Information pushing method and device, equipment and storage medium

Abstract

The embodiment of the invention relates to an information pushing method and device, equipment and a storage medium. The method provided by the invention comprises the following steps: determining a group of push information aiming at a target user and to be pushed; at least based on first information, the recommendation degree of each piece of push information in the set of push information for the target user is determined, and the first information at least indicates historical interaction information related to the target user for the set of push information; determining target push information from the group of push information based on the recommendation degree of each piece of push information in the group of push information; and pushing the target push information to the target user.

Description

Technical Field

[0001] Example embodiments of the present disclosure generally relate to the field of computer technology, and more specifically, to a method, apparatus, device, and storage medium for information push. Background Art

[0002] Information push is a technology that pushes relevant content to users based on user information. This technology builds user profiles by analyzing data such as user browsing history and search history, and then predicts content that users may be interested in and pushes it to users in real time. Summary of the invention

[0003] In a first aspect of the present disclosure, a method for information push is provided, comprising: determining a group of push information for a target user and to be pushed; determining, based at least on first information, a recommendation level of each push information in the group of push information for the target user, wherein the first information at least indicates historical interaction information related to the target user for the group of push information; determining target push information from the group of push information based on the recommendation level of each push information in the group of push information; and pushing the target push information to the target user.

[0004] In a second aspect of the present disclosure, a device for information push is provided, including: a push information determination module, configured to determine a group of push information for a target user and to be pushed; a recommendation degree determination module, configured to determine the recommendation degree of each push information in a group of push information for the target user based on at least first information, wherein the first information at least indicates historical interaction information related to the target user for the group of push information; a target push information determination module, configured to determine target push information from a group of push information based on the recommendation degree of each push information in the group of push information; and a push module, configured to push the target push information to the target user.

[0005] In a third aspect of the present disclosure, an electronic device is provided. The device includes at least one processing unit; and at least one memory, the at least one memory is coupled to the at least one processing unit and stores instructions for execution by the at least one processing unit. When the instructions are executed by the at least one processing unit, the device executes the method of the first aspect.

[0006] In a fourth aspect of the present disclosure, a computer-readable storage medium is provided, wherein a computer program is stored on the computer-readable storage medium, and the computer program can be executed by a processor to implement the method of the first aspect.

[0007] In a fifth aspect of the present disclosure, a computer program product is provided, which includes computer executable instructions, and when the instructions are executed by a processor, the method according to the first aspect of the present disclosure is implemented.

[0008] It should be understood that the contents described in this content section are not intended to limit the key features or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0009] The above and other features, advantages and aspects of the embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. In the accompanying drawings, the same or similar reference numerals represent the same or similar elements, wherein:

[0010] Figure 1 A schematic diagram showing an example environment in which embodiments of the present disclosure can be implemented;

[0011] Figure 2 A flowchart showing an example process of a method for information push according to some embodiments of the present disclosure;

[0012] Figure 3 A schematic diagram showing an example of determining the recommendation degree of push information according to some embodiments of the present disclosure;

[0013] Figure 4 A schematic diagram showing an example of adjusting the recommendation degree according to an embodiment of the present disclosure;

[0014] Figure 5 A schematic diagram showing an example of adjusting the recommendation degree according to an embodiment of the present disclosure;

[0015] Figure 6 A schematic structural block diagram of an apparatus for information push according to some embodiments of the present disclosure is shown; and

[0016] Figure 7 A block diagram of an electronic device is shown in which one or more embodiments of the present disclosure may be implemented. DETAILED DESCRIPTION

[0017] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments set forth herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are only for exemplary purposes and are not intended to limit the scope of protection of the present disclosure.

[0018] It should be noted that the titles of any sections / subsections provided herein are not restrictive. Various embodiments are described throughout this article, and any type of embodiment may be included under any section / subsection. In addition, the embodiments described in any section / subsection may be combined in any manner with any other embodiments described in the same section / subsection and / or different sections / subsections.

[0019] In the description of the embodiments of the present disclosure, the term "including" and similar terms should be understood as open inclusion, that is, "including but not limited to". The term "based on" should be understood as "based at least in part on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The term "some embodiments" should be understood as "at least some embodiments". Other explicit and implicit definitions may be included below. The terms "first", "second", etc. may refer to different or the same objects. Other explicit and implicit definitions may be included below.

[0020] The embodiments of the present disclosure may involve user data, data acquisition and / or use, etc. These aspects are subject to the corresponding laws, regulations and relevant provisions. In the embodiments of the present disclosure, all data collection, acquisition, processing, processing, forwarding, use, etc. are carried out on the premise that the user knows and confirms. Accordingly, when implementing each embodiment of the present disclosure, the type, scope of use, usage scenario, etc. of the data or information that may be involved should be informed to the user and the user's authorization should be obtained in an appropriate manner in accordance with the relevant laws and regulations. The specific notification and / or authorization method can vary according to the actual situation and application scenario, and the scope of the present disclosure is not limited in this respect.

[0021] In this specification and the embodiments, if personal information processing is involved, it will be processed on the premise of having a legal basis (such as obtaining the consent of the subject of personal information, or it is necessary to perform a contract, etc.), and will only be processed within the scope of regulations or agreements. If a user refuses to process personal information other than the necessary information for basic functions, it will not affect the user's use of basic functions.

[0022] As briefly described above, information push technology aims to predict and push content that users may be interested in in real time. In this process, one information push strategy is to combine two types of feature data - user information and push result information, to guide the screening of push information. For example, user information can include the user's gender and age, which can help portray the user's basic portrait; while push result information can include historical click-through rate and conversion rate, which reflects the attractiveness of the content.

[0023] However, there is a significant problem with the above information push strategy: it fails to fully explore the cross-relationship between user information and push result information. In other words, this information push strategy fails to fully consider the complex relationship between user personal characteristics and push information. This limitation limits the accuracy of information push, and the results of information push may not fully meet the personalized needs and interests of users.

[0024] In view of this, an embodiment of the present disclosure provides a scheme for information push. According to the scheme, a group of push information to be pushed for a target user is first determined. Then, based at least on first information, the recommendation level of each push information in the group of push information for the target user is determined, wherein the first information at least indicates historical interaction information related to the group of push information of the target user. Next, based on the recommendation level of each push information in the group of push information, target push information is determined from the group of push information. Finally, the target push information is pushed to the target user.

[0025] It will be more clearly understood through the description below that the scheme of the present disclosure can more accurately capture the interests and preferences of users by considering the historical interaction information between users and push information. Compared with traditional information push strategies, this scheme pays more attention to the personalized needs of users. By deeply analyzing the interaction data between users and push information (such as interaction activity), this scheme can explore the cross-relationship between user information and push result information, thereby improving the accuracy of information push. Accurate information push means that users are more likely to be interested in the pushed information and take further actions, such as clicking, purchasing or sharing. This will directly improve the conversion rate of push information, thereby improving the effectiveness of information push.

