A content pushing method and device

Abstract

The application discloses a content pushing method and device; the application obtains full-amount objects of a content pushing task; the full-amount objects are extracted to obtain a plurality of experimental objects; task data is pushed to the experimental objects based on the content pushing task to obtain feedback information of the experimental objects on the task data, the task data being associated with the content pushing task; data labels of the experimental objects are determined according to the feedback information; a prediction model is trained based on the experimental objects and the data labels to obtain a trained prediction model; target objects of the content pushing task are determined from the full-amount objects through the trained prediction model, so that content pushing is performed through the target objects; and the application can improve content pushing efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of Internet, in particular to a content pushing method and device. BACKGROUND

[0002] In the Internet era, in addition to obtaining information through active search and the like, there are a large number of client-initiated pushed contents, such as advertisements on web pages, etc. In order to improve the return rate of the advertisements, the pushing objects can be selected when the advertisements are pushed. In the prior art, the object selection strategy is determined by professionals through professional knowledge and experience, the object is selected according to the object selection strategy, and finally the pushing object is determined and the content is pushed to the pushing object.

[0003] In the research and practice of the prior art, the present inventors have found that the prior art needs to consume a large amount of manpower, resulting in low content pushing efficiency. SUMMARY

[0004] The embodiments of the present application provide a content pushing method and device, which can improve the content pushing efficiency.

[0005] The embodiments of the present application provide a content pushing method, comprising:

[0006] obtaining full objects of a content pushing task;

[0007] extracting the full objects to obtain a plurality of experimental objects;

[0008] pushing task data associated with the content pushing task to the experimental objects based on the content pushing task to obtain feedback information of the experimental objects on the task data;

[0009] determining data labels of the experimental objects according to the feedback information;

[0010] training a prediction model based on the experimental objects and the data labels to obtain a trained prediction model;

[0011] determining a target object of the content pushing task from the full objects by the trained prediction model, so as to push content through the target object.

[0012] Correspondingly, the present application provides a content pushing device, comprising:

[0013] an obtaining module configured to obtain full objects of a content pushing task;

[0014] an extracting module configured to extract the full objects to obtain a plurality of experimental objects;

[0015] The pushing module is configured to push task data associated with the content pushing task to the experimental object based on the content pushing task, so as to obtain feedback information of the experimental object on the task data.

[0016] The label determining module is configured to determine a data label of the experimental object according to the feedback information.

[0017] The training module is configured to train a prediction model based on the experimental object and the data label, to obtain a trained prediction model.

[0018] The object determining module is configured to determine a target object of the content pushing task from the full-quantity objects by using the trained prediction model, so as to push content through the target object.

[0019] In some embodiments, the feedback information includes positive feedback information and negative feedback information, the data label includes a positive label and a negative label, and the label determining module includes a first determining submodule and a second determining submodule, wherein,

[0020] The first determining submodule is configured to determine the data label of the experimental object as the positive label when the feedback information of the experimental object is the positive feedback information.

[0021] The second determining submodule is configured to determine the data label of the experimental object as the negative label when the feedback information of the experimental object is the negative feedback information.

[0022] In some embodiments, the content pushing device further includes:

[0023] The reference module is configured to extract the full-quantity objects to obtain a plurality of reference objects.

[0024] The history module is configured to obtain historical feedback information of the experimental object.

[0025] The change module is configured to determine reference feedback change information of the reference objects, the reference feedback change information including a change degree of feedback information of the reference objects within a preset time period.

[0026] At this time, the label determining module includes a label determining submodule, wherein:

[0027] The label determining submodule is configured to determine the data label of the experimental object based on the historical feedback information, the reference feedback change information, and the feedback information.

[0028] In some embodiments, the label determining submodule includes a calculation unit and a determination unit, wherein,

[0029] The computing unit is configured to calculate experimental feedback change information of the experimental subject according to the historical feedback information, the reference feedback change information, and the feedback information.

[0030] The determining unit is configured to determine a data label of the experimental subject according to the experimental feedback change information of the experimental subject.

[0031] In some embodiments, the data label includes a positive label and a negative label, and the determining unit is specifically configured to:

[0032] When the experimental feedback change information of the experimental subject is greater than a preset threshold, the data label of the experimental subject is determined as the positive label.

[0033] When the experimental feedback change information of the experimental subject is not greater than the preset threshold, the data label of the experimental subject is determined as the negative label.

[0034] In some embodiments, the preset threshold includes a preset first threshold and a preset second threshold, the data label includes a positive label, a negative label, and a neutral label, and the determining unit is specifically configured to:

[0035] When the experimental feedback change information of the experimental subject is greater than or equal to the preset first threshold, the data label of the experimental subject is determined as the positive label.

[0036] When the experimental feedback change information of the experimental subject is less than the preset first threshold and greater than the preset second threshold, the data label of the experimental subject is determined as the neutral label.

[0037] When the experimental feedback change information of the experimental subject is less than or equal to the preset second threshold, the data label of the experimental subject is determined as the negative label.

[0038] In some embodiments, the object determining module includes a constructing submodule, an input submodule, and a determining submodule, wherein,

[0039] The constructing submodule is configured to perform feature construction on the full-quantity object to obtain a full-quantity feature vector.

[0040] The input submodule is configured to input the full-quantity feature vector into the trained prediction model to obtain label prediction information and a confidence thereof of the full-quantity object.

[0041] The determining submodule is configured to determine a target object of the content pushing task from the full-quantity object based on the label prediction information and the confidence thereof, so as to push content through the target object.

[0042] In some embodiments, the determining submodule is specifically configured to:

[0043] When the predicted label information of the full-quantity object is the target data label, the full-quantity object is determined as a candidate object;

[0044] The target object is determined from all the candidate objects based on the confidence of each candidate object.

