid-range servers), blade servers, mainframe computers, server clusters, or any other suitable arrangement and / or combination. Server 120 may include one or more virtual machines running a virtual operating system, or other computing architectures involving virtualization (e.g., one or more flexible pools of logical storage devices that can be virtualized to maintain virtual storage devices for servers). In various embodiments, server 120 may run one or more services or software applications that provide the functionality described below.

[0030] The computing unit in server 120 can run one or more operating systems, including any of the aforementioned operating systems and any commercially available server operating system. Server 120 can also run any of a variety of additional server applications and / or middleware applications, including HTTP servers, FTP servers, CGI servers, JAVA servers, database servers, etc.

[0031] In some implementations, server 120 may include one or more applications to analyze and merge data feeds and / or event updates received from users of client devices 101, 102, 103, 104, 105, and 106. Server 120 may also include one or more applications to display data feeds and / or real-time events via one or more display devices of client devices 101, 102, 103, 104, 105, and 106.

[0032] In some implementations, server 120 can be a server for a distributed system or a server integrated with blockchain. Server 120 can also be a cloud server, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology. A cloud server is a host product in the cloud computing service system, designed to address the shortcomings of traditional physical hosts and Virtual Private Server (VPS) services, such as high management difficulty and weak business scalability.

[0033] System 100 may also include one or more databases 130. In some embodiments, these databases may be used to store data and other information. For example, one or more of the databases 130 may be used to store information such as audio files and video files. Databases 130 may reside in various locations. For example, a database used by server 120 may be local to server 120, or it may be located away from server 120 and may communicate with server 120 via a network-based or dedicated connection. Databases 130 may be of different categories. In some embodiments, the database used by server 120 may be, for example, a relational database. One or more of these databases may store, update, and retrieve data from and from the databases in response to commands.

[0034] In some embodiments, one or more of the databases 130 may also be used by an application to store application data. The databases used by the application may be different categories of databases, such as key-value stores, object stores, or regular stores supported by a file system.

[0035] Figure 1The system 100 can be configured and operated in various ways to enable the application of the various methods and apparatus described in this disclosure.

[0036] In related technologies, recommender systems typically employ a single-objective modeling mechanism. Existing recommender models often only predict the probability of a user's single satisfying interaction with a specific content block (such as a single click or a single like). However, this conventional approach lacks comprehensive analysis of user behavioral characteristics at different granularities, particularly ignoring macro-level behavioral statistics for the entire page (such as the total number of user interactions and total dwell time). For example, a user may click on a content block on the page but not browse or interact with other content. A single click prediction cannot reflect the user's overall satisfaction with the page, making it difficult for the recommender system to accurately capture the user's true preferences and satisfaction level, thus limiting the effectiveness of content recommendations.

[0037] Based on this, this disclosure provides a training method for a predictive model applied to a page display system. It simultaneously predicts the user's single-point behavior for a single content block and more macroscopic page-level behavioral statistics. By predicting page-level behavioral statistics, it indicates the user's satisfaction with the entire page content. By utilizing the above two dimensions of information, it more comprehensively portrays the user's single-point satisfaction experience and overall satisfaction experience with the page content, so as to improve the content recommendation effect when applied to the page display system.

[0038] Figure 2 A flowchart is shown of a training method 200 for a prediction model applied to a page display according to an exemplary embodiment of the present disclosure. Figure 2 As shown, method 200 includes: Step S201: Based on the sample user characteristics of the sample user, the sample page characteristics of the sample page including multiple content blocks, and the sample content characteristics of each content block in the multiple content blocks, the first prediction network is used to output the first prediction information and the second prediction network is used to output the second prediction information. The first prediction information represents the first probability that the sample user performs a preset interactive behavior on the content block when browsing the sample page, and the second prediction information represents the predicted behavioral statistics of the sample user when browsing the sample page. The behavioral statistics include statistics related to the number of times the sample user performs the preset interactive behavior on the sample page. Step S202: Calculate the first loss value based on the first annotation information corresponding to each content block and the first prediction information corresponding to each content block, wherein the first annotation information corresponding to each content block represents whether the sample user performs a preset interactive behavior for the content block when browsing the sample page; Step S203: Based on the second annotation information and the second prediction information corresponding to each content block, calculate the second loss value, wherein the second annotation information represents the true value of the behavioral statistics of the sample user when browsing the sample page; and Step S204: Adjust the parameters of the prediction model based on the first loss value and the second loss value.

[0039] By applying the above method 200, the first prediction network and the second prediction network can be used to predict the interaction probability of single content blocks and page-level behavioral statistics, respectively. The model parameters are updated together with their respective loss values, so that the trained prediction model can learn the user's satisfaction preferences more comprehensively. The information of the above two dimensions can be used to more comprehensively depict the user's single-point satisfaction experience and overall satisfaction experience of the page content, so as to improve the content recommendation effect when applied to the page display system.

[0040] In some examples, the sample page can be various web pages or application interfaces that provide content display, such as landing pages, news feed pages, or detail pages in content distribution platforms. The multiple content blocks contained in the sample page can be different content display formats, such as landing page image and text cards, short video components, multi-image carousel components, or plain text paragraphs. In one example, the content blocks are arranged sequentially in a certain layout on the page for user browsing and interaction.

[0041] In some examples, sample user features are characteristic information used to characterize the attributes and historical behavioral patterns of sample users. These may include user attributes, historical behavioral characteristics (such as the number of clicks, browsing duration, and interaction frequency within a certain time window), and user preference tag characteristics. Sample page features include characteristic information representing the overall attributes of the sample page. Specific examples may include page entry category information (such as primary or secondary categories), the title length of the page entry, the total number of impressions, clicks, interactions, and comments in the page entry history. Sample content features refer to information used to describe the attributes and quality of content blocks themselves. These may include the title, body text, tags, author information, publication time, content duration or length, and historical consumption data of the content block.

