A method, apparatus, and electronic device for pushing network resources.

Abstract

This invention provides a method, apparatus, and electronic device for pushing network resources, relating to the field of Internet technology. The method includes: upon receiving a network resource acquisition request from a target client, determining the number of historical network resource acquisition requests received from the target client within a specified time period prior to the current moment, as a baseline count; randomly selecting a first number of first network resources from a preset pool of candidate network resources; wherein the first number is negatively correlated with the baseline count; selecting a second number of first network resources as network resources to be pushed, based on the matching degree between each first network resource and the user characteristics of the target user; wherein the target user is a user logged into the target client; and pushing resources to the target client according to the network resources to be pushed. This solution can better meet the personalized needs of users.

Description

Technical Field

[0001] This invention relates to the field of Internet technology, and in particular to a method, apparatus, and electronic device for pushing network resources. Background Technology

[0002] Currently, the client can display online resources on the interface for users to choose from. These resources can include videos, news, and products. While browsing these resources, if a user is not interested in the content displayed, they will actively refresh the client interface. Upon detecting this refresh, the server will select new online resources from a pre-stored pool and push them to the client for re-display.

[0003] In related technologies, when a user actively refreshes the client interface, the server selects the most popular network resources from a pre-stored pool of network resources and pushes them to the client. This fails to meet the user's personalized needs and results in a poor user experience. Summary of the Invention

[0004] The purpose of this invention is to provide a method, apparatus, and electronic device for pushing network resources, so as to better meet user needs and improve user experience. The specific technical solution is as follows:

[0005] In a first aspect of this invention, a method for pushing network resources is provided, the method comprising:

[0006] Upon receiving a network resource acquisition request from a target client, determine the number of network resource acquisition requests historically received from the target client within a specified time period prior to the current moment, and use this as a baseline number;

[0007] A first number of first network resources are randomly selected from a preset pool of candidate network resources; wherein the first number is negatively correlated with the baseline number of times.

[0008] Based on the matching degree between each first network resource and the user characteristics of the target user, a second number of first network resources are selected as network resources to be pushed; wherein, the target user is the user who logs into the target client;

[0009] The resources are pushed to the target client according to the network resources to be pushed.

[0010] Optionally, selecting a second number of first network resources as network resources to be pushed, based on the matching degree between each first network resource and the user characteristics of the target user, includes:

[0011] For each first network resource, the user features of the target user and the resource features of the first network resource are input into a pre-trained resource recommendation model to obtain the first matching degree between the first network resource and the target user; wherein, the resource recommendation model is trained based on the user features of the sample user and the resource features of the sample network resources that match the sample user.

[0012] Based on the determined first matching degree, a second number of network resources are selected from each of the first network resources as network resources to be pushed.

[0013] Optionally, the step of selecting a second number of network resources from the various first network resources as network resources to be pushed, based on the determined first matching degree, includes:

[0014] Identify other users whose user characteristics are most similar to the target user as similar users;

[0015] The network resources selected by the similar users when pushing network resources to the similar users within a preset historical time period are obtained as the second network resources.

[0016] For each first network resource, the similarity of resource features between the first network resource and the second network resource is determined to obtain the second matching degree between the first network resource and the target user;

[0017] Select the second number of network resources with the highest sum of first and second matching degrees from the various first network resources, and use them as the network resources to be pushed.

[0018] Optionally, there are multiple second network resources. For each first network resource, determining the similarity of resource features between the first network resource and the second network resource to obtain a second matching degree between the first network resource and the target user includes:

[0019] For each first network resource, the feature similarity between the resource features of the first network resource and each second network resource is determined, and the average value of the determined feature similarity is calculated to obtain the second matching degree between the first network resource and the target user.

[0020] Optionally, before randomly selecting a first number of first network resources from a preset pool of candidate network resources, the method further includes:

[0021] From the pre-established correspondence between the set of request counts and the set of proportion values, determine the proportion value corresponding to the baseline count as the target proportion value; wherein, in the correspondence, the number of requests and the proportion value are negatively correlated.

[0022] The first number is determined based on the target ratio and the total number of candidate network resources.

[0023] Optionally, selecting a second number of first network resources as network resources to be pushed, based on the matching degree between each first network resource and the user characteristics of the target user, includes:

[0024] The maximum matching degree is determined from the matching degree between each primary network resource and the user characteristics of the target user;

[0025] If the maximum determined matching degree is less than the preset matching degree threshold, return to the step of randomly selecting a first number of first network resources from the preset candidate network resources, until the maximum determined matching degree is not less than the preset matching degree threshold. From each of the currently selected first network resources, according to the matching degree between each first network resource and the user characteristics of the target user, select a second number of network resources as the network resources to be pushed.