[0045] In some embodiments, the training module is specifically configured to:

[0046] The feature construction is performed on the experimental object to obtain an experimental feature vector;

[0047] The model prediction result of the experimental object is obtained through the experimental feature vector and the prediction model;

[0048] The prediction model is trained based on the model prediction result and the data label to obtain a trained prediction model.

[0049] Correspondingly, the embodiments of the present application also provide a storage medium, which stores a computer program, and the computer program is suitable for being loaded by a processor to execute any one of the content pushing methods provided by the embodiments of the present application.

[0050] Correspondingly, the embodiments of the present application also provide a computer device, which comprises a memory, a processor and a computer program stored in the memory and capable of running on the processor, wherein the processor executes the computer program to implement any one of the content pushing methods provided by the embodiments of the present application.

[0051] The present application can obtain full-quantity objects of a content pushing task, extract the full-quantity objects to obtain a plurality of experimental objects, push task data to the experimental objects based on the content pushing task to obtain feedback information of the experimental objects on the task data, the task data being associated with the content pushing task, determine data labels of the experimental objects according to the feedback information, train a prediction model based on the experimental objects and the data labels to obtain a trained prediction model, and determine a target object of the content pushing task from the full-quantity objects through the trained prediction model to push content through the target object.

[0052] The present application can first extract experimental objects from full-quantity objects, push content to the experimental objects, determine data labels of the experimental objects according to feedback information of the experimental objects on the pushed task data, and then train a prediction model according to the experimental objects and the data labels, so as to finally determine a target pushing object from full-quantity data through the trained prediction model to push content, which can significantly improve the efficiency of content pushing, different from a highly dependent human mode. BRIEF DESCRIPTION OF DRAWINGS

[0053] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the description of the embodiments will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort.

[0054] Figure 1 is a scene schematic diagram of a content pushing system provided by an embodiment of the present application;

[0055] Figure 2 is a flow schematic diagram of a content pushing method provided by an embodiment of the present application;

[0056] Figure 3 is another flow schematic diagram of a content pushing method provided by an embodiment of the present application;

[0057] Figure 4 is a structure schematic diagram of a content pushing device provided by an embodiment of the present application;

[0058] Figure 5 is a structure schematic diagram of a computer device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0059] The technical solutions in the embodiments of the present application will be described clearly and completely with reference to the drawings in the embodiments of the present application. Obviously, the embodiments described in the present application are only some embodiments of the present application, not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present application.

[0060] The content pushing system can be integrated in a computer device, which can include at least one of a terminal and a server. The server can be a standalone physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server providing cloud computing services. The terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal and the server can be directly or indirectly connected through wired or wireless communication, which is not limited in the present application.

[0061] Referring to Figure 1The content pushing system can be integrated in a computer device such as a terminal or a server, and the computer device can obtain full objects of a content pushing task; the full objects are extracted to obtain a plurality of experimental objects; task data is pushed to the experimental objects based on the content pushing task to obtain feedback information of the experimental objects on the task data, the task data is associated with the content pushing task; data labels of the experimental objects are determined according to the feedback information; a prediction model is trained based on the experimental objects and the data labels to obtain a trained prediction model; and the target object of the content pushing task is determined from the full objects by using the trained prediction model, so that content pushing is performed through the target object.

[0062] It should be noted that Figure 1 The scenario diagram of the content pushing system shown is only an example, and the content pushing system and scenario described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. It is known to those skilled in the art that, as the content pushing device evolves and new business scenarios appear, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0063] The following will be described in detail. In this embodiment, a content pushing method will be described in detail, which can be integrated in a computer device such as a terminal or a server. Figure 2 As shown in the figure Figure 2 is a flowchart of the content pushing method provided by the embodiments of the present application. The content pushing method can include the following steps.

[0064] 101. Obtain full objects of a content pushing task.

[0065] The full objects include all potential objects that can be pushed with content, and the full objects usually include real users. In the case of testing and experiments, the full objects can also include virtual objects such as test scripts simulating users. The purpose of the present application is to determine target objects from the full objects and push content to the target objects. Therefore, the full objects can include objects that can be pushed with content, such as all registered accounts in a client that provides specific services such as purchase services or reading services. For example, in a content pushing task for a specific web page, the full objects can include all computer devices that have accessed the web page.

[0066] The full objects can be distinguished by full object identifiers such as characters, two-dimensional codes, and texts that can serve as unique identifiers, such as registered account information in a client, an Internet Protocol Address (IP) of a computer accessing a web page, and the like.

[0067] In the present application, the content pushing task includes pushing content to the target object determined from the full object. The content pushing can include various types, such as advertisement pushing, news pushing, activity information pushing, promotion pushing, and play recommendation pushing. The content pushing can be in the form of a pop-up window, a message, or a direct display in the gap of the content browsed by the user. The purpose of the content pushing is to make the object actively view the pushed content. However, in actual application scenarios, not every object that receives the pushed content will actively view it, and it can be directly ignored. At this time, the content pushing becomes an invalid behavior. In this case, the purpose of the content pushing task is to make the object view the content with a higher probability by selecting the target object that is more interested in the pushed content and has a higher possibility of viewing, so as to achieve the purpose of improving the content viewing probability.

[0068] Specifically, the manner of obtaining the full object can include various types, such as directly obtaining from a computer device integrated with the scheme of the present application, or sending a request to a computer device storing full data (such as a server, etc.), receiving the full object returned by the computer device according to the request, etc.

[0069] For example, the content pushing task can be advertisement pushing on a personal social page of a user using an instant messaging software. The account identifier of all accounts using the personal social page function in the instant messaging software can be obtained (i.e., the full object is obtained).

[0070] 102. Extracting the full object to obtain a plurality of experimental objects.