[0042] The construction of a sample dataset including the aforementioned sample feature information can be achieved by collecting records of actual user interactions with the landing page within a historical time window. In one example, user behavior data such as exposure, clicks, browsing duration, completion rate, likes, comments, and shares for each content block on the landing page can be collected. Positive and negative samples can then be defined based on these real interactive behaviors. For instance, for labeling single-point satisfaction, if a user clicks on a content block and the browsing duration or completion rate exceeds a certain threshold, or performs interactive behaviors such as liking, sharing, or commenting, it is labeled as a positive sample, indicating user satisfaction with that content block. Conversely, if a user only browses the page but does not perform any of the aforementioned interactive behaviors on that content block, it is labeled as a negative sample. For labeling overall page satisfaction, statistical information such as the cumulative number of content blocks clicked, the number of interactions, the browsing duration or browsing step length during the current page browsing process can be collected to distinguish between positive and negative samples.

[0043] Based on this, the first annotation information can be directly extracted from the user's interaction behavior records. If the user interacts with a certain content block, the first annotation information is 1; otherwise, it is 0. The second annotation information is obtained by statistically analyzing the user's behavior logs within the sample page. For example, it can be calculated by counting the total number of times the user clicks on all content blocks, or the user's total browsing time, total browsing steps, etc., thereby generating the second annotation information corresponding to the second prediction information.

[0044] In some examples, the first and second prediction networks can be constructed using a multilayer perceptron structure such as a fully connected deep neural network. Feature mapping information is extracted step by step through one or more hidden layers, and then prediction information is output in the output layer through a non-linear activation function.

[0045] In some examples, preset interactive behaviors refer to predefined, observable actions that reflect a user's level of satisfaction with the content. For example, preset interactive behaviors may include, but are not limited to, at least one of the following: clicking on a content block, spending more than a set threshold on a content block, achieving a content browsing or viewing completion rate exceeding a set threshold, liking the content, commenting on the content, or sharing the content. Preset interactive behaviors are statistically analyzed and predicted to reflect users' positive feedback and satisfaction tendencies with the page content.

[0046] In some examples, behavioral statistics refer to quantitative metrics derived from the user's actions during page browsing, used to measure user satisfaction at the overall page level. In some examples, behavioral statistics may include the total number of content blocks on the page that triggered preset interactive behaviors, such as effective clicks or in-depth reading of several content blocks, or information such as total browsing time, total swipe length, and total number of interactions. By aggregating user behavior statistics, the overall engagement and satisfaction level of the user with the content page can be reflected.

[0047] According to some embodiments, the prediction model further includes a feature encoding network, the depth of which is not less than the depth of the first prediction network or the depth of the second prediction network. The method further includes: inputting sample user features, sample page features, and sample content features into the feature encoding network to obtain the feature encoding result output by the feature encoding network. In step S201, based on the sample user features of the sample user, the sample page features of the sample page including multiple content blocks, and the sample content features of each content block in the multiple content blocks, outputting first prediction information using the first prediction network and outputting second prediction information using the second prediction network includes: inputting the feature encoding result into the first prediction network and the second prediction network respectively to obtain the first prediction information and the second prediction information.

[0048] According to the above embodiments, by introducing a feature encoding network with a depth no less than that of the first or second prediction network into the prediction model, the original, high-dimensional, sparse input features can be mapped to a low-dimensional, dense, shared feature space. This shared feature encoding mechanism allows the first and second prediction networks to perform predictions based on higher-quality abstract semantic features, rather than independently extracting information from the original features. This is equivalent to enabling feature sharing between the two prediction tasks, which helps reduce the overall number of parameters and computational overhead of the model, thereby achieving more efficient online prediction. This, in turn, facilitates more efficient determination of page display content in real-world page display scenarios. For example, page display content can be flexibly customized according to user preferences to improve user experience.

[0049] In some examples, the feature encoding network can be a deeper feedforward fully connected neural network model than the first and second prediction networks. For instance, the feature encoding network can be composed of multiple fully connected layers stacked sequentially, extracting abstract semantic representations from the input features through layer-by-layer feature transformation and combination. The shared feature encoding results output by this feature encoding network can then be fed into the subsequent first and second prediction networks.

[0050] In some examples, the calculation of the first and second loss values ​​can be based on the appropriate loss function selected according to the prediction task. For example, when modeling the two prediction information as a probabilistic classification task, the cross-entropy loss function can be used to calculate the error between the predicted probability and the labeled 0 / 1 true value, respectively. When the second prediction information is a regression numerical prediction, the mean squared error can be used to calculate the second loss value. Alternatively, log loss can also be used to calculate the loss value. As long as the prediction model can be optimized and trained based on supervised learning, this disclosure does not limit the specific form of the loss function.

[0051] In step S204, the joint loss can be determined based on the first loss value and the second loss value, the gradient can be calculated using the backpropagation algorithm, and the weights and bias parameters of the prediction model can be updated along the gradient descent direction using an optimizer such as Adam.

[0052] According to some embodiments, adjusting the parameters of the prediction model based on the first loss value and the second loss value in step S204 includes: determining the weights corresponding to the first loss value and the second loss value respectively; and adjusting the parameters of the prediction model based on the weighted calculation result of the first loss value and the second loss value.