[0026] In summary, the information push solution provided by the embodiments of the present disclosure achieves more accurate information push by integrating and analyzing historical interaction information between users and content, thereby significantly improving user experience and information conversion rate.

[0027] Various example implementations of the solution will be described in detail below in conjunction with the accompanying drawings.

[0028] Figure 1 A schematic diagram of an example environment 100 is shown in which embodiments of the present disclosure can be implemented.

[0029] Reference Figure 1 , the electronic device 120 may manage one or more push information related to one or more objects. In the present disclosure, an object may include various recommendable objects, examples of which may include applications, physical goods / services, virtual goods / services, digital content / physical content, etc. "Push information" refers to content presented to recommend a corresponding object. Examples of push information may include advertisements.

[0030] One or more client devices 110-1, 110-2, ..., 110-3, etc. (for ease of discussion, collectively or individually referred to as client device 110) are associated with electronic device 120. Electronic device 120 can present push information to users 140-1, 140-2, ..., 140-3, etc. (for ease of discussion, collectively or individually referred to as user 140) with the help of client device 110. User 140 can view the corresponding object by clicking on the push information, and perform subsequent operations (such as purchase, like, collection, etc.) based on information related to the object. As an example, applications, websites, web pages, and other accessible platforms can be set in electronic device 120. Client device 110 can be installed with an application for accessing electronic device 120, or client device 110 can also access electronic device 120 in other suitable ways.

[0031] The electronic device 120 may be configured to provide a set of push information related to a set of objects to the user 140 based on a corresponding policy (e.g., provided or presented on the client device 110). A set of push information may, for example, include one or more push information 130-1, 130-2, ... 130-3 (for ease of discussion, collectively or individually referred to as push information 130) in the push information database 131.

[0032] In the present disclosure, user 140 may be any potential consumer that can be served, such as an individual, a group, an organization, or any other entity.

[0033] In some embodiments, the electronic device 120 may provide corresponding push information 122 to the user 140 based on the request of the push requester 150-1, 150-2, ..., 150-3, etc. (collectively or individually referred to as "push requester" 152). As an example, in the scenario of advertising delivery, the push requester is sometimes also referred to as an advertiser. In the scenario of business push, the push requester is sometimes also referred to as a business line.

[0034] In some embodiments, the client device 110 can be any type of mobile terminal, fixed terminal or portable terminal, including mobile phones, desktop computers, laptop computers, notebook computers, netbook computers, tablet computers, media computers, multimedia tablets, personal communication systems (PCS) devices, personal navigation devices, personal digital assistants (PDAs), audio / video players, digital cameras / camcorders, positioning devices, television receivers, radio broadcast receivers, e-book devices, gaming devices, or any combination of the foregoing, including accessories and peripherals of these devices or any combination thereof. In some embodiments, the client device 110 can also support any type of interface for the user (such as "wearable" circuits, etc.).

[0035] The electronic device 120 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content distribution networks, and big data and artificial intelligence platforms. The electronic device 120 may include, for example, a computing system / server, such as a mainframe, an edge computing node, a computing device in a cloud environment, and the like. The electronic device 120 may provide background services for applications in the client device 110 that support content presentation.

[0036] A communication connection may be established between the electronic device 120 and the client device 110. The communication connection may be established in a wired manner or a wireless manner. The communication connection may include, but is not limited to, a Bluetooth connection, a mobile network connection, a universal serial bus connection, a wireless fidelity connection, etc., and the embodiments of the present disclosure are not limited in this respect. In the embodiments of the present disclosure, the electronic device 120 and the client device 110 may implement signaling interaction through the communication connection between the two.

[0037] It should be understood that the structure and function of the various elements in the environment 100 are described for exemplary purposes only and do not imply any limitation on the scope of the present disclosure.

[0038] Figure 2 A flowchart showing an example process 200 of a method for information push according to some embodiments of the present disclosure is shown. Figure 3 A schematic diagram of an example 300 of determining a recommendation level 330 of push information according to some embodiments of the present disclosure is shown. The process 200 may be implemented at the electronic device 120, as shown below in conjunction with Figure 3 The process 200 is described.

[0039] In block 210 , the electronic device 120 determines a set of push information for a target user (eg, the user 140 ) and to be pushed.

[0040] In some embodiments, process 200 may be applied to products or scenarios related to traffic distribution. Traffic distribution involves how to effectively push candidate information (such as articles, videos, advertisements, etc.) to interested users.

[0041] As an example, products or scenarios related to traffic distribution may involve search engines, web pages, applications (APPs), and mini-programs, etc. For example, in search engines, traffic distribution aims to ensure that users can see the most relevant and valuable results when searching. On web pages, traffic distribution aims to help website owners display important content or advertisements to users who are most likely to be interested. For APPs and mini-programs, traffic distribution aims to ensure that users can access the content and functions that best meet their needs when using these applications.

[0042] As an example, the information push method of the embodiment of the present disclosure can be applied to the traffic distribution scenario of the APP, that is, the information push method of the embodiment of the present disclosure can be used to instruct the APP to provide users with content that is most likely to attract them (that is, the target push information mentioned below).

[0043] In block 220, the electronic device 120 determines a recommendation level 330 of each push information 130 in a set of push information for a target user based at least on the first information. The first information at least indicates historical interaction information 301 related to the target user for a set of push information.

[0044] In some embodiments, the historical interaction information 301 may be derived from prior knowledge related to a set of push information about the target user. As an example, the prior knowledge may be information or data about the target user determined based on expert experience. Such information may include, but is not limited to, the user's personal preferences, risk ratings, consumption habits, search history, purchase records, etc., which are usually collected through the user's previous online behavior, questionnaires, registration information, etc.

[0045] In some embodiments, as an example, the historical interaction information 301 may be determined based on the historical interaction behavior between the user and a certain push information 130. The historical interaction behavior may include, for example, whether the target user has clicked on the push information 130, and whether the target user has purchased and commented on the product or content (such as shopping, video watching, music listening, etc.) corresponding to the push information 130 after clicking on the push information 130.

[0046] In some embodiments, the historical interaction information 301 may also reflect the degree of interaction activity between the user and a certain push information 130. The degree of interaction activity may indicate the frequency and depth of interaction of the target user with a certain push information 130. For example, the number of times the target user has watched, liked, or commented on a certain video reflects the target user's high interest in and active interaction with the information. This degree of interaction activity provides the electronic device 120 with instant feedback from the target user on certain content.

[0047] In some embodiments, by analyzing the historical interaction information 301, the electronic device 120 can evaluate the target user's interest and preference for at least one push information 130 in a group of push information. Such analysis may include whether the target user's preference for a product or content corresponding to a certain push information 130 has increased within a specific time period, or whether the target user's interaction with a product or content corresponding to a certain push information 130 has become more frequent.

[0048] Next, based on the analysis results of the historical interaction information 301 , the electronic device 120 may calculate a recommendation level 330 for each push information 130 . The recommendation level 330 reflects the interest level of the target user in a certain push information 130 .