[0026] Optionally, the second number is the total number of network resources that the target client's display interface can display.

[0027] In a second aspect of the present invention, a network resource push device is also provided, the device comprising:

[0028] The count determination module is used to determine, upon receiving a network resource acquisition request sent by the target client, the number of times the target client has historically sent network resource acquisition requests within a specified time period prior to the current moment, and use this number as a baseline count.

[0029] The first selection module is used to randomly select a first number of first network resources from a preset pool of candidate network resources; wherein the first number is negatively correlated with the baseline number of times.

[0030] The second selection module is used to select a second number of first network resources as network resources to be pushed, based on the matching degree between each first network resource and the user characteristics of the target user; wherein, the target user is the user who logs into the target client;

[0031] The push module is used to push resources to the target client according to the network resources to be pushed.

[0032] In a third aspect of the present invention, an electronic device is also provided, including a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus.

[0033] Memory, used to store computer programs;

[0034] The processor, when executing a program stored in memory, implements any of the above-described methods for pushing network resources.

[0035] In another aspect of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, and when the computer program is executed by a processor, it implements any of the above-described methods for pushing network resources.

[0036] In another aspect of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the above-described methods for pushing network resources.

[0037] The network resource push method provided in this embodiment of the invention, upon receiving a network resource acquisition request sent by a target client, determines the number of times the target client has historically sent network resource acquisition requests within a specified time period prior to the current moment, as a baseline number; randomly selects a first number of first network resources from a preset pool of candidate network resources; wherein the first number is negatively correlated with the baseline number; selects a second number of first network resources as network resources to be pushed according to the matching degree between each first network resource and the user characteristics of the target user; wherein the target user is a user logged into the target client; and pushes resources to the target client according to the network resources to be pushed. In this scheme, the number of randomly selected first network resources is negatively correlated with the number of network resource acquisition requests sent by the target client in the past. Therefore, an increase in the number of network resource acquisition requests sent by the target client in the past can determine a smaller number of first network resources. Consequently, when selecting network resources to be pushed from the first network resources according to the matching degree, it is easier to select more network resources with a lower degree of matching with user characteristics. When the number of network resource acquisition requests sent by the target client in the past is high, it often means that the user is dissatisfied with the network resources currently pushed with a high degree of matching with user characteristics. This scheme, while taking into account the matching degree of user characteristics, can push more network resources with a lower degree of matching as the number of network resource acquisition requests sent by the target client in the past increases. Therefore, it can better meet user needs and improve user experience. Attached Figure Description

[0038] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below.

[0039] Figure 1 This is a flowchart of the method for pushing network resources in an embodiment of the present invention;

[0040] Figure 2This is another flowchart of the method for pushing network resources in an embodiment of the present invention;

[0041] Figure 3 This is a schematic diagram of a method for pushing network resources in an embodiment of the present invention;

[0042] Figure 4 This is a schematic diagram of the structure of the network resource push device in an embodiment of the present invention;

[0043] Figure 5 This is a schematic diagram of the structure of an electronic device in an embodiment of the present invention. Detailed Implementation

[0044] The technical solutions of the present invention will now be described with reference to the accompanying drawings in the embodiments of the present invention.

[0045] To better meet users' personalized needs, embodiments of the present invention provide a method, apparatus, and electronic device for pushing network resources. The method for pushing network resources can be applied to a server and includes:

[0046] Upon receiving a network resource acquisition request from the target client, determine the number of network resource acquisition requests previously received from the target client within a specified time period prior to the current moment, and use this as the baseline number;

[0047] A first number of first network resources are randomly selected from a preset pool of candidate network resources; wherein the first number is negatively correlated with the baseline number.

[0048] Based on the matching degree between each first network resource and the user characteristics of the target user, a second number of first network resources are selected as network resources to be pushed; among them, the target user is the user who has logged into the target client;

[0049] Based on the network resources to be pushed, resources are pushed to the target client.

[0050] In this embodiment, the number of randomly selected first network resources is negatively correlated with the number of network resource acquisition requests sent by the target client in the past. Therefore, an increase in the number of network resource acquisition requests sent by the target client in the past can determine a smaller number of first network resources. Consequently, when selecting network resources to be pushed from the first network resources according to the matching degree, it is easier to select more network resources with a lower degree of matching with user characteristics. When the number of network resource acquisition requests sent by the target client in the past is large, it often means that the user is dissatisfied with the network resources that are currently pushed with a high degree of matching with user characteristics. This solution, while taking into account the matching degree of user characteristics, can push more network resources with a lower degree of matching as the number of network resource acquisition requests sent by the target client in the past increases. Therefore, it can better meet user needs and improve user experience.