[0071] Specifically, the manner of extracting the full object to obtain the experimental object can include various types, such as random extraction, or grouping the full object according to the properties of the full object, extracting in each group, and finally integrating the extraction results of each group to obtain the experimental object. For example, the full object can be sampled multiple times within a period of time, and the results of each sampling are integrated to obtain the experimental object. The specific selection can be flexible in the application process, and will not be described here.

[0072] For example, the full object can be divided into several groups according to the age information carried by the account identifier, and randomly extracted in each group. Finally, the extraction results of all groups are integrated to obtain the experimental object.

[0073] 103. Pushing task data associated with the content pushing task to the experimental object based on the content pushing task to obtain feedback information of the experimental object to the task data.

[0074] The task data can include data of content that the content pushing task wants to push. For example, when the content pushing task is an advertisement pushing task, the task data can include an advertisement to be pushed. For another example, when the content pushing task is a voting pushing task, the task data can include voting information. In addition, the task data can also include questionnaire content, promotion content, and the like.

[0075] The feedback information can include reaction information of the experimental object based on the task data. The feedback information can include a staying time of the object on a page where the task data is located, a frequency of an operation pointed by content of the task data, and the like. For example, a frequency of using a service corresponding to advertisement content by the object, a staying time of the object on a page where the advertisement content is located, and the like.

[0076] The feedback information can include a response of the experimental object to the task data, such as clicking an advertisement, taking a coupon, purchasing a product, watching a video, filling out a questionnaire, clicking a link, clicking a message, and the like. The feedback information can also include ignoring of the task data by the experimental object, that is, when the experimental object does not respond to the task data in any way, it is recorded as a kind of feedback information.

[0077] Specifically, based on the content pushing task pushing the task data to the experimental object, the task data corresponding to the content pushing task can be determined, and then the task data is pushed to each experimental object respectively. According to a reaction of the experimental object to content displayed on a page based on the task data, or a reaction of the experimental object to a page containing the content within a period of time, feedback information of the experimental object to the task data is determined.

[0078] For example, the task data (advertisement data) corresponding to the content pushing task (advertisement task) is determined, the advertisement data is pushed to the experimental object, and feedback information of the experimental object to the advertisement data is obtained according to a reaction of the experimental object.

[0079] 104. According to the feedback information, a data tag of the experimental object is determined.

[0080] The data tag can include a representation of the feedback information of the experimental object. The data tag can include characters that are easy to distinguish and read. Different data tags refer to different feedback information. Specifically, the process of determining the data tag of the experimental object according to the feedback information can be flexible according to the difference of the feedback information. For example, the feedback information can be a continuous integer. An interval can be set. The data tag of the experimental object corresponding to the feedback information is determined according to an interval where the feedback information is located. For another example, the feedback information can be a limited number of simple texts. At this time, a data tag can be set for each text, and then the data tag of each experimental object is determined.

[0081] In some embodiments, the feedback information includes positive feedback information and negative feedback information, the data label includes a positive label and a negative label, and the step of "extracting the full amount of objects to obtain a plurality of experimental objects" can include:

[0082] When the feedback information of the experimental object is positive feedback information, the data label of the experimental object is determined to be a positive label; when the feedback information of the experimental object is negative feedback information, the data label of the experimental object is determined to be a negative label.

[0083] For example, when the content pushing task is an advertisement pushing task, the feedback information of the experimental object can be whether the advertisement is clicked. If an experimental object clicks the advertisement, the feedback information of the experimental object is positive feedback information, and if an experimental object does not click the advertisement, the feedback information of the experimental object is negative feedback information. Then, the data label of the experimental object can be determined according to the feedback information of the experimental object. When the feedback information of an experimental object is positive feedback information, the data label of the experimental object is a positive label; when the feedback information of an experimental object is negative feedback information, the data label of the experimental object is a negative label.

[0084] In some embodiments, the content pushing method further includes:

[0085] extracting the full amount of objects to obtain a plurality of reference objects; obtaining historical feedback information of the experimental object; determining reference feedback change information of the reference objects, the reference feedback change information including a change degree of feedback information of the reference objects within a preset time period;

[0086] At this time, the step of "determining the data label of the experimental object according to the feedback information" can include:

[0087] determining the data label of the experimental object based on the historical feedback information, the reference feedback change information, and the feedback information.

[0088] In some embodiments, determining the data label of the experimental object needs to be based on the historical feedback information and the feedback information of the experimental object, and the reference feedback change information of the reference object,

[0089] The reference object can include a part of the full amount of objects, and the reference object can correspond to the experimental object, similar to the experimental group and the control group. The reference object can exclude the interference of factors other than content pushing on the determination of the data label to a certain extent.

[0090] The extraction manner of the reference objects can be similar to the extraction manner of the experimental objects, which will not be described herein. The present application extracts the reference objects and the experimental objects from the full objects according to a certain proportion. The proportion can be determined according to the number of the full objects, the number of the experimental objects and the number of the reference objects, and the like. For example, the proportion can be 1%. In addition, the number of the reference objects and the number of the experimental objects can be flexibly determined based on the training requirements of the prediction model and the like. For example, the number of the reference objects and the number of the experimental objects can be set to 100,000.

[0091] For example, the content pushing is to push an advertisement on an information stream page (the information stream can include a video, a text, an image, and the like), an advertisement can be pushed to the information stream page of the experimental object within a preset time period, and the average stay time (i.e., feedback information) of the object on the information stream page after the advertisement is pushed is recorded. The information stream page of the reference object is not pushed with an advertisement, and the average stay time of the reference object on the information stream page at the starting stage and the ending stage of the preset time period is recorded, respectively, and then the reference feedback change information of the reference object is obtained.