[0049] Then, through blocks 230 and 240, the electronic device 120 can filter a group of push information according to the recommendation level 330 of each push information 130. In this way, when the target user visits an APP, website or other platform, the target user first sees the content that best matches his or her interests and needs (i.e., the target push information).

[0050] In some embodiments, the electronic device 120 determines the recommendation level 330 of each push information 130 in a group of push information based on the first information and the second information.

[0051] In some embodiments, when determining the recommendation level 330 of each push information 130 in a group of push information, the electronic device 120 may consider the second information in addition to the historical interaction information 301. The second information may provide more comprehensive reference information for the electronic device 120, thereby enhancing the accuracy and personalization of the information push.

[0052] In some embodiments, the second information indicates at least one of the following: push activation information 302 related to the target user for at least one push information 130 in a group of push information; historical push result information 303 related to a group of push information, the historical push result information 303 at least indicating the historical click-through rate and / or historical conversion rate related to a group of push information; or user information 304 related to the target user.

[0053] As an example, the push activation information 302 may be derived from prior knowledge about a set of push information related to the target user. The push activation information 302 may indicate whether the target user has activated (e.g., completed the initial setup or registration process) the product or content corresponding to the push information 130 and has begun to formally use the function of the product or content.

[0054] As an example, if a target user downloads or installs a product (such as an application, software, or service) for the first time, they usually need to complete some initial settings, such as creating an account, agreeing to the terms of service, and providing necessary personal information. Once the user completes these steps and starts using the core functions of the product, their push activation information 302 can be considered "activated." If a user downloads the product but does not complete the initial settings or registration process, or creates an account but never uses the core functions of the product, their push activation information 302 can be considered "unactivated."

[0055] As an example, the historical push result information 303 may indicate a set of characteristics of the push information itself. For example, the historical click-through rate associated with a certain push information 130 may be the ratio of the number of times the push information 130 has been clicked to the number of times it has been exposed (or displayed to users) in the past month. In this way, the historical click-through rate may reflect the degree of acceptance of the push information 130 by the user group over the past period of time. For example, the historical conversion rate associated with a certain push information 130 may be the ratio of the number of times the push information 130 has been converted to the number of times it has been clicked in the past month. It should be noted that "converted" here may refer to the user's purchase, use or download of the product or content corresponding to the push information 130. In this way, the historical click-through rate may reflect the proportion of these push information 130 that has been converted into actual actions.

[0056] As an example, the user information 304 may indicate the characteristics of the target user itself. For example, the user information 304 related to the target user may indicate the target user's age, gender, geographic location, occupation, education level, etc. These characteristics may provide support for the electronic device 120 to construct a user portrait.

[0057] In some embodiments, the electronic device 120 can combine the first information and the second information to calculate a comprehensive recommendation level 330 for each push information 130. In addition to considering the personalized historical interaction between the target user and the push information 130, the recommendation level 330 also considers the activation state between the target user and the push information 130, the target user's own characteristics, and the push information 130's own characteristics. In this way, the electronic device 120 can fully explore the complex connections between these information. For example, after browsing a certain type of content, user A may be more inclined to click on another type of content that seems unrelated but actually has some kind of internal connection. This connection may not be obvious, but hidden in a large number of features.

[0058] In some embodiments, according to actual needs, the second information can also indicate more content. For example, the second information can also indicate the participation status of the target user for at least one push information 130. The participation status may include, but is not limited to, indicating whether the target user has opened high-value service content such as membership services on the product or content corresponding to the push information 130. For another example, the second information can also indicate the user's personal preferences, risk ratings, consumption habits, search history, purchase records, etc. The above-mentioned various contents can be derived from the prior knowledge described above. In other words, the embodiments of the present disclosure can determine the recommendation level 330 of the push information 130 in combination with any suitable content in the prior knowledge. In this way, the second information can fully reflect the personalized instructions related to the target user, and based on such second information, it is beneficial for the machine learning model mentioned below to use personalized knowledge to offset the domain bias that may arise in the learning process.

[0059] In addition, with the help of the first information and the second information, the electronic device 120 can also explore the potential correlation between the target user's behavior performance on different push information 130. For example, users who frequently click on push information 130-1 often also show a high click interest in push information 130-2. When evaluating the recommendation level 330 of push information 130, the electronic device 120 will take into account the correlation between user behaviors between such different push information 130, thereby achieving accurate push across products or content. This method not only meets the diverse needs of users, but also further optimizes the user experience and improves user satisfaction.

[0060] In some embodiments, the electronic device 120 may take the first information and the second information as input, and then use a trained machine learning model to determine the recommendation level 330 of the pushed information 130 .

[0061] In some embodiments, in order to enable the machine learning model to more accurately determine the recommendation level 330 of the push information 130, the embodiments of the present disclosure optimize the samples used by the machine learning model during the training process.

[0062] Specifically, when training a machine learning model, the initial sample set may include a group of push information samples, and historical interaction information samples, push activation information samples, historical push result information samples, and historical push result information samples related to this group of push information samples. Based on the description below, it can be understood that the embodiments of the present disclosure enable the machine learning model to have multiple goals, so each sample in a group of push information samples can be classified into exposure samples (i.e., push information 130 samples displayed to users), click samples (i.e., push information 130 samples displayed to users and clicked by users), and conversion samples (i.e., push information 130 samples clicked by users and executed such as purchase behaviors).

[0063] There is a strict temporal relationship between exposure samples, click samples, and conversion samples (for example, click occurs after exposure, and conversion occurs after click). Therefore, in a set of push information samples, the embodiment of the present disclosure selects all click samples as positive samples, which naturally include all conversion samples. In addition, the embodiment of the present disclosure performs negative sampling on the part of the training sample set except the click samples to obtain negative samples. Then, the positive and negative samples are fused to form a complete training sample set.

[0064] As an example, for a training sample set, the embodiment of the present disclosure can divide it into sparse features according to the properties of the features. sparse ) and dense features dense ). Sparse features usually refer to those features whose values ​​are mostly zero and only a few non-zero values, while dense features refer to those features whose values ​​are relatively evenly distributed.

[0065] For sparse features in the training sample set, the embodiment of the present disclosure first performs data cleaning, removes abnormal data and fills in the default value of each feature. Then, these sparse features are converted into dense features using encoders and neural networks to improve the expression ability of information.

[0066] For dense features in the training sample set, the embodiment of the present disclosure first performs data cleaning and default value filling. Then, the mean and standard deviation of these dense features are calculated. In order to avoid dimensionality problems, the embodiment of the present disclosure can standardize the dense features through Formula 1:

[0067]

[0068] Afterwards, the embodiment of the present disclosure inputs the processed training samples into the machine learning model to train the machine learning model, and enables the trained machine learning model to determine the recommendation degree 330 of the push information 130 based on the historical interaction information 301, the push activation information 302, the historical push result information 303 and the historical push result information 303.