[0051] The following is a detailed description of a method for pushing network resources provided by an embodiment of the present invention, with reference to the accompanying drawings.

[0052] like Figure 1 As shown, the method for pushing network resources provided in this embodiment of the invention may include the following steps:

[0053] S101, upon receiving a network resource acquisition request sent by the target client, determine the number of network resource acquisition requests previously sent by the target client within a specified time period before the current moment, and use this number as a baseline number;

[0054] The aforementioned network resources can be any network resources available for user browsing, such as videos, news, and products. The aforementioned target client can be any client with network resource display capabilities; it can be an electronic device, such as a mobile phone or computer, or a software program running on an electronic device. The target client's display interface is the interface through which the network resources are displayed.

[0055] The specified duration can be set based on experience and needs, such as one minute, two minutes, etc. Network resource acquisition requests can be sent by the target user logged into the target client during historical refresh operations on the displayed interface. To determine the number of historical network resource acquisition requests received from the target client within the specified duration prior to the current moment, the timestamp of each received request can be recorded. This timestamp can then be used to determine the baseline count of historical network resource acquisition requests received from the target client within the specified duration prior to the current moment. Alternatively, the target client can record the timestamp of each refresh operation performed by the user on the target client's displayed interface, for example, in the client's log information. The timestamps of each recorded refresh operation can then be obtained from the log information to determine the baseline count of historical refresh operations performed by the user within the specified duration prior to the current moment. Historical network resource acquisition requests may include those currently received from the target client.

[0056] S102, randomly select a first number of first network resources from the preset candidate network resources; wherein, the first number is negatively correlated with the baseline number;

[0057] In one implementation, a correspondence between a set of numbers and a set of frequencies can be established in advance, in which the number and the frequency are negatively correlated. Then, after determining the base frequency, the number corresponding to the base frequency can be determined as the first number.

[0058] In one implementation, before randomly selecting a first number of first network resources from a preset pool of candidate network resources, a target ratio value can be determined from a pre-established correspondence between a set of request counts and a set of ratio values, corresponding to a baseline count; wherein, in the correspondence, the number of requests and the ratio value are negatively correlated; and then the first number is determined based on the target ratio value and the total number of candidate network resources.

[0059] The aforementioned set of request counts can contain at least one set of request counts, and the set of proportion values ​​can contain at least one proportion value. Each request count corresponds to a proportion value, and in this correspondence, the larger the request count, the smaller the proportion value. This correspondence can be a negatively correlated function. Therefore, after determining the baseline count, this baseline count can be substituted into this function to determine the corresponding proportion value. After determining the target proportion value, the product of the total number of candidate network resources and the target proportion value can be calculated to obtain the first number.

[0060] The aforementioned preset candidate network resources can be network resources used for push notifications on the server. For example, they can be all or some of the network resources online on the server. After determining the first number, a first number of network resources can be randomly selected from the preset candidate network resources as the first network resource.

[0061] S103, based on the matching degree between each first network resource and the user characteristics of the target user, select a second number of first network resources as network resources to be pushed; wherein, the target user is the user who has logged into the target client;

[0062] The second number mentioned above may be the total number of network resources that can be displayed on the target client's display interface, or the second number may be less than the total number of network resources that can be displayed on the target client's display interface; the user characteristics of the target user may be determined based on the personal information set by the target user and / or the target user's selection behavior of historically pushed network resources.

[0063] In one implementation, the target user's user characteristics can be tags. For example, if the target client is a video playback client, and the target user's personal information includes a preference for comedy, then the preference for comedy can be used as the target user's tag. Alternatively, the user's preference type can be determined according to preset rules, such as selecting the most frequently chosen type from historically pushed online resources as the user's preference type. Then, online resources with the target user's preference type can be considered as online resources with a high degree of matching with the target user's user characteristics, and thus, online resources with the target user's preference type can be selected from the first set of online resources as the online resources to be pushed. Alternatively, the target user's tags can be processed into vector form to obtain the target user's user characteristics. For example, the bag-of-words model can be used to vectorize the target user's tags, and the tags of the first online resources can also be vectorized to obtain the resource characteristics of the first online resources. Then, the similarity between the target user's user characteristics and the resource characteristics of each of the first online resources can be calculated as the matching degree, and the second number of online resources with the highest similarity among the first online resources can be selected as the online resources to be pushed.