[0092] The reference feedback change information includes the change degree of the feedback information of the reference object within the preset time period. The calculation manner of the reference feedback change information can include multiple manners. For example, the average stay time at the ending stage is directly divided by the average stay time at the starting stage. For another example, the content pushing task can be to push a promotion information of a shopping website, and the reference feedback change information can be the difference between the average order number of the object at the ending stage and the average order number of the object at the starting stage. In different content pushing tasks, the calculation manner and the expression form of the reference feedback change information will change accordingly, which can be flexibly determined according to the actual situation in the application scene, which will not be limited herein.

[0093] The historical feedback information can include the reaction of the experimental object before the task data is pushed. For example, the historical feedback information can include the average stay time of the experimental object on the information stream page before the advertisement is pushed to the information stream page. For another example, the historical feedback information can include the average order number of the experimental object before the promotion information is pushed, and the like.

[0094] The process of determining the data label according to the historical feedback information and the feedback information of the experimental object and the reference feedback change information of the reference object can include multiple manners. For example, the fusion manner such as weighted average, difference value, and sum can be used, or the ranking and comparison manner can be used.

[0095] In some embodiments, the step of “determining the data label of the experimental object based on the historical feedback information, the reference feedback change information, and the feedback information” can include:

[0096] According to the historical feedback information, the feedback information, and the reference feedback change information, the experimental feedback change information of the experimental object is calculated; and the data label of the experimental object is determined through the experimental feedback change information.

[0097] The experimental feedback change information can include the change degree of the feedback information of the experimental object in a preset time period, and the preset time period can be the period during which the task is pushed. Specifically, to determine the data label of the experimental object, the initial feedback change information of the experimental object can be calculated first. The calculation method of the initial feedback change information is similar to that of the reference feedback change information. For example, the feedback information can be divided by the historical feedback information, or the feedback information can be subtracted by the historical feedback information, etc. Specifically, it can be flexibly set in actual application. Then, based on the initial feedback change information and the reference feedback change information, the experimental feedback change information of the experimental object is determined. For example, the size of the initial feedback change information and the reference feedback change information of the experimental object can be compared, and the experimental feedback change information is obtained according to the comparison result. For another example, the initial feedback change information and the reference feedback change information can be weighted and summed to obtain the experimental feedback change information, etc.

[0098] Finally, the data label of the experimental object is determined through the experimental feedback change information. For example, the data label corresponding to the experimental feedback change information of the experimental object can be determined according to the mapping relationship.

[0099] For example, the initial feedback change information C of the experimental object 11 can be determined according to the historical feedback information L and the feedback information F. Then, the size of the initial feedback change information C of the experimental object 11 and the reference feedback change information K is compared, and the experimental feedback change information S of the experimental object 11 is determined according to the comparison result, and the data label of the experimental object 11 is determined as the data label 1.

[0100] In some embodiments, the data label includes a positive label and a negative label. The step of “determining the data label of the experimental object according to the experimental feedback change information of the experimental object” can include:

[0101] When the experimental feedback change information of the experimental object is greater than a preset threshold, the data label of the experimental object is determined as a positive label. When the experimental feedback change information of the experimental object is not greater than the preset threshold, the data label of the experimental object is determined as a negative label.

[0102] For example, the experimental feedback change information of the experimental subject obtained by calculation can be a real number. When determining the data label, the size relationship between the experimental feedback change information of the experimental subject and the preset threshold value can be compared, and the data label of the experimental subject can be obtained according to the comparison result. For example, the preset threshold value can be 0. If the experimental feedback change information of an experimental subject is greater than 0, it can be determined that the data label of the experimental subject is a positive label. If the experimental feedback change information of an experimental subject is less than or equal to 0, it can be determined that the data label of the experimental subject is a negative label. The positive label and the negative label can be identified by characters, such as the positive label can be identified as 1, and the negative label can be identified as 0.

[0103] In some embodiments, the preset threshold value includes a preset first threshold value and a preset second threshold value, the data label includes a positive label, a negative label and a neutral label, and the step of "determining the data label of the experimental subject according to the experimental feedback change information of the experimental subject" can include:

[0104] When the experimental feedback change information of the experimental subject is greater than or equal to the preset first threshold value, the data label of the experimental subject is determined to be a positive label. When the experimental feedback change information of the experimental subject is less than the preset first threshold value and greater than the preset second threshold value, the data label of the experimental subject is determined to be a neutral label. When the experimental feedback change information of the experimental subject is less than or equal to the preset second threshold value, the data label of the experimental subject is determined to be a negative label.

[0105] For example, the experimental feedback change information of the experimental subject can be a real number. When determining the data label, the size relationship between the experimental feedback change information of the experimental subject and the preset threshold value can be compared, and the data label of the experimental subject can be obtained according to the comparison result. The preset threshold value can include multiple, and the data label can also include multiple. For example, the preset threshold value can include a preset first threshold value and a preset second threshold value, and the data label can include a positive label, a negative label and a neutral label. For example, the preset first threshold value and the preset second threshold value can be 1 and -1 respectively. When the experimental feedback change information of an experimental subject is greater than or equal to 1, it can be determined that the data label of the experimental subject is a positive label. When the experimental feedback change information of an experimental subject is less than 1 and greater than -1, it can be determined that the data label of the experimental subject is a neutral label. When the experimental feedback change information of an experimental subject is less than or equal to -1, it can be determined that the data label of the experimental subject is a negative label.

[0106] 105、based on the experimental subject and the data label, training the prediction model to obtain a trained prediction model.

[0107] The prediction model can include a model capable of predicting the reaction of the experimental object to the content push. The prediction model can include a classification model. Common classification models can include a tree model, a linear regression model, a gradient boosting model (such as an extreme gradient boosting model (XGBoost)), a random forest model, a long short-term memory network model (LSTM), a neural network, and the like. In practice, the specific model can be flexibly selected, and the present disclosure is not limited in this regard.