[0069] In some embodiments, the electronic device 120 constructs a first set of feature representations 311 related to the first recommendation evaluation task 310 for candidate push information in a group of push information based on the historical push result information 303 and the user information 304. And the electronic device 120 constructs a second set of feature representations 321 related to the second recommendation evaluation task 320 for the candidate push information based on the historical push result information 303, the user information 304, the push activation information 302, and the historical interaction information 301. Then, the electronic device 120 determines the recommendation degree 330 of the candidate push information by executing the first recommendation evaluation task 310 and the second recommendation evaluation task 320 based on the first set of feature representations 311 and the second set of feature representations 321.

[0070] In this way, the electronic device 120 achieves a more accurate and personalized push information recommendation evaluation through the dual recommendation evaluation tasks, thereby significantly improving the accuracy of information push and user satisfaction.

[0071] In some embodiments, the target group feature representation in the first group feature representation 311 and the second group feature representation 321 is determined in the following manner: First, in the first information and the second information, the electronic device 120 extracts features related to the target group feature representation to obtain at least one first feature. Then, the electronic device 120 fills the missing values ​​of each first feature in the at least one first feature with a predetermined value to obtain at least one second feature. Then, the electronic device 120 encodes the at least one second feature to determine the target group feature representation.

[0072] As an example, for the first set of feature representations, the electronic device 120 can extract the first feature from the historical push result information 303 and the historical push result information 303. For the second set of feature representations, the electronic device 120 can extract the first feature from the historical interaction information 301, the push activation information 302, the historical push result information 303 and the historical push result information 303.

[0073] As an example, the predetermined value may include but is not limited to 0, mean value, median, mode or any appropriate reasonable value, etc. Through this process, the electronic device 120 fills the missing value of the first feature with the predetermined value, thereby ensuring the integrity of the feature data.

[0074] As an example, in the first recommendation evaluation task 310, the electronic device 120 may use the encoder of the machine learning model to encode the second features obtained based on the historical push result information 303 and the historical push result information 303 to obtain a first set of feature representations 311. In the second recommendation evaluation task 320, the electronic device 120 may use the encoder of the machine learning model to encode the second features obtained based on the historical interaction information 301, the push activation information 302, the historical push result information 303, and the historical push result information 303 to obtain a second set of feature representations 321.

[0075] In some embodiments, the electronic device 120 converts the sparse feature representation into a dense feature representation by compressing the sparse feature representation in response to determining that there is a sparse feature representation corresponding to the sparse feature in the first group of feature representations 311 and the second group of feature representations 321. Then, the electronic device 120 determines the recommendation degree of the candidate push information by executing the first recommendation degree evaluation task 310 and the second recommendation degree evaluation task 320 based at least on the dense feature representation.

[0076] As an example, the identification of sparse feature representation can be achieved by counting the proportion of non-zero or valid values ​​in the feature vector. For example, if the sparsity of a feature vector exceeds a preset threshold, it is regarded as a sparse feature representation.

[0077] For the identified sparse feature representation, the electronic device 120 converts it into a dense feature representation through operations such as compression, and the conversion process includes but is not limited to: selecting the most predictive features from the sparse features, removing redundant or irrelevant features; mapping high-dimensional sparse features to low-dimensional space while retaining the original information as much as possible; and converting sparse features into a more compact representation through a specific algorithm. In this way, the sparsity of the first set of feature representations 311 and the second set of feature representations 321 can be reduced, and the density and information content of the features can be increased, thereby providing more high-quality and efficient data input for subsequent machine learning models.

[0078] As an example, in the first recommendation evaluation task 310, the electronic device 120 can use the field-wise pooling layer (Field-wise Pooling Layer) of the machine learning model to summarize or fuse the features in the first set of feature representations 311 to obtain a feature representation 312-1 related to the historical push result information 303 of the user information 304 and a feature representation 312-2 related to the user information 304. Subsequently, the electronic device 120 can use the machine learning model to splice the feature representation 312-1 and the feature representation 312-2 to obtain the feature representation 313. Then, the electronic device 120 can use the multi-layer perceptron (MultiLayer Perception) 314 in the machine learning model to identify the spliced ​​features, thereby obtaining the output of the first recommendation evaluation task 310, such as the predicted conversion probability 315 mentioned below.

[0079] As an example, in the second recommendation evaluation task 320, the electronic device 120 can use the field-level pooling layer of the machine learning model to summarize or fuse the features in the second set of feature representations 311 to obtain feature representations 322-1 related to historical interaction information 301 and push activation information, feature representations 322-2 related to historical push result information, and feature representations 322-3 related to user information 304. Subsequently, the electronic device 120 can use the machine learning model to splice the feature representations 322-1, 322-2, and 322-3 to obtain feature representations 323. Subsequently, the electronic device 120 can use the multi-layer perceptron 324 in the machine learning model to identify the spliced ​​features, thereby obtaining the output of the second recommendation evaluation task 320, such as the predicted click probability 325 mentioned below.

[0080] In some embodiments, the electronic device 120 determines the predicted conversion probability 315 of the candidate push information for the target user by performing the first recommendation evaluation task 310 based on the first set of feature representations 311. Furthermore, the electronic device 120 determines the predicted click probability 325 of the candidate push information for the target user by performing the second recommendation evaluation task 320 based on the second set of feature representations 321. Then, the electronic device 120 determines the recommendation degree 330 of the candidate push information based at least on the predicted conversion probability 315 and the predicted click probability 325.

[0081] In some embodiments, the predicted click probability 325 may refer to the predicted probability of the push information 130 being clicked by the user (Click-through rate, CTR). The predicted click probability 325 may reflect the attractiveness of the push information 130 to the target user. A higher predicted click probability 325 means that the push information 130 can successfully attract the user's attention and stimulate their interest, thereby increasing the user's interaction opportunities with the push information 130.

[0082] In some embodiments, the predicted conversion rate 315 (CVR) at least indicates the predicted probability that the target user will perform a target operation based on the content provided by the push information 130 after clicking on at least one push information 130. As an example, "click" here may refer to the user clicking, browsing, or otherwise accepting the push information 130. As an example, the predicted conversion probability 315 reflects an indicator of the possibility of conversion of the target user from the initial contact with the push information 130 to the final further action (such as purchasing a product, registering an account, filling out a questionnaire, etc.).

[0083] In some embodiments, the electronic device 120 uses a trained machine learning model to respectively perform the first recommendation evaluation task 310 and the second recommendation evaluation task 320 to obtain a predicted conversion probability 315 and a predicted click probability 325 .

[0084] In some embodiments, at least some of the feature representations in the first set of feature representations 311 and the second set of feature representations are shared. In this way, the first recommendation evaluation task 310 can learn from various information contained in the second set of feature representations 321, thereby alleviating the problem of data sparsity.

[0085] After obtaining the predicted conversion probability 315 and the predicted click probability 325 , the electronic device 120 may perform weighted calculation on the two indicators to determine the recommendation degree 330 of the candidate push information.