[0053] According to the above embodiments, by setting adjustable weight coefficients for the first loss value and the second loss value respectively, the influence of the two prediction tasks on the model parameter update can be flexibly controlled during joint training, thereby better balancing the relationship between the two optimization objectives and avoiding the situation where one task dominates the training process, causing the performance of the other task to decline.

[0054] For example, in some business scenarios, if the focus is more on overall user satisfaction of the page and the model prioritizes the accuracy of predicting page-level behavioral statistics, then a higher weight can be assigned to the second loss value. Conversely, if the focus is more on the accurate recommendation and ranking of individual content blocks, then a higher weight can be assigned to the first loss value. By flexibly adjusting the weights, the same model architecture can be adapted to different product needs.

[0055] For example, a weight adjustment strategy can be set so that the relative weights of the two loss values ​​change dynamically as the training process progresses. Alternatively, the weights of the first and second loss values ​​can be automatically determined based on their numerical ratio. For instance, a smaller weight can be applied to the loss term with the larger value, and a larger weight can be applied to the loss term with the smaller value. This achieves a dynamic balance between the two loss magnitudes during training, preventing gradient suppression during backpropagation due to an excessively large loss value.

[0056] In one example, the training process of the prediction model can be implemented using multiple batches of training datasets, allowing for repeated evaluation of training effectiveness to obtain the optimal model. For instance, the AUC (Area Under Count) metric can be calculated after training on a certain number of batches of data to evaluate the model's training performance, selecting the prediction model with the highest accuracy. This leads to more accurate page content recommendations.

[0057] The trained prediction model described above can be directly applied to online page display services. For example, when a user initiates a page request, the system can obtain the user characteristics of the target user, the page characteristics of the current page, and the content characteristics of each candidate content block, and input these features into the trained prediction model. The feature encoding network in the model first processes these features to obtain shared feature encodings. Then, the first and second prediction networks output first and second prediction information for each content block based on the shared feature encodings, respectively. Based on this, the system selects the content blocks to be displayed first and determines the content layout of the page to present a page that better matches the user's preferences.

[0058] According to one aspect of this disclosure, a page display method is also provided. Figure 3 A flowchart of a page display method 300 according to an exemplary embodiment of the present disclosure is shown. Figure 3 As shown, method 300 includes: Step S301: For each content block among multiple content blocks associated with the target page, based on the user characteristics of the target user, the page characteristics of the target page, and the content characteristics of the content block, determine first prediction information and second prediction information. The first prediction information represents the first probability that the target user will perform a preset interactive behavior on the content block when browsing the target page. The second prediction information represents the predicted behavioral statistics of the target user when browsing the target page including the content block. The behavioral statistics include statistics related to the number of times the target user performs the preset interactive behavior on the target page. Step S302: Based on the first prediction information and the second prediction information, determine the display strategy for the target page.

[0059] By applying the above method 300, we can obtain single-point satisfaction prediction information and page-level behavioral statistics prediction information for each content block. We can comprehensively evaluate the presentation value of the content block from the perspectives of single-point satisfaction and overall satisfaction. Then, based on the prediction results of these two dimensions, we can formulate a display strategy to more accurately determine which content blocks should be highlighted or prioritized, thereby improving the user's overall satisfaction with the final page and the depth of browsing.

[0060] In some examples, the target page refers to the aggregated page that needs to be generated and displayed for the user, such as the landing page that the user enters after clicking on a piece of content in the recommended information stream. The landing page is used to aggregate multiple related pieces of content (e.g., arranged in the form of an information stream) for the user to swipe through.

[0061] In some examples, sample user features are feature information used to characterize the attributes and historical behavior patterns of sample users. These may include user attributes, historical behavioral features of users (such as the number of clicks, browsing duration, interaction frequency, etc. within a certain time window), and user preference tag features.

[0062] In some examples, sample page features include feature information characterizing the overall attributes of the sample page. For example, these could specifically include entry point classification information (such as primary or secondary categories), the title length of the entry point, the total number of impressions, clicks, interactions, and comments in the page's history. Multiple content blocks associated with the target page refer to the set of content for which a display strategy needs to be determined. These content blocks can be different types of content information units planned to be displayed on the landing page, i.e., various media content units used to fill the page, such as mixed text and image modules, video playback modules, and multi-image carousel cards included on the landing page. These content blocks can be pre-recalled by the upstream recall module based on user interests and page themes, and then prioritized or laid out based on the prediction information obtained in step S301 to obtain the final displayed page. In one example, these content blocks are multiple content blocks already fixed within the target page, and the system needs to decide their display order or whether to collapse some content. In another example, these content blocks are a pool of candidate content resources corresponding to the target page. The system needs to select a portion of the content from these blocks for assembly and display. In this case, the display strategy includes which content blocks to select and how to sort them.

[0063] Content features of a content block refer to the feature information that characterizes the attributes and semantics of the content block itself. For example, they can be obtained by performing feature engineering and vectorization encoding on the original attribute information of the content block, such as its title, body text, author, tags, publication time, content duration, content type, and historical click rate.

[0064] In some examples, preset interactive behaviors refer to predefined, observable actions that reflect a user's level of satisfaction with the content. For example, preset interactive behaviors may include, but are not limited to, at least one of the following: clicking on a content block, spending more than a set threshold on a content block, achieving a content browsing or viewing completion rate exceeding a set threshold, liking the content, commenting on the content, or sharing the content. Preset interactive behaviors are statistically analyzed and predicted to reflect users' positive feedback and satisfaction tendencies with the page content.