[0108] For example, the prediction model can be a gradient boosting model. The prediction model can be trained according to the experimental object and the data label to obtain a trained prediction model A.

[0109] In some embodiments, the step of training the prediction model based on the experimental object and the data label to obtain a trained prediction model can include:

[0110] The experimental object is subjected to feature construction to obtain an experimental feature vector. The experimental feature vector and the prediction model are used to obtain a model prediction result of the experimental object. The prediction model is trained based on the model prediction result and the data label to obtain a trained prediction model.

[0111] Specifically, during the feature construction, the experimental object identifier, the content prediction task, and the like of the experimental object can be flexibly performed. For example, common processes of the feature construction can include binning, one-hot encoding, hashing trick, embedding, log transformation, scaling, normalization, feature interaction, and the like.

[0112] The experimental feature vector can be input into the prediction model to obtain a model prediction result of the experimental object. The loss value of the current training can be calculated according to the data label of the experimental object and the model prediction result. The parameters of the prediction model are adjusted to realize the training process to obtain a trained prediction model. Specifically, the algorithm for adjusting the parameters can include a plurality of algorithms, such as a stochastic gradient descent (SGD), a momentum SGD, an adaptive gradient (AdaGrad), and the like.

[0113] 106. determining, by the trained prediction model, a target object from the full set of objects for the content pushing task to be performed on the target object.

[0114] For example, the full set of objects other than the experimental objects can be determined as target full set of objects, and the target full set of objects can be sequentially input into the trained prediction model to obtain output information of each target full set of object, and then the target object can be selected from the target full set of objects according to the output information, and the content pushing task can be performed on the target object.

[0115] In some embodiments, the step of "determining, by the trained prediction model, a target object from the full set of objects for the content pushing task to be performed on the target object" can include:

[0116] performing feature construction on the full set of objects to obtain a full set of feature vectors, inputting the full set of feature vectors into the trained prediction model to obtain label prediction information and confidence thereof of the full set of objects, and determining a target object from the full set of objects for the content pushing task to be performed on the target object based on the label prediction information and the confidence thereof.

[0117] Before the full set of objects is input into the trained prediction model, feature construction needs to be performed on the full set of objects, and the manner and steps of feature construction are similar to those of the experimental objects, which will not be described here. After the feature construction on the full set of objects is performed to obtain a full set of feature vectors, the full set of feature vectors can be input into the trained prediction model to obtain label prediction information and confidence of the full set of objects.

[0118] The label prediction information can include a prediction of a data label of the global object, and the data label is determined by feedback information. The label prediction information / data label represents the performance of the corresponding full set of object in the content pushing task (such as clicking on an advertisement, increasing the number of purchases, etc.). The confidence can include the credibility of the label prediction information. The higher the confidence, the higher the credibility of the label prediction information.

[0119] In some embodiments, the step of "determining a target object from the full set of objects for the content pushing task based on the label prediction information and the confidence thereof" can include:

[0120] When the predicted label information of the full set of objects is a target data label, the full set of objects is determined as a candidate object. The target object is determined from all candidate objects based on the confidence of each candidate object.

[0121] For example, the target data label can be data label 1, and the full set of objects with data label 1 is retained as a candidate object. Each candidate object is ranked according to the confidence of the label prediction information, and then the target object is determined from the candidate objects according to the ranking information.

[0122] The present application can first extract the experimental object from the full quantity object, and push the content to the experimental object, determine the data label of the experimental object according to the feedback information of the experimental object to the pushed task data, and then train the prediction model according to the experimental object and the data label of the experimental object. Finally, the target pushing object is determined from the full quantity data through the trained prediction model to push the content, which can significantly improve the efficiency of content pushing, which is different from the highly dependent way.

[0123] According to the method described in the above embodiment, the following will be further illustrated by example.

[0124] The present application will take the content pushing system integrated in the computer equipment as an example to introduce the content pushing method, as shown in Figure 3 Figure 3 is a flowchart of the content pushing method provided by the embodiment of the present application. The content pushing method can include:

[0125] 201, the computer equipment obtains the full quantity object of the content pushing task.

[0126] For example, the content pushing task can be an e-commerce advertisement pushing task, the full quantity object can be all users who have used the e-commerce, and the computer equipment can read the full quantity users of the e-commerce advertisement pushing task from the database.

[0127] 202, the computer equipment extracts the full quantity object to obtain a plurality of experimental objects.

[0128] For example, the computer equipment can randomly extract 1% from all users who have used the e-commerce as experimental objects.

[0129] 203, the computer equipment extracts the full quantity object to obtain a plurality of reference objects.

[0130] For example, the computer equipment can randomly extract 1% from all users who have used the e-commerce as reference objects.

[0131] 204, the computer equipment pushes the task data to the experimental object based on the content pushing task to obtain the feedback information of the experimental object, and the task data is associated with the content pushing task.

[0132] For example, the computer equipment can push the e-commerce advertisement (i.e. task data) to the experimental object, and the e-commerce advertisement can be presented to the experimental object in the form of text, image, animation, video, etc. The specific presentation mode can include displaying the e-commerce advertisement on the page frequently browsed by the experimental object, and then obtaining the feedback information of the experimental object. The feedback information can be the purchase frequency of the experimental object within one week on the e-commerce after the e-commerce advertisement is pushed to the experimental object, wherein the e-commerce advertisement is determined according to the e-commerce advertisement pushing task.

[0133] ​205、The computer device obtains historical feedback information of the experimental object, and determines reference feedback change information of the reference object.

[0134] For example, the computer device obtains the historical feedback information of the experimental object before performing e-commerce advertisement pushing on the experimental object, and the historical feedback information includes the average weekly purchase frequency (average value of weekly purchase frequency) of the experimental object on the e-commerce platform.