[0086] In the above process, the electronic device 120 uses a machine learning model to comprehensively consider the historical interaction information 301, the push activation information 302, the historical push result information 303, the user information 304 and the complex relationship between them, and calculates the recommendation degree 330 of the push information 130 based on the two indicators of the predicted conversion probability 315 and the predicted click probability 325. In this way, it can ensure that the push information 130 not only fits the personal preferences of the target user, but also has a high conversion potential, thereby providing the user with a more accurate and personalized push experience.

[0087] In some embodiments, the electronic device 120 obtains the predicted revenue information 340 of the candidate push information for the target user. Then, the electronic device 120 determines the recommendation degree 330 of the candidate push information based on the predicted conversion probability 315 , the predicted click probability 325 and the predicted revenue information 340 .

[0088] In some embodiments, the predicted revenue information 340 at least indicates the target user's lifetime value (LTV) on the product related to the candidate push information. As an example, the lifetime value indicates all the value that the target user may generate during the use of the product or service.

[0089] In some embodiments, the predicted revenue information 340 may be pre-configured based on expert experience. In this way, for each push information 130 in a group of push information, the electronic device 120 can directly find the predicted revenue information 340 corresponding to each push information 130 through a search operation, thereby eliminating the calculation process of the revenue information 340 and improving processing efficiency.

[0090] In this way, when determining the recommendation degree 330 of the candidate push information, the electronic device 120 also introduces the predicted revenue information 340 in addition to the predicted conversion probability 315 and the predicted click probability 325. The predicted revenue information 340 allows the determination process of the recommendation degree 330 to additionally consider the value of the push information 130, which is conducive to preferentially pushing push information 130 with higher value to the target user, thereby improving the return rate of the information push service.

[0091] In some embodiments, during the training of the machine learning model, for a group of push information samples of the same user, the machine learning model can sort these push information 130 samples based on the size of LTV. Next, the electronic device 120 can set a larger LTV weight (for example, "1") for the push information 130 samples with a higher LTV ranking (for example, the top 30%) in the group of push information samples, and set a smaller LTV weight (for example, "0") for the push information 130 samples with a lower ranking (for example, the bottom 70%) in the group of push information samples. In this way, during the training process, the machine learning model can pay more attention to the push information 130 samples with a larger LTV, thereby improving the training effect.

[0092] In some embodiments, the machine learning model technology determines the predicted penetration conversion rate based on the predicted conversion rate 315 and the predicted click rate 325, and then determines the recommendation level 330 based on the predicted penetration conversion rate. As an example, the penetration conversion rate may refer to the ratio of the probability of the push information 130 being converted to the probability of being exposed.

[0093] Specifically, the machine learning model can determine the predicted penetration conversion rate pCTCVR through Formula 2 and Formula 3:

[0094] p(click=1,conversion=1|x)=p(click=1|x)×p(conversion=

[0095] 1|click=1,x);

[0096] Formula 2 can be simplified to Formula 3:

[0097] pCTCVR=pCTR×pCVR;

[0098] Formula 3

[0099] Wherein p(click=1|x) represents the probability of the target user clicking (click) given feature x, p(conversion=1|click=1,x) represents the probability of the target user converting (conversion) given that the target user has already clicked (click) and feature x, pCTR represents the output of the second recommendation evaluation task 320 (e.g., predicted click probability 325), and pCVR represents the output of the first recommendation evaluation task 310 (e.g., predicted conversion probability 315).

[0100] The calculation process of pCTR and pCVR is described below in combination with Formula 4 to Formula 7.

[0101] As an example, the electronic device 120 first determines the input CTR_Emb (eg, feature representation 323 ) of the multilayer perceptron 324 in the second recommendation evaluation task 320 through Formula 4.

[0102] CTR_Emb=Concat(Feature user ,Feature item ,Feature cross );

[0103] Formula 4

[0104] Among them, Feature user represents the feature representation related to the user information 304 determined in the second recommendation evaluation task 320 (eg, feature representation 322-3), Feature item Indicates the feature representation related to the historical push result information 303 determined in the second recommendation evaluation task 320 (for example, feature representation 322-2), Feature crossIt represents the feature representation related to the historical interaction information 301 and the push activation information 302 determined in the second recommendation degree evaluation task 320 (eg, the feature representation 322 - 1 ).

[0105] As an example, the electronic device 120 may also determine the input CVR_Emb (eg, the feature representation 313 ) of the multilayer perceptron 324 in the first recommendation evaluation task 310 by using Formula 5.

[0106] CVR_Emb=Concat(Feature user ′,Feature item ′,Feature cross ′);

[0107] Formula 5

[0108] Among them, Feature user ' represents the feature representation related to the user information 304 determined in the first recommendation evaluation task 310 (eg, feature representation 312-2), Feature item ′ represents the feature representation related to the historical push result information 303 determined in the first recommendation evaluation task 310 (for example, feature representation 312-1), Feature cross ′ represents the feature representation related to the historical interaction information 301 and the push activation information 302 determined in the first recommendation evaluation task 310. It should be noted that, as described above, the second set of feature representations 321 can be shared with the first set of feature representations 311. Therefore, Figure 3 Although not shown in the figure, in fact, the first recommendation evaluation task 310 can also extract features from the second set of feature representations 321 to obtain Feature cross ′.

[0109] In some embodiments, the output DNN (CTR_Emb) of the second recommendation evaluation task 320 may be processed by an activation function (eg, a sigmoid function) to obtain pCTR, which may be specifically expressed by Formula 6.

[0110] pCTR=sigmoid(DNN(CTR_Emb));

[0111] Formula 6

[0112] In some embodiments, the output DNN (DNN(CVR_Emb) of the first recommendation evaluation task 310 may be processed by an activation function (eg, a sigmoid function) to obtain pCVR. Specifically, it may be expressed by Formula 7.

[0113] pCVR=sigmoid(DNN(CVR_Emb));

[0114] Formula 7

[0115] After the pCTR and pCVR are determined, the pCTCVR can be determined based on Formula 3, and then the pCTCVR is combined with the corresponding LTV to obtain the recommendation level 330 of the push information 130.

[0116] In this way, the electronic device 120 comprehensively and additionally considers maximizing the value of the push information 130. When the predicted click probability and the predicted conversion probability are the same, giving priority to pushing to the user can generate higher quality push information 130, thereby improving the quality of information push.

[0117] In some embodiments, a focal loss function may be used to train the machine learning model.

[0118] As an example, the loss function Focal_Loss related to pCTR ctr The loss function Focal_Loss related to pCTCVR can be calculated by formula 8. ctcvr It can be calculated by formula 9.

[0119] Focal_Loss ctr =-(1-p t_ctr ) γ log(p t_ctr );

[0120] Formula 8

[0121] Focal_Loss ctcvr =-(1-p t_ctcvr ) γ log(p t_ctcvr );

[0122] Formula 9

[0123] Among them, γ is an adjustable focusing parameter, p t_ctr It can be calculated by formula 10, P t_ctcvr It can be calculated by formula 11.