[0064] In another implementation, a digital model representing the target user can be established based on the target user's information, such as gender, age, and / or user behavior information, such as their selection behavior of historically pushed network resources. This digital model can serve as the user's user characteristics. For example, this digital model can be a vector, where each element represents a piece of information about the target user. For instance, each element could represent: age, gender (1 for male, 0 for female), whether they like current affairs news (1 for liking, 0 for disliking), whether they like comedy, and whether they like action movies. For a 30-year-old male user, if their preference for current affairs news is pre-determined based on their selection behavior of historically pushed network resources, their digital model could be represented as a vector: (30, 1, 1, 0, 0). In this implementation, selecting a second number of network resources as the network resources to be pushed, based on the matching degree between each first network resource and the target user's user characteristics, includes the following steps:

[0065] Step A1: For each first network resource, input the user features of the target user and the resource features of the first network resource into the pre-trained resource recommendation model to obtain the first matching degree between the first network resource and the target user; wherein, the resource recommendation model is trained based on the user features of the sample user and the resource features of the sample network resources that match the sample user.

[0066] In this implementation, the resource features of the aforementioned online resources can be pre-defined. For example, a bag-of-words model can be used to extract textual features such as tags, titles, and descriptions of the online resources, which can then be used as the resource features of that online resource. Alternatively, if a cover image of the online resource exists, Convolutional Neural Networks (CNNs) can be used to extract features from the cover image, such as color and texture features, to obtain image features that can also be used as the resource features of that online resource. Furthermore, the textual and image features of the online resource can be merged to obtain the resource features.

[0067] The resource recommendation model described above can be any model used for resource recommendation, such as CAN (Co-Action Network, feature-based collaborative interaction model) or LR (Logistic Regression) model. The resource recommendation model can be trained based on the user characteristics of sample users and the resource characteristics of sample network resources that match the sample users. Furthermore, the resource recommendation model can include an attention mechanism to enhance the model's information mining and representation, thereby improving the accuracy and interpretability of the recommendations. The sample users can be any user. For example, if a sample user has selected a previously recommended network resource, that network resource can be considered a positive sample network resource matching the sample user; if the sample user has not selected a previously recommended network resource, that network resource can be considered a negative sample network resource that does not match the sample user. Alternatively, if a sample user has rated a previously recommended network resource, and the score exceeds a first preset threshold, that network resource can be considered a positive sample network resource matching the sample user; if the score is below a second preset threshold, that network resource can be considered a negative sample network resource that does not match the sample user. Of course, the method for determining sample network resources is not limited to these. The specific training method for the resource recommendation model will be introduced below, and will not be repeated here.

[0068] After inputting the user characteristics of the target user and the resource characteristics of the network resource into the pre-trained resource recommendation model, the resource recommendation model will output the correlation between the network resource and the target user, and the output correlation can be used as the first matching degree between the network resource and the target user.

[0069] Step A2: Based on the determined first matching degree, select a second number of network resources from each of the first network resources as network resources to be pushed.

[0070] In this step, the second number of network resources with the highest first matching degree can be directly selected from each first network resource as the network resources to be pushed.

[0071] Because related technologies typically push resources based solely on user interests or the popularity of online resources, the pushed resources tend to be highly similar and monotonous. Even with multiple refreshes of the display interface, the types of pushed resources rarely change significantly after each refresh, easily leading to user boredom. Furthermore, the higher the frequency of user refreshes, the greater the user's need for novel online resources. In this embodiment, the first number is negatively correlated with the baseline number. This allows for the determination of a smaller first number when user requests are frequent, thereby increasing the randomness of the pushed online resources and making it easier for the server to push novel resources, thus satisfying users' demand for novel online resources. Furthermore, by selecting online resources that match the target user's user characteristics, user preferences can be taken into account, thereby increasing user acceptance of the push results, improving user experience, and enhancing user stickiness.

[0072] S104: Push resources to the target client according to the network resources to be pushed.

[0073] After identifying the network resources to be pushed, the resources can be sent to the target client. This can be done by sending the network resources themselves, or their associated information, such as title text, links associated with the resource, or cover images associated with the resource. The target client can then display the network resources in its interface. In one implementation, the client can display the network resources sequentially in descending order of their matching degree. Furthermore, the client can personalize the display of the network resources; for example, if the target client initially displays a cover image, that cover image can be dynamically displayed, further enhancing the user's personalized experience and satisfaction.