[0065] In some examples, behavioral statistics refer to quantitative metrics derived from the user's actions during page browsing, used to measure user satisfaction at the overall page level. In some examples, behavioral statistics may include the total number of content blocks on the page that triggered preset interactive behaviors, such as effective clicks or in-depth reading of several content blocks, or information such as total browsing time, total swipe length, and total number of interactions. By aggregating user behavior statistics, the overall engagement and satisfaction level of the user with the content page can be reflected.

[0066] According to some embodiments, the second prediction information includes a second probability that the predicted behavioral statistics meet preset conditions, the preset conditions including that the number of times the target user performs preset interactive behaviors on the target page is not less than a number threshold.

[0067] By applying the above method, the second prediction information is designed as the probability value of the predicted behavioral statistics meeting preset conditions. This is equivalent to transforming a page-level regression prediction task or statistical task into a binary classification task, simplifying the representation of the model output. This probabilistic output format facilitates unified processing and weighted fusion with the probability value of single-point satisfactory prediction (the first prediction information), thereby enabling more convenient multi-objective joint optimization. It also helps simplify model parameters, achieving more efficient training and inference.

[0068] In some other embodiments, the second prediction information can also be a specific predicted value of behavioral statistics, such as the model directly outputting the predicted number of user-clicked content blocks, or outputting a predicted satisfaction score. The model can flexibly choose between regression output or classification output according to actual needs.

[0069] According to some embodiments, behavioral statistics also include information on the total step length or total time a target user spends browsing the target page. This allows for a deeper understanding of user retention, accurately reflecting user satisfaction with the overall page content. For example, total step length information can record the total distance a user swipes across the entire page or the total number of content blocks viewed, while total time information can record the total time a user spends from entering the page to leaving it, enabling a more comprehensive assessment of the user's overall engagement and satisfaction.

[0070] According to some embodiments, when the second prediction information includes a second probability, the preset conditions also include total step length information not less than a step length threshold or total duration information not less than a duration threshold. Therefore, by configuring preset conditions for browsing step length or duration information, the values ​​of behavioral statistics can be mapped to the classification output results of the second prediction information, more accurately representing the user's satisfaction probability with the overall page content, and thus achieving more accurate recommendations.

[0071] According to some embodiments, in step S301, determining the first and second prediction information based on the user characteristics of the target user, the page characteristics of the target page, and the content characteristics of the content block includes: based on the user characteristics, page characteristics, and the content characteristics of the content block, outputting the first prediction information using a first prediction network and outputting the second prediction information using a second prediction network. The first and second prediction networks are trained using sample data labeled with reference output information. By applying the trained prediction networks, the two types of prediction information can be obtained more accurately and efficiently, thus achieving more efficient page display.

[0072] In some examples, the first and second prediction networks mentioned above are trained using the aforementioned training method for prediction models applied to page display (such as method 200).

[0073] According to some embodiments, method 300 further includes: inputting user features, page features, and content features of the content block into a feature encoding network to obtain feature encoding results, wherein the depth of the feature encoding network is not less than the depth of the first prediction network or the depth of the second prediction network, wherein outputting first prediction information using the first prediction network and outputting second prediction information using the second prediction network based on user features, page features, and content features of the content block includes: inputting the feature encoding results into the first prediction network and the second prediction network respectively to obtain the first prediction information and the second prediction information.

[0074] Therefore, by introducing a feature encoding network with a depth no less than that of the first or second prediction network into the prediction model, the original, high-dimensional, sparse input features can be mapped to a low-dimensional, dense, shared feature space. This shared feature encoding mechanism allows the first and second prediction networks to make predictions based on higher-quality abstract semantic features, rather than independently extracting information from the original features. This is equivalent to enabling feature sharing between the two prediction tasks, which helps reduce the overall number of parameters and computational overhead of the model, thus achieving more efficient online prediction. This, in turn, facilitates more efficient determination of page display content in real-world page display scenarios, such as flexibly customizing page display content based on user preferences to improve user experience.

[0075] In some examples, the feature encoding network described above is obtained by simultaneously optimizing the first and second prediction networks using the training method of the prediction model applied to page display.

[0076] According to some embodiments, method 300 further includes: performing normalization processing on user features, page features, and content features of the content block, wherein inputting user features, page features, and content features of the content block into a feature encoding network to obtain feature encoding results includes: inputting the normalized user features, the normalized page features, and the normalized content features of the content block into a feature encoding network to obtain feature encoding results.

[0077] By normalizing the various input features, the dimensional differences between different feature dimensions can be effectively eliminated, making the input data distribution more uniform and stable. This reduces the probability of unstable model training or slow convergence caused by excessive differences in dimensionality and value range between different features, which is conducive to obtaining a lighter prediction model and thus achieving more efficient online page content recommendation.

[0078] Normalization can be achieved in various ways. For example, for continuous behavioral features such as age, historical click count, and browsing duration, truncated normalization can be used. This involves first cropping extreme values ​​beyond a specified quantile range to the boundary values, and then performing max-min normalization or Z-score standardization. For categorical attribute features such as gender, education level, and city, they can be converted to integer codes through numerical mapping, and then these codes can be normalized. This unified preprocessing effectively eliminates differences in feature dimensions, facilitating the development of more efficient and lightweight prediction models, and ultimately achieving more efficient online page content recommendation.

[0079] According to some embodiments, determining the display strategy for a target page based on first and second prediction information includes: determining display strategies for multiple content blocks based on the first and second prediction information, wherein the display strategy for each content block includes increasing the display priority of the content block, decreasing the display priority of the content block, or not displaying the content block. Thus, priority adjustment actions for specific content blocks can be obtained based on the prediction information of the above two dimensions. For any content block, its final display priority depends not only on the single-point probability of a user clicking on the content block and making a satisfactory consumption, but also on its expected contribution to the overall satisfaction of the entire page after the content block is placed on it. This achieves content block filtering or sorting from a globally optimal perspective to obtain a target page that better matches user preferences and improves user experience.