[0124]

[0125] In this way, on the one hand, the problem of data imbalance can be alleviated, and on the other hand, the prediction accuracy of difficult-to-classify samples can be improved.

[0126] In some embodiments, the electronic device 120 may use a distributed computing method to determine the recommendation level 330 of each push information 130 in a group of push information. In this way, the calculation speed of the recommendation level 330 may be increased, and the user experience may be improved.

[0127] In block 230 , the electronic device 120 determines target push information from a group of push information based on the recommendation level 330 of each push information 130 in the group of push information.

[0128] In some embodiments, the electronic device 120 may select one or more push information from a group of push information as target push information based on the recommendation levels 330 of the respective push information 130 .

[0129] In some embodiments, for the selected one or more target push information, the electronic device 120 may present the target push information to the user through block 240. However, in actual scenarios, it is necessary to ensure that the push traffic distribution of each target push information is stable and maintains a certain level.

[0130] Based on this, in some embodiments, the electronic device 120 adjusts the recommendation degree 330 of each push information 130 in a group of push information based on the flow control information of each push information 130 in a group of push information, wherein the flow control information indicates the maximum push flow of each push information 130 in a group of push information within a predetermined period. Then, the electronic device 120 determines the target push information from the group of push information based on the adjusted recommendation degree 330 of each push information 130 in the group of push information.

[0131] Figure 4 A schematic diagram illustrating an example 400 of adjusting the recommendation level 330 according to an embodiment of the present disclosure is shown.

[0132] Reference Figure 4 In box 410, the electronic device 120 uses a machine learning model to determine the recommendation level 330 of each push information 130 in a set of push information. In box 420, the electronic device 120 obtains the flow control information of each push information 130. As an example, in addition to indicating the maximum push flow of the push information 130 within a predetermined period, the flow control information may also indicate the product corresponding to the push information 130 and a candidate solution related to the push information 130 in a set of candidate solutions (such as candidate solution 441 and candidate solution 442, which specifically indicates which one may be pre-configured). In box 430, the electronic device 120 adjusts the recommendation level 330 of each push information 130 based on a candidate solution related to each push information 130.

[0133] As an example, candidate solution 441 indicates that the first push flow rate is used as the maximum push flow rate of the push information 130. Candidate solution 442 indicates that the second push flow rate is used as the maximum push flow rate of the push information 130. The first push flow rate may be a static flow rate pre-set for the push information 130, and the second push flow rate may be a dynamic flow rate determined based on the static flow rate and the trend of the push information 130 in the historical push process.

[0134] As an example, in addition to the above content, the traffic control information may also include the historical exposure amount of the push information 130. The historical exposure amount may indicate the total number of times the push information 130 is presented to the user.

[0135] Figure 5 A schematic diagram showing an example 500 of adjusting a recommendation level according to an embodiment of the present disclosure is shown.

[0136] Reference Figure 5 In box 510, based on the historical exposure of a group of push information, the electronic device 120 determines the predicted total exposure of a group of push information within a predetermined period. In box 520, the electronic device 120 selects a candidate solution related to each push information 130. In box 530, based on the historical exposure of each push information 130 in a group of push information, the electronic device 120 determines the predicted exposure of each push information 130. In box 540, the electronic device 120 adjusts the recommendation level of each push information 130 based on the predicted total exposure, the predicted exposure of each push information 130, and the maximum push traffic of each push information 130.

[0137] As an example, the predicted exposure of each push information 130 can be stored in a database or a cache device, which is helpful to improve the calculation speed.

[0138] As an example, adjusting the recommendation level 330 of the push information 130 includes, but is not limited to, increasing the recommendation level 330 or decreasing the recommendation level 330 .

[0139] In some embodiments, if the push traffic of the candidate push information in a group of push information exceeds the maximum push traffic of the candidate push information within a predetermined period, the electronic device 120 reduces the recommendation level 330 of the candidate push information. If the push traffic of the candidate push information in a group of push information does not exceed the maximum push traffic of the candidate push information within a predetermined period, the electronic device 120 may not adjust the recommendation level 330 of the candidate push information.

[0140] As an example, the electronic device 120 can also use a proportional-integral-derivative (PID) controller to directly adjust the network traffic of the push information 130, thereby accurately controlling the push traffic of each push information 130 so that each push information 130 is kept at no more than a maximum push traffic.

[0141] In this way, the electronic device 120 can ensure that the exposure of each push information 130 matches the actual needs of different scenarios. In addition, this link also adds an extra layer of protection to the entire information push process, thereby solving problems such as excessive exposure of some push information 130 caused by upstream configuration errors or algorithm deviations, and thus making the exposure of each push information 130 stable after reaching the expected standard.

[0142] In block 240 , the electronic device 120 pushes the target push information to the target user.

[0143] In some embodiments, for one or more selected target push information, the electronic device 120 can also sort these target push information based on their respective recommendation levels 330, so as to facilitate the priority push of the most relevant and valuable push content to the user. As an example, the presentation of the target push information can be completed through the interface 150, such as displaying these target push information on the homepage or push page of an APP. Among them, the target push information with the highest recommendation level 330 can be located in the most prominent position in the interface 150, so that the user can see the target push information first when opening the interface 150.

[0144] It can be clearly understood from the various embodiments described above that the embodiments of the present disclosure can more accurately capture the interests and preferences of users by considering the historical interaction information 301 between users and push information 130. Compared with traditional information push strategies, the embodiments of the present disclosure pay more attention to the personalized needs of users. By deeply analyzing the interaction data between users and push information 130 (such as interactive activity), the embodiments of the present disclosure can explore the cross-relationship between user information and push result information, thereby improving the accuracy of information push. Accurate information push means that users are more likely to be interested in the pushed information and take further actions, such as clicking, purchasing or sharing. This will directly improve the conversion rate of push information 130, thereby improving the effectiveness of information push.

[0145] In summary, the information push solution provided by the embodiment of the present disclosure achieves more accurate and personalized information push by integrating and analyzing the historical interaction information 301 between users and content, thereby significantly improving the user experience and information conversion rate.

[0146] The embodiments of the present disclosure also provide corresponding devices for implementing the above methods or processes. Figure 6 A schematic structural block diagram of an apparatus 600 for information push according to some embodiments of the present disclosure is shown. The apparatus 600 may be implemented as or included in the electronic device 120. Each module / component in the apparatus 600 may be implemented by hardware, software, firmware or any combination thereof.

[0147] Reference Figure 6 , the device 600 includes a push information determination module 610, a recommendation degree determination module 620, a target push information determination module 630 and a push module 640. In some embodiments. The push information determination module 610 is configured to determine a group of push information for a target user and to be pushed. The recommendation degree determination module 620 is configured to determine the recommendation degree of each push information in a group of push information for the target user based on at least the first information, wherein the first information at least indicates the historical interaction information related to the target user for the group of push information. The target push information determination module 630 is configured to determine the target push information from a group of push information based on the recommendation degree of each push information in the group of push information. The push module 640 is configured to push the target push information to the target user.