[0135] The reference feedback change information can reflect the change degree of the feedback information of the reference object during the period of pushing e-commerce advertisements to the experimental object. For example, the computer device obtains the weekly average purchase frequency 1 of the reference object before performing e-commerce advertisement pushing on the experimental object, and the weekly average purchase frequency 2 of the reference object after performing e-commerce advertisement pushing on the experimental object, and then obtains the reference feedback change information 1 of the reference object 1 by subtracting the weekly average purchase frequency 1 from the weekly average purchase frequency 2. The reference feedback change information X of the reference object is obtained by averaging the reference feedback change information of all reference objects.

[0136] 206、The computer device determines the data label of the experimental object based on the historical feedback information, the reference feedback change information, and the feedback information.

[0137] For example, the data label includes a positive label (which can be marked by the number 1) and a negative label (which can be marked by the number 2). The computer device calculates the experimental feedback change information of the experimental object based on the historical feedback information, the reference feedback change information, and the feedback information, and compares the experimental feedback change information with a set threshold value. If the experimental feedback change information is greater than the set threshold value, the data label of the experimental object is determined to be 1, and if the experimental feedback change information is not greater than the set threshold value, the data label of the experimental object is determined to be 2.

[0138] 207、The computer device trains the prediction model based on the experimental object and the data label, and obtains a trained prediction model.

[0139] 208、The computer device determines the target object of the content pushing task from the full object through the trained prediction model, so as to push the content to the target object.

[0140] For example, the trained prediction model outputs the label prediction information and the confidence of each full object. The full object with the data label of 1 is retained as a candidate object. Each candidate object is ranked according to the confidence of the label prediction information, and then the target object is determined from the candidate object according to the ranking information, and the e-commerce advertisement is pushed to the target object.

[0141] The present application can first extract the experimental object from the full quantity object, and push the content to the experimental object, determine the data label of the experimental object according to the feedback information of the experimental object to the pushed task data, and then train the prediction model according to the experimental object and the data label of the experimental object. Finally, the target pushing object is determined from the full quantity data through the trained prediction model to push the content, which can significantly improve the efficiency of content pushing, unlike the highly dependent way of people.

[0142] In order to better implement the content pushing method provided by the embodiments of the present application, the embodiments of the present application also provide a device based on the above-mentioned content pushing method. The meanings of the terms are the same as those in the above-mentioned content pushing method, and the specific implementation details can be referred to the description in the method embodiment.

[0143] As shown in Figure 4 , Figure 4 The content pushing device provided by an embodiment of the present application has the structure as shown in the figure, which can include a sample acquisition module 301, an extraction module 302, a pushing module 303, a label determination module 304, a training module 305 and an object determination module 306, wherein:

[0144] The acquisition module 301 is used to acquire the full quantity object of the content pushing task;

[0145] The extraction module 302 is used to extract the full quantity object to obtain a plurality of experimental objects;

[0146] The pushing module 303 is used to push the task data to the experimental object based on the content pushing task, so as to obtain the feedback information of the experimental object to the task data, and the task data is associated with the content pushing task;

[0147] The label determination module 304 is used to determine the data label of the experimental object according to the feedback information;

[0148] The training module 305 is used to train the prediction model based on the experimental object and the data label, and obtain the trained prediction model;

[0149] The object determination module 306 is used to determine the target object of the content pushing task from the full quantity object through the trained prediction model, so as to push the content through the target object.

[0150] In some embodiments, the feedback information includes positive feedback information and negative feedback information, the data label includes positive label and negative label, and the label determination module includes a first determination sub-module and a second determination sub-module, wherein,

[0151] The first determination sub-module is used to determine the data label of the experimental object as the positive label when the feedback information of the experimental object is the positive feedback information;

[0152] The second determining sub-module is configured to determine that the data label of the experimental object is a negative label when the feedback information of the experimental object is negative feedback information.

[0153] In some embodiments, the content pushing apparatus further comprises:

[0154] The reference module is configured to extract the full-amount object to obtain a plurality of reference objects.

[0155] The history module is configured to obtain historical feedback information of the experimental object.

[0156] The change module is configured to determine reference feedback change information of the reference objects, the reference feedback change information comprising a change degree of feedback information of the reference objects within a preset time period.

[0157] At this time, the label determining module comprises a label determining sub-module, wherein:

[0158] The label determining sub-module is configured to determine the data label of the experimental object based on the historical feedback information, the reference feedback change information, and the feedback information.

[0159] In some embodiments, the label determining sub-module comprises a calculation unit and a determination unit, wherein:

[0160] The calculation unit is configured to calculate experimental feedback change information of the experimental object according to the historical feedback information, the reference feedback change information, and the feedback information.

[0161] The determination unit is configured to determine the data label of the experimental object according to the experimental feedback change information of the experimental object.

[0162] In some embodiments, the data label comprises a positive label and a negative label, and the determination unit is specifically configured to:

[0163] determine that the data label of the experimental object is the positive label when the experimental feedback change information of the experimental object is greater than a preset threshold value;

[0164] determine that the data label of the experimental object is the negative label when the experimental feedback change information of the experimental object is not greater than the preset threshold value.

[0165] In some embodiments, the preset threshold value comprises a preset first threshold value and a preset second threshold value, the data label comprises a positive label, a negative label, and a neutral label, and the determination unit is specifically configured to:

[0166] determine that the data label of the experimental object is the positive label when the experimental feedback change information of the experimental object is greater than or equal to the preset first threshold value;

[0167] When the experimental feedback change information of the experimental subject is less than the preset first threshold and greater than the preset second threshold, it is determined that the data label of the experimental subject is a neutral label;

[0168] When the experimental feedback change information of the experimental subject is less than or equal to the preset second threshold, it is determined that the data label of the experimental subject is a negative label.