[0080] According to some embodiments, determining a display strategy for multiple content blocks based on first and second prediction information includes: determining third prediction information for each content block in the multiple content blocks by performing a fusion calculation based on the first and second prediction information; sorting the multiple content blocks based on the third prediction information for each content block in the multiple content blocks; and determining a display strategy for the multiple content blocks based on the sorting result.

[0081] By fusing the first and second prediction information, two prediction information of different granularities can be integrated into a unified ranking score. This allows for a unified comparison and ranking of multiple content blocks, simplifying the logic for determining display strategies and enabling more efficient and convenient page content layout.

[0082] The third predictive information can be calculated in several ways. For example, it can be obtained by directly weighting and summing the first predictive information (single-point satisfaction probability) and the second predictive information (page satisfaction probability), or by multiplying the two probability values ​​and using the product as the comprehensive score. In one example, the third predictive information can be calculated by further combining the prediction results of other dimensions based on the first and second predictive information. For instance, an independent user preference prediction model can be introduced to predict whether a user will like the content, and then this probability of liking can be weighted and fused with the first and second predictive information to obtain a more accurate comprehensive ranking score.

[0083] In one example, the display strategy for determining multiple content blocks based on the first and second prediction information can also be implemented in other ways. For example, when the system determines that the first and second prediction probabilities of a candidate content are both higher than a certain probability threshold, its display strategy is set to higher priority (such as placing it on the first screen of the page). When both prediction probabilities are lower than the probability threshold, it can be determined not to display the content block, thereby reducing the possibility of invalid exposure of page content.

[0084] According to some embodiments, determining the first prediction information and the second prediction information based on the user characteristics of the target user, the page characteristics of the target page, and the content characteristics of the content block includes: determining the first prediction information and the second prediction information based on the user characteristics, the page characteristics, the content characteristics of the content block, and the scene characteristics, wherein the scene characteristics are used to characterize the scene information of the target user browsing the target page, and the scene information includes at least one of browsing time information, network environment information, or the channel information of the target user entering the target page.

[0085] By incorporating scene features into the feature input, it is possible to more accurately capture and distinguish the different page browsing intentions of the same user at different times or in different environments, thereby further improving the accuracy of content recommendation results in diverse scenarios.

[0086] In some examples, the specific content of the scene features includes, but is not limited to: channel ID (e.g., indicating which application channel the user entered the page from through one-hot encoding or binarization), time period (e.g., mapping specific hour values ​​of a day to a certain range and then normalizing them), whether it is a weekday or holiday (using 0 and 1 for binarization), network environment type (e.g., WiFi or mobile network, represented by binarization), and the number of times the user refreshes during the current session (i.e., how many requests were made during the current use, which can be normalized), etc.

[0087] In some examples, the application of scene features can be integrated into the processing steps of the aforementioned embodiments. For instance, after normalizing the scene features, they can be concatenated with user features, page features, content features, etc., and then input together into a feature encoding network for unified feature fusion encoding, so that the underlying feature encoding results take into account scene information.

[0088] According to one aspect of this disclosure, a prediction model for page display is also provided. According to an exemplary embodiment provided by this disclosure, the prediction model for page display includes a first prediction network for outputting first prediction information based on user characteristics of a target user, page characteristics of a target page, and content characteristics of content blocks associated with the target page, wherein the first prediction information characterizes a first probability that the target user performs a preset interactive behavior on a content block when browsing the target page; and a second prediction network for outputting second prediction information based on user characteristics, page characteristics, and content characteristics, wherein the second prediction information characterizes predicted behavioral statistics of the target user when browsing a target page including content blocks, and the behavioral statistics include statistics related to the number of times the target user performs a preset interactive behavior on the target page.

[0089] According to some embodiments, the prediction model further includes: a feature encoding network for outputting feature encoding results based on user features, page features, and content features, wherein the first prediction network and the second prediction network are used to output first prediction information and second prediction information respectively based on the feature encoding results.

[0090] Figure 4 A schematic diagram of the structure of a prediction model 400 applied to page display according to an exemplary embodiment of the present disclosure is shown. Figure 4 As shown, the prediction model 400 includes: a feature encoding network 410, a first prediction network 420, and a second prediction network 430. See also... Figure 4 The input end of the feature coding network 410 is used to receive input features (such as user features of the target user, page features of the target page, content features of content blocks, and possible scene features mentioned in the previous embodiments), and to perform feature extraction and fusion coding on the input features, thereby outputting the feature coding result.

[0091] The first prediction network 420 receives the aforementioned feature encoding results at its input and performs forward prediction calculations based on these results to output first prediction information (e.g., the aforementioned single-point interaction probability for a specific content block). The second prediction network 430 also receives the aforementioned feature encoding results at its input and performs forward prediction calculations based on these results to output second prediction information (e.g., the aforementioned expected achievement probability reflecting page-level behavioral statistics). Through application... Figure 4 The network architecture shown is a shared underlying feature encoding network, which utilizes two prediction networks for top-level expert prediction. The prediction model 400 can share underlying feature representations between the two tasks of single-point interaction prediction and macro-behavioral statistical prediction. It achieves more accurate dual-objective prediction while effectively reducing the overall number of model parameters and online inference overhead. It uses the information from the above two dimensions to more comprehensively characterize the user's single-point satisfaction experience and overall satisfaction experience with the page content, so as to improve the content recommendation effect when applied to the page display system.