[0148] In some embodiments, the recommendation degree determination module 620 is further configured to: determine the recommendation degree of each push information in a group of push information based on the first information and the second information; wherein the second information indicates at least one of the following: push activation information related to the target user for at least one push information in a group of push information; historical push result information related to a group of push information, the historical push result information at least indicating the historical click-through rate and / or historical conversion rate related to a group of push information; or user information related to the target user.

[0149] In some embodiments, the recommendation degree determination module 620 is also configured to: construct a first set of feature representations related to a first recommendation degree evaluation task for candidate push information in a group of push information based on historical push result information and user information; construct a second set of feature representations related to a second recommendation degree evaluation task for the candidate push information based on historical push result information, user information, historical interaction information and push activation information; and determine the recommendation degree of the candidate push information by executing the first recommendation degree evaluation task and the second recommendation degree evaluation task based on the first set of feature representations and the second set of feature representations.

[0150] In some embodiments, the recommendation degree determination module 620 is further configured to: determine the predicted conversion probability of the candidate push information for the target user by executing a first recommendation degree evaluation task based on a first set of feature representations; determine the predicted click probability of the candidate push information for the target user by executing a second recommendation degree evaluation task based on a second set of feature representations; and determine the degree of recommendation of the candidate push information based at least on the predicted conversion probability and the predicted click probability.

[0151] In some embodiments, the predicted conversion probability at least indicates a predicted probability that the target user will perform a target operation based on the content provided by the push information after clicking on at least one push information.

[0152] In some embodiments, the recommendation degree determination module 620 is further configured to: obtain predicted revenue information of the candidate push information for the target user; and determine the recommendation degree of the candidate push information based on the predicted click probability, predicted conversion probability and predicted revenue information.

[0153] In some embodiments, the predicted revenue information at least indicates the target user's lifetime value on the product related to the candidate push information.

[0154] In some embodiments, at least some of the feature representations in the first set of feature representations and the second set of feature representations are shared.

[0155] In some embodiments, the target group feature representation in the first group feature representation and the second group feature representation is determined in the following manner: in the first information and the second information, feature extraction is performed on information related to the target group feature representation to obtain at least one first feature; the missing values ​​of each first feature in the at least one first feature are filled with predetermined values ​​to obtain at least one second feature; and the at least one second feature is encoded to determine the target group feature representation.

[0156] In some embodiments, the recommendation degree determination module 620 is further configured to: in response to determining that there is a sparse feature representation corresponding to a sparse feature in the first group of feature representations and the second group of feature representations, convert the sparse feature representation into a dense feature representation by compressing the sparse feature representation; and determine the recommendation degree of the candidate push information by executing the first recommendation degree evaluation task and the second recommendation degree evaluation task at least based on the dense feature representation.

[0157] In some embodiments, the target push information determination module 630 is further configured to: adjust the recommendation level of each push information in a group of push information based on the traffic control information of each push information in a group of push information, wherein the traffic control information indicates the maximum push traffic of each push information in a group of push information within a predetermined period; and determine the target push information from a group of push information based on the adjusted recommendation level of each push information in a group of push information.

[0158] In some embodiments, the target push information determination module 630 is further configured to: if the push flow of candidate push information in a group of push information exceeds the maximum push flow of the candidate push information within a predetermined period, reduce the recommendation level of the candidate push information.

[0159] Figure 7 1 is a block diagram of an electronic device 700 in which one or more embodiments of the present disclosure may be implemented. The electronic device 700 may be used to implement, for example, Figure 1 The electronic device 120 shown. It should be understood that Figure 7 The electronic device 700 shown is merely exemplary and should not constitute any limitation on the functionality and scope of the embodiments described herein.

[0160] Reference Figure 7 , the electronic device 700 is in the form of a general electronic device. The components of the electronic device 700 may include, but are not limited to, one or more processors or processing units 710, a memory 720, a storage device 730, one or more communication units 740, one or more input devices 750, and one or more output devices 760. The processing unit 710 may be an actual or virtual processor and is capable of performing various processes according to a program stored in the memory 720. In a multi-processor system, multiple processing units execute computer executable instructions in parallel to improve the parallel processing capability of the electronic device 700.

[0161] The electronic device 700 typically includes a plurality of computer storage media. Such media can be any available media accessible to the electronic device 700, including but not limited to volatile and non-volatile media, removable and non-removable media. The memory 720 can be a volatile memory (e.g., register, cache, random access memory (RAM)), a non-volatile memory (e.g., read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory) or some combination thereof. The storage device 730 can be a removable or non-removable medium, and can include a machine-readable medium, such as a flash drive, a disk, or any other medium, which can be used to store information and / or data and can be accessed within the electronic device 700.

[0162] The electronic device 700 may further include additional removable / non-removable, volatile / non-volatile storage media. Figure 7 As shown in , a disk drive for reading or writing from a removable, non-volatile disk (e.g., a "floppy disk") and an optical drive for reading or writing from a removable, non-volatile optical disk may be provided. In these cases, each drive may be connected to the bus (not shown) by one or more data media interfaces. The memory 720 may include a computer program product 725 having one or more program modules that are configured to perform various methods or actions of various embodiments of the present disclosure.

[0163] The communication unit 740 implements communication with other electronic devices through a communication medium. Additionally, the functions of the components of the electronic device 700 can be implemented with a single computing cluster or multiple computing machines that can communicate through a communication connection. Therefore, the electronic device 700 can operate in a networked environment using a logical connection with one or more other servers, a network personal computer (PC), or another network node.

[0164] The input device 750 may be one or more input devices, such as a mouse, a keyboard, a tracking ball, etc. The output device 760 may be one or more output devices, such as a display, a speaker, a printer, etc. The electronic device 700 may also communicate with one or more external devices (not shown) through the communication unit 740 as needed, such as a storage device, a display device, etc., communicate with one or more devices that allow a user to interact with the electronic device 700, or communicate with any device that allows the electronic device 700 to communicate with one or more other electronic devices (e.g., a network card, a modem, etc.). Such communication may be performed via an input / output (I / O) interface (not shown).

[0165] According to an exemplary implementation of the present disclosure, a computer-readable storage medium is provided, on which computer-executable instructions are stored, wherein the computer-executable instructions are executed by a processor to implement the method described above. According to an exemplary implementation of the present disclosure, a computer program product is also provided, which is tangibly stored on a non-transitory computer-readable medium and includes computer-executable instructions, and the computer-executable instructions are executed by a processor to implement the method described above.

[0166] Various aspects of the present disclosure are described herein with reference to the flowcharts and / or block diagrams of the methods, devices, equipment, and computer program products implemented according to the present disclosure. It should be understood that each box in the flowchart and / or block diagram and the combination of each box in the flowchart and / or block diagram can be implemented by computer-readable program instructions.