[0169] In some embodiments, the object determining module includes a construction submodule, an input submodule, and a determination submodule, wherein,

[0170] The construction submodule is configured to perform feature construction on the full-quantity objects to obtain full-quantity feature vectors;

[0171] The input submodule is configured to input the full-quantity feature vectors into the trained prediction model to obtain label prediction information and confidence thereof of the full-quantity objects;

[0172] The determination submodule is configured to determine a target object of the content pushing task from the full-quantity objects based on the label prediction information and the confidence thereof, so as to perform content pushing through the target object.

[0173] In some embodiments, the determination submodule is specifically configured to:

[0174] When the prediction label information of the full-quantity object is the target data label, it is determined that the full-quantity object is a candidate object;

[0175] Based on the confidence of each candidate object, the target object is determined from all the candidate objects.

[0176] In some embodiments, the training module is specifically configured to:

[0177] Perform feature construction on the experimental objects to obtain experimental feature vectors;

[0178] Obtain model prediction results of the experimental objects through the experimental feature vectors and the prediction model;

[0179] Train the prediction model based on the model prediction results and the data labels to obtain the trained prediction model.

[0180] In the present application, the obtaining module 301 obtains the full quantity object of the content pushing task, the extracting module 302 extracts the full quantity object to obtain a plurality of experimental objects, the pushing module 303 pushes task data to the experimental objects based on the content pushing task to obtain feedback information of the experimental objects on the task data, the task data is associated with the content pushing task, then the label determining module 304 determines the data label of the experimental object according to the feedback information, the training module 305 trains the prediction model based on the experimental object and the data label to obtain the trained prediction model, and finally the object determining module 306 determines the target object of the content pushing task from the full quantity object through the trained prediction model to push the content through the target object.

[0181] The present application can first extract the experimental object from the full quantity object, push the content to the experimental object, determine the data label of the experimental object according to the feedback information of the experimental object on the pushed task data, and then train the prediction model according to the experimental object and the data label of the experimental object, and finally determine the target pushing object from the full quantity data through the trained prediction model to push the content, which can significantly improve the efficiency of content pushing compared with the highly dependent human mode.

[0182] In addition, the present application also provides a computer device, which can be a terminal or a server, as shown in Figure 5 The computer device structure related to the embodiments of the present application is shown, and specifically:

[0183] The computer device can include a processor 401 with one or more processing cores, a memory 402 with one or more computer readable storage media, a power supply 403, an input unit 404, and the like. Those skilled in the art can understand that the computer device structure shown in Figure 5 The computer device structure shown in the present application does not constitute a limitation on the computer device, and can include more or fewer components than shown, or combine certain components, or different component arrangements. Among them:

[0184] The processor 401 is the control center of the computer device, which connects all parts of the computer device through various interfaces and lines, executes the software programs and / or modules stored in the memory 402 and calls the data stored in the memory 402, performs various functions and processes data of the computer device, and thus performs overall detection of the computer device. Optionally, the processor 401 can include one or more processing cores; preferably, the processor 401 can integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user pages and application programs, etc., and the modem processor mainly processes wireless communication. It can be understood that the above-mentioned modem processor can also not be integrated into the processor 401.

[0185] The memory 402 can be used to store software programs and modules, and the processor 401 executes various function applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, application programs required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; and the data storage area can store data created according to the use of the computer device, etc. In addition, the memory 402 can include a high-speed random access memory, and can also include a non-volatile memory, such as at least one magnetic disk storage device, a flash memory device, or other volatile solid-state memory device. Accordingly, the memory 402 can also include a memory controller to provide the processor 401 with access to the memory 402.

[0186] The computer device further includes a power supply 403 for supplying power to various components. Preferably, the power supply 403 can be logically connected to the processor 401 through a power management system, so as to realize functions such as management of charging, discharging, and power consumption management through the power management system. The power supply 403 can also include one or more than one direct current or alternating current power supply, a recharging system, a power supply fault detection circuit, a power supply converter or inverter, a power supply state indicator, and the like.

[0187] The computer device can further include an input unit 404, which can be used to receive input digital or character information, and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function controls.

[0188] Although not shown, the computer device can also include a display unit and the like, which will not be described here. Specifically, in the present embodiment, the processor 401 in the computer device loads one or more than one executable file corresponding to the process of an application program into the memory 402 according to the following instructions, and runs the application program stored in the memory 402 by the processor 401, so as to realize various functions, as follows:

[0189] obtain a full object of a content pushing task; extract the full object to obtain a plurality of experimental objects; push task data associated with the content pushing task to the experimental objects based on the content pushing task to obtain feedback information of the experimental objects on the task data; determine a data label of the experimental object according to the feedback information; train a prediction model based on the experimental object and the data label to obtain a trained prediction model; and determine a target object of the content pushing task from the full object by using the trained prediction model, so as to push content through the target object.

[0190] The specific implementation of each operation can refer to the foregoing embodiments, which will not be described here.

[0191] According to an aspect of the present application, a computer program product or computer program is provided, which includes computer instructions stored in a computer readable storage medium. A processor of a computer device reads the computer instructions from the computer readable storage medium, and the processor executes the computer instructions, so that the computer device performs the method provided in various optional implementations of the above embodiments.

[0192] Those skilled in the art can understand that all or part of the steps in the various methods of the above embodiments can be completed by a computer program, or by relevant hardware controlled by a computer program, which can be stored in a computer readable storage medium and loaded and executed by a processor.

[0193] To this end, the embodiments of the present application further provide a storage medium, which stores a computer program capable of being loaded by a processor to execute the steps in any of the content pushing methods provided by the embodiments of the present application. For example, the computer program can execute the following steps:

[0194] obtaining a full object of a content pushing task; extracting the full object to obtain a plurality of experimental objects; pushing task data associated with the content pushing task to the experimental objects based on the content pushing task to obtain feedback information of the experimental objects on the task data; determining a data label of the experimental object according to the feedback information; training a prediction model based on the experimental object and the data label to obtain a trained prediction model; and determining a target object of the content pushing task from the full object by using the trained prediction model, so as to push content through the target object.