[0092] According to one aspect of this disclosure, a training apparatus for a prediction model applied to page display is also provided, wherein the prediction model includes a first prediction network and a second prediction network. Figure 5 A structural block diagram of a training apparatus 500 for a prediction model applied to a page display according to an exemplary embodiment of the present disclosure is shown. Figure 5 As shown, the device 500 includes: The output unit 501 is configured to output first prediction information using a first prediction network and second prediction information using a second prediction network, based on the sample user characteristics of the sample user, the sample page characteristics of the sample page including multiple content blocks, and the sample content characteristics of each content block in the multiple content blocks. The first prediction information represents the first probability that the sample user performs a preset interactive behavior on the content block when browsing the sample page, and the second prediction information represents the predicted behavioral statistics of the sample user when browsing the sample page. The behavioral statistics include statistics related to the number of times the sample user performs the preset interactive behavior on the sample page. The first calculation unit 502 is configured to calculate a first loss value based on the first annotation information corresponding to each content block and the first prediction information corresponding to each content block, wherein the first annotation information corresponding to each content block characterizes whether the sample user performs a preset interactive behavior for the content block when browsing the sample page; The second calculation unit 503 is configured to calculate a second loss value based on the second annotation information and the second prediction information corresponding to each content block, wherein the second annotation information characterizes the true value of the behavioral statistics of the sample user when browsing the sample page; and The adjustment unit 504 is configured to adjust the parameters of the prediction model based on the first loss value and the second loss value.

[0093] According to some embodiments, the prediction model further includes a feature encoding network, the depth of which is not less than the depth of the first prediction network or the depth of the second prediction network. The device 500 further includes an encoding unit configured to input sample user features, sample page features and sample content features into the feature encoding network to obtain the feature encoding result output by the feature encoding network. The output unit is configured to input the feature encoding result into the first prediction network and the second prediction network respectively to obtain first prediction information and second prediction information.

[0094] According to some embodiments, the adjustment unit is configured to: determine the weights corresponding to the first loss value and the second loss value respectively; and adjust the parameters of the prediction model based on the weighted calculation result of the first loss value and the second loss value.

[0095] According to one aspect of this disclosure, a page display device is also provided. Figure 6 A structural block diagram of a page display device 600 according to an exemplary embodiment of the present disclosure is shown. Figure 6 As shown, the device 600 includes: The first determining unit 601 is configured to, for each of a plurality of content blocks associated with a target page, determine first prediction information and second prediction information based on the user characteristics of the target user, the page characteristics of the target page, and the content characteristics of the content block. The first prediction information represents a first probability that the target user will perform a preset interactive behavior on the content block when browsing the target page. The second prediction information represents predicted behavioral statistics of the target user when browsing the target page including the content block, including statistics related to the number of times the target user performs the preset interactive behavior on the target page. The second determining unit 602 is configured to determine a display strategy for the target page based on the first prediction information and the second prediction information.

[0096] According to some embodiments, the first determining unit 601 is configured to: output first prediction information using a first prediction network and output second prediction information using a second prediction network based on user features, page features and content features of the content block, wherein the first prediction network and the second prediction network are trained using sample data labeled with reference output information.

[0097] According to some embodiments, the device 600 further includes an input unit configured to input user features, page features, and content features of the content block into a feature encoding network to obtain feature encoding results, wherein the depth of the feature encoding network is not less than the depth of the first prediction network or the depth of the second prediction network, wherein the first determination unit 601 is configured to input the feature encoding results into the first prediction network and the second prediction network respectively to obtain first prediction information and second prediction information.

[0098] According to some embodiments, the apparatus 600 further includes a normalization processing unit configured to perform normalization processing on user features, page features, and content features of the content block, wherein the input unit is configured to input the normalized user features, the normalized page features, and the normalized content features of the content block into a feature encoding network to obtain feature encoding results.

[0099] According to some embodiments, the second determining unit 602 is configured to determine a display strategy for a plurality of content blocks based on the first prediction information and the second prediction information, wherein the display strategy for each content block includes increasing the display priority of the content block, decreasing the display priority of the content block, or not displaying the content block.

[0100] According to some embodiments, the second determining unit 602 is configured to: determine third prediction information for each content block among a plurality of content blocks by performing a fusion calculation based on first prediction information and second prediction information; sort the plurality of content blocks based on the third prediction information for each content block among the plurality of content blocks; and determine a display strategy for the plurality of content blocks based on the sorting result.

[0101] According to some embodiments, the first determining unit 601 is configured to: determine first prediction information and second prediction information based on user characteristics, page characteristics, content characteristics of the content block and scene characteristics, wherein the scene characteristics are used to characterize the scene information of the target user browsing the target page, and the scene information includes at least one of browsing time information, network environment information or channel information of the target user entering the target page.

[0102] The collection, storage, use, processing, transmission, provision, and disclosure of user personal information involved in the technical solution disclosed herein comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0103] According to one aspect of this disclosure, an electronic device is also provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the training method or page display method of the prediction model applied to page display described above.

[0104] According to one aspect of this disclosure, a non-transitory computer-readable storage medium storing computer instructions is also provided, wherein the computer instructions are used to cause the computer to perform the above-described training method or page display method for a prediction model applied to page display.

[0105] According to one aspect of this disclosure, a computer program product is also provided, comprising a computer program, wherein the computer program, when executed by a processor, implements the above-described training method or page display method for a prediction model applied to page display.