[0074] In this embodiment, the number of randomly selected first network resources is negatively correlated with the number of network resource acquisition requests sent by the target client in the past. Therefore, an increase in the number of network resource acquisition requests sent by the target client in the past can determine a smaller number of first network resources. Consequently, when selecting network resources to be pushed from the first network resources according to the matching degree, it is easier to select more network resources with a lower degree of matching with user characteristics. When the number of network resource acquisition requests sent by the target client in the past is large, it often means that the user is dissatisfied with the network resources that are currently pushed with a high degree of matching with user characteristics. This solution, while taking into account the matching degree of user characteristics, can push more network resources with a lower degree of matching as the number of network resource acquisition requests sent by the target client in the past increases. Therefore, it can better meet user needs and improve user experience.

[0075] Optionally, in another embodiment of the present invention, the above-mentioned selection of a second number of network resources from each of the first network resources as network resources to be pushed, based on the determined first matching degree, may include:

[0076] Step B1: Identify other users whose user characteristics are most similar to the target user, and designate them as similar users;

[0077] In this step, user characteristics of multiple users can be pre-recorded. These users can be those who selected network resources to be pushed to them within a preset historical time period. Then, when pushing network resources to the target user, the user characteristics of the target user and other users can be used to identify other users with the highest similarity to the target user's characteristics, and these are considered similar users. There are various ways to determine the similarity of user characteristics; for example, collaborative filtering algorithms can be used.

[0078] Step B2: Obtain the network resources selected by similar users when pushing network resources to similar users within a preset historical time period, and use them as the second network resources;

[0079] The aforementioned preset historical time period can be set based on experience and needs. For example, it could be the period within one week or one day prior to the current moment. When a user clicks, comments, likes, or gives a rating exceeding a first preset threshold on the client's display interface, the network resource can be considered as the network resource selected by the user. Of course, this is not the only option.

[0080] Step B3: For each first network resource, determine the similarity of resource features between the first network resource and the second network resource to obtain the second matching degree between the first network resource and the target user;

[0081] In this step, for each first network resource, the similarity of resource features between the first network resource and the second network resource can be calculated, and the obtained similarity can be used as the second matching degree between the first network resource and the target user. When there are multiple second network resources, for each first network resource, the feature similarity of resource features between the first network resource and each of the second network resources can be determined, and the average value of the determined feature similarities can be calculated to obtain the second matching degree between the first network resource and the target user.

[0082] Step B4: Select the second number of network resources with the highest sum of first and second matching degrees from each first network resource as the network resources to be pushed.

[0083] After determining the first matching degree and the second matching degree of each network resource, the first matching degree and the second matching degree of each network resource can be added together, and then the second number of network resources with the highest sum of the first matching degree and the second matching degree can be selected from each first network resource as the network resources to be pushed.

[0084] In this embodiment, other users with the highest similarity to the target user's user characteristics are identified as similar users; network resources selected by similar users when pushing network resources to similar users within a preset historical time period are obtained as second network resources; and for each first network resource, the similarity of the resource characteristics of the first network resource and the second network resource is determined to obtain the second matching degree between the first network resource and the target user; then, a second number of network resources with the highest sum of the first matching degree and the second matching degree are selected from the first network resources as network resources to be pushed. This approach can take into account the network resources selected by other users similar to the target user, thereby further improving the accuracy of the pushed network resources.

[0085] In another embodiment of the invention, such as Figure 2 As shown, the method for pushing network resources may include the following steps:

[0086] S201, Upon receiving a network resource acquisition request sent by the target client, determine the number of network resource acquisition requests previously sent by the target client within a specified time period before the current moment, and use this number as a baseline count;

[0087] S202, randomly select a first number of first network resources from the preset candidate network resources; wherein, the first number is negatively correlated with the baseline number;

[0088] The steps S201-S202 above are similar to steps S101-S102 above, and will not be repeated here.

[0089] S203, determine the maximum matching degree from the matching degree between each first network resource and the user characteristics of the target user;

[0090] S204, Determine whether the determined maximum matching degree is less than the preset matching degree threshold;

[0091] If it is less than, return to step S202; if it is not less than, proceed to step S205.

[0092] S205, From the currently selected first network resources, select a second number of network resources as network resources to be pushed, according to the matching degree between each first network resource and the user characteristics of the target user.

[0093] S206: Push resources to the target client according to the network resources to be pushed.

[0094] The steps S205-S206 above are similar to steps S103-S104 above, and will not be repeated here.