[0167] These computer-readable program instructions can be provided to a processing unit of a general-purpose computer, a special-purpose computer, or other programmable data processing device, thereby producing a machine, so that when these instructions are executed by the processing unit of the computer or other programmable data processing device, a device that implements the functions / actions specified in one or more boxes in the flowchart and / or block diagram is generated. These computer-readable program instructions can also be stored in a computer-readable storage medium, and these instructions cause the computer, programmable data processing device, and / or other equipment to work in a specific manner, so that the computer-readable medium storing the instructions includes a manufactured product, which includes instructions for implementing various aspects of the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0168] Computer-readable program instructions can be loaded onto a computer, other programmable data processing apparatus, or other device so that a series of operational steps are performed on the computer, other programmable data processing apparatus, or other device to produce a computer-implemented process, so that the instructions executed on the computer, other programmable data processing apparatus, or other device implement the functions / actions specified in one or more boxes in the flowchart and / or block diagram.

[0169] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system, method and computer program product according to multiple implementations of the present disclosure. In this regard, each square box in the flow chart or block diagram can represent a part of a module, program segment or instruction, and a part of a module, program segment or instruction includes one or more executable instructions for realizing the logical function of the specification. In some implementations as replacements, the function marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or flow chart, and the combination of the square boxes in the block diagram and / or flow chart can be realized by a special hardware-based system that performs the function or action of the specification, or can be realized by a combination of special hardware and computer instructions.

[0170] The above descriptions of various implementations of the present disclosure are exemplary, non-exhaustive, and not limited to the disclosed implementations. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described implementations. The determination of the terms used herein is intended to best explain the principles of the implementations, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the various implementations disclosed herein.

Claims

1. A method for information push, comprising: Determine a set of push information for target users and to be pushed; Determining, based at least on first information, a recommendation level of each push information in the set of push information for the target user, wherein the first information at least indicates historical interaction information related to the target user for the set of push information; Determining target push information from the group of push information based on the recommendation degree of each push information in the group of push information; as well as Push the target push information to the target user.

2. The method according to claim 1, wherein determining the recommendation degree of each push information in the set of push information for the target user based at least on the first information comprises: Determining the recommendation level of each push information in the group of push information based on the first information and the second information; The second information indicates at least one of the following: Push activation information related to the target user for at least one push information in the set of push information; Historical push result information related to the set of push information, the historical push result information at least indicating a historical click rate and / or a historical conversion rate related to the set of push information; or User information related to the target user.

3. The method according to claim 2, wherein determining the recommendation level of each push information in the set of push information based on the first information and the second information comprises: Based on the historical push result information and the user information, determining a first set of feature representations related to a first recommendation evaluation task for candidate push information in the set of push information; Determining a second set of feature representations related to a second recommendation evaluation task for the candidate push information based on the historical push result information, the user information, the historical interaction information, and the push activation information; as well as Based on the first set of feature representations and the second set of feature representations, the recommendation degree of the candidate push information is determined by executing the first recommendation degree evaluation task and the second recommendation degree evaluation task.

4. The method according to claim 3, wherein determining the recommendation degree of the candidate push information by performing the first recommendation degree evaluation task and the second recommendation degree evaluation task based on the first set of feature representations and the second set of feature representations comprises: Based on the first set of feature representations, determining a predicted conversion probability of the candidate push information for the target user by performing the first recommendation evaluation task; Based on the second set of feature representations, determining a predicted click probability of the candidate push information for the target user by performing the second recommendation degree evaluation task; as well as The recommendation degree of the candidate push information is determined based at least on the predicted conversion probability and the predicted click probability.

5. The method according to claim 4, wherein the predicted conversion probability at least indicates the predicted probability that the target user will perform a target operation based on the content provided by the push information after clicking on at least one of the push information.

6. The method according to claim 4, wherein determining the recommendation degree of the candidate push information based at least on the predicted conversion probability and the predicted click probability comprises: Obtaining predicted revenue information of the candidate push information for the target user; as well as The recommendation degree of the candidate push information is determined based on the predicted click probability, the predicted conversion probability and the predicted revenue information. 7 . The method according to claim 6 , wherein the predicted revenue information at least indicates the life cycle value of the target user on the product related to the candidate push information. The method according to claim 4 , wherein at least part of the feature representations in the first set of feature representations and the second set of feature representations are shared.

9. The method according to claim 3, wherein the target group feature representations in the first group feature representations and the second group feature representations are determined by: In the first information and the second information, feature extraction is performed on information related to the target group feature representation to obtain at least one first feature; Filling the missing value of each first feature in the at least one first feature with a predetermined value to obtain at least one second feature; and The at least one second feature is encoded to determine the target set feature representation.

10. The method according to claim 3, wherein determining the recommendation degree of the candidate push information by performing the first recommendation degree evaluation task and the second recommendation degree evaluation task based on the first set of feature representations and the second set of feature representations comprises: In response to determining that there is a sparse feature representation corresponding to a sparse feature in the first set of feature representations and the second set of feature representations, converting the sparse feature representation into a dense feature representation by compressing the sparse feature representation; as well as At least based on the dense feature representation, the recommendation degree of the candidate push information is determined by executing the first recommendation degree evaluation task and the second recommendation degree evaluation task.

11. The method according to claim 1, wherein determining target push information from the group of push information based on the recommendation level of each push information in the group of push information comprises: adjusting the recommendation degree of each push information in the group of push information based on the flow control information of each push information in the group of push information, wherein the flow control information indicates the maximum push flow of each push information in the group of push information within a predetermined period; as well as Based on the adjusted recommendation levels of each push information in the group of push information, target push information is determined from the group of push information.

12. The method according to claim 11, wherein adjusting the recommendation level of each push information in the group of push information based on the flow control information of each push information in the group of push information comprises: If the push flow of the candidate push information in the group of push information within the predetermined period exceeds the maximum push flow of the candidate push information, the recommendation level of the candidate push information is reduced.

13. A device for information push, comprising: A push information determination module, configured to determine a set of push information to be pushed to a target user; a recommendation degree determination module, configured to determine a recommendation degree of each push information in the set of push information for the target user based at least on first information, wherein the first information at least indicates historical interaction information related to the target user for the set of push information; a target push information determination module, configured to determine target push information from the group of push information based on the recommendation degree of each push information in the group of push information; as well as The push module is configured to push the target push information to the target user.

14. An electronic device comprising: at least one processing unit; as well as At least one memory, the at least one memory being coupled to the at least one processing unit and storing instructions for execution by the at least one processing unit, the instructions causing the electronic device to perform the method according to any one of claims 1 to 12 when executed by the at least one processing unit.

15. A computer-readable storage medium having a computer program stored thereon, wherein the computer program can be executed by a processor to implement the method according to any one of claims 1 to 12.

16. A computer program product comprising computer executable instructions, wherein the computer executable instructions, when executed by a processor, implement the method according to any one of claims 1 to 12.

Images