[0106] refer to Figure 7 The present invention describes a structural block diagram of an electronic device 700 that can serve as a server or client of the present disclosure, which is an example of a hardware device that can be applied to various aspects of the present disclosure. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0107] like Figure 7 As shown, device 700 includes a computing unit 701, which can perform various appropriate actions and processes based on a computer program stored in read-only memory (ROM) 702 or a computer program loaded into random access memory (RAM) 703 from storage unit 708. The RAM 703 may also store various programs and data required for the operation of device 700. The computing unit 701, ROM 702, and RAM 703 are interconnected via bus 704. Input / output (I / O) interface 705 is also connected to bus 704.

[0108] Multiple components in device 700 are connected to I / O interface 705, including: input unit 706, output unit 707, storage unit 708, and communication unit 709. Input unit 706 can be any type of device capable of inputting information to device 700. Input unit 706 can receive input numerical or character information and generate key signal inputs related to user settings and / or function control of the electronic device, and may include, but is not limited to, a mouse, keyboard, touchscreen, trackpad, trackball, joystick, microphone, and / or remote control. Output unit 707 can be any type of device capable of presenting information, and may include, but is not limited to, a monitor, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 708 may include, but is not limited to, a hard disk and an optical disk. Communication unit 709 allows device 700 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth devices, 802.11 devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0109] The computing unit 701 can be various general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above, such as training methods for predictive models applied to page display or page display methods. For example, in some embodiments, the training methods for predictive models applied to page display or page display methods can be implemented as computer software programs tangibly contained in a machine-readable medium, such as storage unit 708. In some embodiments, part or all of the computer program can be loaded and / or installed on device 700 via ROM 702 and / or communication unit 709. When the computer program is loaded into RAM 703 and executed by the computing unit 701, one or more steps of the training methods for predictive models applied to page display or page display methods described above can be performed. Alternatively, in other embodiments, the computing unit 701 may be configured by any other suitable means (e.g., by means of firmware) to perform a training method or a page display method for a prediction model applied to a page display.

[0110] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0111] The program code used to implement the methods of this disclosure may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing apparatus, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0112] In the context of this disclosure, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0113] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0114] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), the Internet, and blockchain networks.

[0115] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact via communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other. Servers can be cloud servers, servers in distributed systems, or servers incorporating blockchain technology.

[0116] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired result of the technical solution disclosed in this disclosure can be achieved, and this is not limited herein.

[0117] While embodiments or examples of this disclosure have been described with reference to the accompanying drawings, it should be understood that the methods, systems, and devices described above are merely exemplary embodiments or examples, and the scope of the invention is not limited by these embodiments or examples, but only by the granted claims and their equivalents. Various elements in the embodiments or examples may be omitted or replaced by their equivalents. Furthermore, the steps may be performed in a different order than that described in this disclosure. Further, various elements in the embodiments or examples may be combined in various ways. Importantly, as the technology evolves, many elements described herein can be replaced by equivalents that appear after this disclosure.

Claims

1. A method of training a prediction model applied to page display, wherein, The prediction model includes a first prediction network and a second prediction network, and the method includes: Based on the sample user characteristics of the sample user, the sample page characteristics of the sample page including multiple content blocks, and the sample content characteristics of each content block in the multiple content blocks, the first prediction network outputs first prediction information and the second prediction network outputs second prediction information. The first prediction information represents the first probability that the sample user performs a preset interactive behavior on the content block when browsing the sample page. The second prediction information represents the predicted behavioral statistics of the sample user when browsing the sample page. The behavioral statistics include statistics related to the number of times the sample user performs the preset interactive behavior on the sample page. Based on the first annotation information corresponding to each content block and the first prediction information corresponding to each content block, a first loss value is calculated, wherein the first annotation information corresponding to each content block characterizes whether the sample user performs the preset interactive behavior for the content block when browsing the sample page; Based on the second annotation information and the second prediction information corresponding to each content block, a second loss value is calculated, wherein the second annotation information represents the true value of the behavioral statistics of the sample users when browsing the sample page; and The parameters of the prediction model are adjusted based on the first loss value and the second loss value.

2. The method of claim 1, wherein, The prediction model further includes a feature encoding network, the depth of which is not less than the depth of the first prediction network or the depth of the second prediction network. The method further includes: The sample user features, sample page features, and sample content features are input into the feature encoding network to obtain the feature encoding result output by the feature encoding network. The step of using the first prediction network to output first prediction information and the second prediction network to output second prediction information based on the sample user characteristics of the sample user, the sample page characteristics of the sample page including multiple content blocks, and the sample content characteristics of each content block in the multiple content blocks includes: The feature encoding results are input into the first prediction network and the second prediction network respectively to obtain the first prediction information and the second prediction information.

3. The method of claim 1 or 2, wherein, The adjustment of the parameters of the prediction model based on the first loss value and the second loss value includes: Determine the weights corresponding to the first loss value and the second loss value; and The parameters of the prediction model are adjusted based on the weighted calculation results of the first loss value and the second loss value.

4. A page display method, comprising: For each of a plurality of content blocks associated with a target page, based on the user characteristics of the target user, the page characteristics of the target page, and the content characteristics of the content block, first prediction information and second prediction information are determined. The first prediction information represents a first probability that the target user will perform a preset interactive behavior on the content block when browsing the target page. The second prediction information represents predicted behavioral statistics of the target user when browsing the target page including the content block, including statistics related to the number of times the target user performs the preset interactive behavior on the target page. Based on the first prediction information and the second prediction information, a display strategy for the target page is determined.

5. The method of claim 4, wherein, The determination of the first prediction information and the second prediction information based on the user characteristics of the target user, the page characteristics of the target page, and the content characteristics of the content block includes: Based on the user features, the page features, and the content features of the content block, the first prediction information is output using a first prediction network and the second prediction information is output using a second prediction network. The first prediction network and the second prediction network are trained using sample data labeled with reference output information.