[0095] In this embodiment, user needs can be better met and user experience improved. Furthermore, if the determined maximum matching degree is less than a preset matching degree threshold, the process returns to randomly selecting a first number of first network resources from a preset pool of candidate network resources until the determined maximum matching degree is not less than the preset matching degree threshold. From the currently selected first network resources, a second number of network resources are selected based on the matching degree between each first network resource and the target user's user characteristics as the network resources to be pushed. This ensures that the determined first network resources include those with a high matching degree with the target user's user characteristics, thereby ensuring that at least one network resource among the network resources to be pushed matches the user's interests, further improving the user experience.

[0096] In another embodiment of the present invention, the training process of the above-mentioned resource recommendation model includes the following steps:

[0097] Step C1: Obtain the user characteristics of the sample users and their historical selection information; wherein, the historical selection information represents: the positive sample network resources selected by the sample users and the negative sample network resources not selected when pushing sample network resources to the sample users; the label of the positive sample network resource indicates that the positive sample network resource matches the sample user, and the label of the negative sample network resource indicates that the negative sample network resource does not match the sample user.

[0098] In this step, when pushing sample network resources to sample users, the sample network resources selected by the sample users can be regarded as positive sample network resources. For example, if the sample user clicks, comments, likes, or gives a rating exceeding a first preset threshold for the pushed sample network resource, it can be regarded as the sample user selecting the network resource. If the sample user does not select the sample network resource when it is pushed to the sample user, the sample network resource can be regarded as a negative sample network resource. For example, if the sample user does not click on the pushed sample network resource, or gives the sample network resource a rating lower than a second preset threshold, it can be regarded as the sample user not selecting the network resource.

[0099] Furthermore, given the positive and negative sample network resources, labels can be assigned to each resource. The label for a positive sample network resource indicates a match between the resource and the user, while the label for a negative sample network resource indicates a mismatch. For example, a positive sample network resource can be labeled with 1, and a negative sample network resource can be labeled with 0.

[0100] Step C2: Input the user characteristics of the sample users and the resource characteristics of each sample network resource into the resource recommendation model of the initial structure to obtain the prediction result corresponding to the sample network resource.

[0101] After identifying positive and negative sample network resources for a sample user, the user characteristics of the sample user and the resource characteristics of each sample network resource can be input into the resource recommendation model of the initial structure to obtain the prediction result corresponding to the sample network resource, that is, to predict the matching degree between the sample network resource and the sample user.

[0102] Step C3: Based on the obtained prediction results and the label of each sample network resource, adjust the model parameters of the resource recommendation model of the initial structure until the model converges, and obtain the trained resource recommendation model.

[0103] Having obtained the prediction result for each sample network resource, the prediction result and the label of that sample network resource can be input into the loss function of the resource recommendation model to calculate the loss value. Then, the model parameters of the initial resource recommendation model can be adjusted based on the obtained loss value, for example, using gradient descent, until the model converges, resulting in a trained resource recommendation model. Furthermore, during training, a portion of the predicted results and labels of sample network resources can be used as the training set, and another portion as the validation set. Based on the processing results of the resource recommendation model on the validation set, the model can be adjusted and optimized to improve the accuracy and diversity of network resource recommendations using this model.

[0104] Furthermore, in this embodiment, different users can be periodically used as sample users, and the latest user characteristics of the sample users and the historical selection information of the sample users can be obtained to train the resource recommendation model, so that the resource recommendation model can be adjusted in a timely manner to adapt to changes in user interests.

[0105] In this embodiment, a resource recommendation model is trained by using the positive sample network resources selected by the sample users and the negative sample network resources not selected when pushing sample network resources to sample users. The deep learning capability is used to mine the correlation between user features and network resources, so that the resource recommendation model can learn user interests and preferences. Therefore, when using the trained resource recommendation model to push network resources, the accuracy of the pushed network resources can be improved.

[0106] To make it easier to understand, the following will be combined with... Figure 3 The method for pushing network resources provided in the embodiments of the present invention will be described, such as... Figure 3 As shown, the following parts may be involved in implementing this method:

[0107] The user interface, or the display interface of the target client, is used to display network resources and detect refresh operations of the target user, such as pull-to-refresh. When the target user performs a refresh operation, the target client sends a network resource acquisition request.

[0108] The adaptive strategy determination end is used to determine the recommendation strategy based on the target user's historical refresh operation behavior on the display interface. That is, based on the number of network resource acquisition requests received from the target client within a specified time period before the current time, the first number of the first network resources to be randomly selected is determined.

[0109] The recommendation module, based on the user characteristics of the target user and the resource characteristics of the network resources, uses a pre-trained resource recommendation model and a determined recommendation strategy to determine the network resources to be pushed to the target client.