6. The method of claim 5, further comprising: The user features, page features, and content features of the content block are input into a feature encoding network to obtain a feature encoding result, wherein the depth of the feature encoding network is not less than the depth of the first prediction network or the depth of the second prediction network. The step of using a first prediction network to output the first prediction information and a second prediction network to output the second prediction information based on the user features, the page features, and the content features of the content block includes: The feature encoding results are input into the first prediction network and the second prediction network respectively to obtain the first prediction information and the second prediction information.

7. The method of claim 6, further comprising: Normalization processing is performed on the user features, the page features, and the content features of the content block. The step of inputting the user features, the page features, and the content features of the content block into a feature encoding network to obtain the feature encoding result includes: The normalized user features, the normalized page features, and the normalized content features of the content block are input into the feature encoding network to obtain the feature encoding result.

8. The method of any one of claims 4-7, wherein, The step of determining the display strategy for the target page based on the first prediction information and the second prediction information includes: Based on the first prediction information and the second prediction information, a display strategy for the plurality of content blocks is determined, wherein the display strategy for each content block includes increasing the display priority of the content block, decreasing the display priority of the content block, or not displaying the content block.

9. The method of claim 8, wherein, The step of determining the display strategy for the plurality of content blocks based on the first prediction information and the second prediction information includes: A third prediction information for each content block among the plurality of content blocks is determined by performing a fusion calculation based on the first prediction information and the second prediction information. Based on the third prediction information of each of the plurality of content blocks, the plurality of content blocks are sorted; and The display strategy for the multiple content blocks is determined based on the sorting results.

10. The method of any one of claims 4-9, wherein, The second prediction information includes a second probability that the predicted behavioral statistics meet preset conditions, wherein the preset conditions include that the number of times the target user performs the preset interactive behavior on the target page is not less than a number threshold.

11. The method of any one of claims 4-10, wherein, The behavioral statistics also include the total step length or total duration of the target user browsing the target page.

12. The method of claim 11, wherein, When the second prediction information includes the second probability, the preset condition also includes that the total step size information is not less than the step size threshold or the total duration information is not less than the duration threshold.

13. The method of any one of claims 4-12, wherein, The determination of the first prediction information and the second prediction information based on the user characteristics of the target user, the page characteristics of the target page, and the content characteristics of the content block includes: Based on the user characteristics, the page characteristics, the content characteristics of the content block, and the scene characteristics, the first prediction information and the second prediction information are determined. The scene characteristics are used to characterize the scene information of the target user browsing the target page. The scene information includes at least one of browsing time information, network environment information, or the channel information of the target user entering the target page.

14. A prediction model applied to page display, comprising: A first prediction network is used to output first prediction information based on the user characteristics of a target user, the page characteristics of a target page, and the content characteristics of content blocks associated with the target page. The first prediction information represents a first probability that the target user will perform a preset interactive behavior on the content block when browsing the target page. The second prediction network is used to output second prediction information based on the user features, the page features, and the content features. The second prediction information represents the predicted behavioral statistics of the target user when browsing the target page including the content block. The behavioral statistics include statistics related to the number of times the target user performs the preset interactive behavior on the target page.

15. The prediction model of claim 14, further comprising: A feature encoding network is used to output feature encoding results based on the user features, the page features, and the content features. The first prediction network and the second prediction network are used to output the first prediction information and the second prediction information respectively based on the feature encoding results.

16. A device for training a prediction model applied to page display, wherein, The prediction model includes a first prediction network and a second prediction network, and the device includes: The output unit is configured to output first prediction information using the first prediction network and second prediction information using the second prediction network, based on the sample user characteristics of the sample user, the sample page characteristics of the sample page including multiple content blocks, and the sample content characteristics of each content block in the multiple content blocks. The first prediction information represents the first probability that the sample user performs a preset interactive behavior on the content block when browsing the sample page. The second prediction information represents the predicted behavioral statistics of the sample user when browsing the sample page. The behavioral statistics include statistics related to the number of times the sample user performs the preset interactive behavior on the sample page. The first calculation unit is configured to calculate a first loss value based on the first annotation information corresponding to each content block and the first prediction information corresponding to each content block, wherein the first annotation information corresponding to each content block characterizes whether the sample user performs the preset interactive behavior for the content block when browsing the sample page; The second calculation unit is configured to calculate a second loss value based on the second annotation information and the second prediction information corresponding to each content block, wherein the second annotation information represents the true value of the behavioral statistics of the sample user when browsing the sample page; and The adjustment unit is configured to adjust the parameters of the prediction model based on the first loss value and the second loss value.

17. A page display device, comprising: A first determining unit is configured to, for each of a plurality of content blocks associated with a target page, determine first prediction information and second prediction information based on user characteristics of a target user, page characteristics of the target page, and content characteristics of the content block. The first prediction information represents a first probability that the target user will perform a preset interactive behavior on the content block when browsing the target page. The second prediction information represents predicted behavioral statistics of the target user when browsing the target page including the content block, the behavioral statistics including statistics related to the number of times the target user performs the preset interactive behavior on the target page. The second determining unit is configured to determine a display strategy for the target page based on the first prediction information and the second prediction information.

18. An electronic device, comprising: At least one processor; as well as A memory that is communicatively connected to the at least one processor; in The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-13.

19. A non-transitory computer readable storage medium having stored thereon computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-13.

20. A computer program product comprising a computer program, wherein, When the computer program is executed by a processor, it implements the method according to any one of claims 1-13.

Images