[0110] The data collection and preprocessing end is used to collect user characteristics, behavioral information, and network resource characteristics from multiple users, and to preprocess the collected data, such as removing duplicates, filling in missing data, and removing abnormal data, to ensure the quality and format consistency of the collected data.

[0111] The user modeling module is used to build models representing users based on collected user characteristics and behavioral information, and to calculate the similarity between users using collaborative filtering algorithms.

[0112] The resource feature extraction end is used to extract features from network resources to obtain the resource features of the network resources.

[0113] The model training end is used to train and validate the resource recommendation model, resulting in a fully trained resource recommendation model.

[0114] Based on the same inventive concept, embodiments of the present invention also provide a network resource push device, such as... Figure 4 As shown, the device includes:

[0115] The number of requests for historical refresh operations performed by the target user who logged into the target client on the display interface within a specified time period before the current moment is determined when a refresh operation for the display interface of the target client is detected, and the number of requests is used as a baseline number.

[0116] The first selection module 402 is used to randomly select a first number of first network resources from a preset pool of candidate network resources; wherein the first number is negatively correlated with the baseline number of times.

[0117] The second selection module 403 is used to select a second number of first network resources as network resources to be pushed, based on the matching degree between each first network resource and the user characteristics of the target user; wherein, the target user is the user who logs into the target client;

[0118] The push module 404 is used to push resources to the target client according to the network resources to be pushed.

[0119] Optionally, the second selection module 403 includes:

[0120] The first matching degree determination submodule is used to input the user features of the target user and the resource features of the first network resource into a pre-trained resource recommendation model for each first network resource to obtain the first matching degree between the first network resource and the target user; wherein, the resource recommendation model is trained based on the user features of the sample user and the resource features of the sample network resources that match the sample user.

[0121] The selection submodule is used to select a second number of network resources from each first network resource based on the determined first matching degree, as the network resources to be pushed.

[0122] Optionally, the selection submodule includes:

[0123] The similar user determination unit is used to determine other users whose user characteristics are most similar to the target user, as similar users;

[0124] The acquisition unit is used to acquire the network resources selected by the similar users when pushing network resources to the similar users within a preset historical time period, and use them as the second network resources.

[0125] The second matching degree determination unit is used to determine the similarity of resource features between the first network resource and the second network resource for each first network resource, and obtain the second matching degree between the first network resource and the target user.

[0126] The selection unit is used to select a second number of network resources with the highest sum of first matching degree and second matching degree from each first network resource, as the network resources to be pushed.

[0127] Optionally, there are multiple second network resources, and the second matching degree determination unit is specifically used for:

[0128] For each first network resource, the feature similarity between the resource features of the first network resource and each second network resource is determined, and the average value of the determined feature similarity is calculated to obtain the second matching degree between the first network resource and the target user.

[0129] Optionally, the device further includes:

[0130] The target ratio value determination module is used to determine the ratio value corresponding to the benchmark number of requests from a pre-established correspondence between the request count set and the ratio value set before the first selection module 402 randomly selects a first number of first network resources from the preset candidate network resources. The target ratio value is defined as such. In the correspondence, the request count is negatively correlated with the ratio value.

[0131] The first number determination module determines the first number based on the target ratio value and the total number of candidate network resources.

[0132] Optionally, the second selection module 403 is specifically used to determine the maximum matching degree from the matching degree between each first network resource and the user characteristics of the target user; if the determined maximum matching degree is less than a preset matching degree threshold, the first selection module 402 is triggered to perform the step of randomly selecting a first number of first network resources from the preset candidate network resources until the determined maximum matching degree is not less than the preset matching degree threshold, and from the currently selected first network resources, according to the matching degree between each first network resource and the user characteristics of the target user, a second number of network resources are selected as network resources to be pushed.

[0133] Optionally, the second number is the total number of network resources that the target client's display interface can display.

[0134] This invention also provides an electronic device, such as... Figure 5 As shown, it includes a processor 501, a communication interface 502, a memory 503, and a communication bus 504, wherein the processor 501, the communication interface 502, and the memory 503 communicate with each other through the communication bus 504.

[0135] Memory 503 is used to store computer programs;

[0136] When the processor 501 executes the program stored in the memory 503, it implements the network resource push method described in any of the above embodiments.

[0137] The communication bus mentioned above can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This communication bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0138] The communication interface is used for communication between the aforementioned terminal and other devices.

[0139] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0140] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0141] In another embodiment of the present invention, a computer-readable storage medium is also provided, wherein a computer program is stored therein, and when the computer program is executed by a processor, it implements the network resource push method described in any of the above embodiments.

[0142] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to execute any of the network resource push methods described in the above embodiments.

[0143] In this embodiment of the invention, the acquisition, storage, use, processing, transmission, provision, and disclosure of event data, type information, object information, identification information, statistical information, attribute data, transfer data, paper data, and user personal data are all carried out with the authorization of the user or the rights holder.

[0144] In the above embodiments, implementation can be achieved entirely or partially through software, hardware, firmware, or any combination thereof. When implemented using software, it can be implemented entirely or partially in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that integrates one or more available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid state disk (SSD)).

[0145] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0146] The various embodiments in this specification are described in a related manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0147] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention are included within the scope of protection of the present invention.

Claims

1. A method for pushing network resources, characterized in that, The method includes: Upon receiving a network resource acquisition request from a target client, determine the number of network resource acquisition requests historically received from the target client within a specified time period prior to the current moment, and use this as a baseline number; A first number of first network resources are randomly selected from a preset pool of candidate network resources; wherein the first number is negatively correlated with the baseline number of times. Based on the matching degree between each first network resource and the user characteristics of the target user, a second number of first network resources are selected as network resources to be pushed; wherein, the target user is the user who logs into the target client; The resources are pushed to the target client according to the network resources to be pushed.

2. The method according to claim 1, characterized in that, The step of selecting a second number of first network resources as network resources to be pushed, based on the matching degree between each first network resource and the user characteristics of the target user, includes: For each first network resource, the user features of the target user and the resource features of the first network resource are input into a pre-trained resource recommendation model to obtain the first matching degree between the first network resource and the target user; wherein, the resource recommendation model is trained based on the user features of the sample user and the resource features of the sample network resources that match the sample user. Based on the determined first matching degree, a second number of network resources are selected from each of the first network resources as network resources to be pushed.

3. The method according to claim 2, characterized in that, The step of selecting a second number of network resources from each of the first network resources based on the determined first matching degree as network resources to be pushed includes: Identify other users whose user characteristics are most similar to the target user as similar users; The network resources selected by the similar users when pushing network resources to the similar users within a preset historical time period are obtained as the second network resources. For each first network resource, the similarity of resource features between the first network resource and the second network resource is determined to obtain the second matching degree between the first network resource and the target user; Select the second number of network resources with the highest sum of first and second matching degrees from the various first network resources, and use them as the network resources to be pushed.

4. The method according to claim 3, characterized in that, The second network resource is multiple. For each first network resource, determining the similarity of resource features between the first network resource and the second network resource to obtain a second matching degree between the first network resource and the target user includes: For each first network resource, the feature similarity between the resource features of the first network resource and each second network resource is determined, and the average value of the determined feature similarity is calculated to obtain the second matching degree between the first network resource and the target user.

5. The method according to claim 1, characterized in that, Before randomly selecting a first number of first network resources from a preset pool of candidate network resources, the method further includes: From the pre-established correspondence between the set of request counts and the set of proportion values, determine the proportion value corresponding to the baseline count as the target proportion value; wherein, in the correspondence, the number of requests and the proportion value are negatively correlated. The first number is determined based on the target ratio and the total number of candidate network resources.

6. The method according to claim 1, characterized in that, The step of selecting a second number of first network resources as network resources to be pushed, based on the matching degree between each first network resource and the user characteristics of the target user, includes: The maximum matching degree is determined from the matching degree between each primary network resource and the user characteristics of the target user; If the maximum determined matching degree is less than the preset matching degree threshold, return to the step of randomly selecting a first number of first network resources from the preset candidate network resources, until the maximum determined matching degree is not less than the preset matching degree threshold. From each of the currently selected first network resources, according to the matching degree between each first network resource and the user characteristics of the target user, select a second number of network resources as the network resources to be pushed.

7. The method according to claim 1, characterized in that, The second number is the total number of network resources that the target client's display interface can show.

8. A device for pushing network resources, characterized in that, The device includes: The count determination module is used to determine, upon receiving a network resource acquisition request sent by the target client, the number of times the network resource acquisition requests previously sent by the target client were received within a specified time period before the current time, and use this number as a baseline count. The first selection module is used to randomly select a first number of first network resources from a preset pool of candidate network resources; wherein the first number is negatively correlated with the baseline number of times. The second selection module is used to select a second number of first network resources as network resources to be pushed, based on the matching degree between each first network resource and the user characteristics of the target user; wherein, the target user is the user who logs into the target client; The push module is used to push resources to the target client according to the network resources to be pushed.

9. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other through the communication bus; Memory, used to store computer programs; A processor, when executing a program stored in memory, implements the steps of the method described in any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the method described in any one of claims 1-7.

Images