[0195] The storage medium can include a read only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0196] Since the computer program stored in the storage medium can execute the steps in any of the content pushing methods provided by the embodiments of the present application, the beneficial effects of any of the content pushing methods provided by the embodiments of the present application can be achieved, which are described in detail in the foregoing embodiments and will not be described here.

[0197] The above describes in detail the content pushing method and device provided by the embodiments of the present application. The principles and implementation manners of the present application are described by using specific examples. The above description of the embodiments is only used to help understand the method of the present application and its core idea. Meanwhile, for those skilled in the art, the specific implementation manners and application ranges will be changed according to the idea of the present application. In conclusion, the content of the specification should not be understood as a limitation of the present application.

Claims

1. A content push method characterized by, include: Retrieve the full object of the content push task; Multiple experimental objects were obtained by extracting from the full set of objects; Based on the content push task, task data is pushed to the experimental subjects to obtain feedback information from the experimental subjects on the task data. The task data is associated with the content push task. Extracting from the full set of objects yields multiple reference objects; Obtain the historical feedback information of the experimental subjects; Determine the reference feedback change information of the reference object, wherein the reference feedback change information includes the degree of change of the feedback information of the reference object within a preset time period; Determining the data label of the experimental subject based on the feedback information includes: calculating the experimental feedback change information of the experimental subject based on the historical feedback information of the experimental subject, the reference feedback change information of the reference subject, and the feedback information of the experimental subject; and determining the data label of the experimental subject based on the experimental feedback change information. Based on the experimental objects and the data labels, the prediction model is trained to obtain the trained prediction model; The method involves using a trained prediction model to determine the target object for the content push task from the full set of objects, and then pushing content through the target object. This includes: constructing features from the full set of objects to obtain a full set of feature vectors; inputting the full set of feature vectors into the trained prediction model to obtain the label prediction information and confidence scores of the full set of objects; and determining the target object for the content push task from the full set of objects based on the label prediction information and confidence scores, so as to push content through the target object.

2. The method according to claim 1, characterized in that, The feedback information includes positive feedback information and negative feedback information, and the data labels include positive labels and negative labels. Determining the data labels of the experimental subjects based on the feedback information includes: When the feedback information from the experimental subject is positive, the data label of the experimental subject is determined to be a positive label; When the feedback information from the experimental subject is negative, the data label of the experimental subject is determined to be a negative label.

3. The method according to claim 1, characterized in that, The data labels include positive and negative labels. Determining the data labels of the experimental subjects based on their experimental feedback changes includes: When the experimental feedback change information of the experimental object is greater than a preset threshold, the data label of the experimental object is determined to be a positive label; When the experimental feedback change information of the experimental subject is not greater than the preset threshold, the data label of the experimental subject is determined to be a negative label.

4. The method according to claim 3, characterized in that, The preset thresholds include a first preset threshold and a second preset threshold; the data labels include positive labels, negative labels, and neutral labels; and determining the data labels of the experimental subjects based on their experimental feedback change information includes: When the experimental feedback change information of the experimental subject is greater than or equal to a preset first threshold, the data label of the experimental subject is determined to be a positive label. When the experimental feedback change information of the experimental subject is less than the preset first threshold and greater than the preset second threshold, the data label of the experimental subject is determined to be a neutral label; When the experimental feedback change information of the experimental subject is less than or equal to the preset second threshold, the data label of the experimental subject is determined to be a negative label.

5. The method according to claim 1, characterized in that, The step of determining the target object for the content push task from the full set of objects based on the tag prediction information and its confidence level includes: When the predicted label information of all objects is the target data label, the all objects are determined to be candidate objects; The target object is determined from all candidates based on the confidence level of each candidate object.

6. The method according to any one of claims 1 to 5, characterized in that, The step of training the prediction model based on the experimental objects and the data labels to obtain the trained prediction model includes: The experimental object is used to construct features to obtain an experimental feature vector; The model prediction results for the experimental object are obtained using the experimental feature vector and the prediction model. Based on the model prediction results and the data labels, the prediction model is trained to obtain the trained prediction model.

7. A content push device, characterized in that, include: The acquisition module is used to obtain the full object of the content push task; The extraction module is used to extract multiple experimental objects from the full set of objects. The push module is used to push task data to the experimental subject based on the content push task, so as to obtain the experimental subject's feedback information on the task data, wherein the task data is associated with the content push task; The reference module is used to extract multiple reference objects from the full set of objects. The history module is used to obtain historical feedback information of the experimental subjects; A change module is used to determine the reference feedback change information of the reference object, wherein the reference feedback change information includes the degree of change of the feedback information of the reference object within a preset time period; The label determination module is used to determine the data labels of the experimental subjects based on the feedback information. The label determination module includes a label determination submodule, which is used to calculate the experimental feedback change information of the experimental object based on the historical feedback information of the experimental object, the reference feedback change information of the reference object, and the feedback information of the experimental object; and to determine the data label of the experimental object based on the experimental feedback change information. The training module is used to train the prediction model based on the experimental object and the data label to obtain the trained prediction model. The object determination module is used to determine the target object of the content push task from the full set of objects through a trained prediction model, so as to push content through the target object; The object determination module includes a construction submodule, an input submodule, and a determination submodule, wherein... The construction submodule is used to construct features from the full set of objects to obtain a full set of feature vectors; The input submodule is used to input the full feature vector into the trained prediction model to obtain the label prediction information and confidence of the full set of objects; The determination submodule is used to determine the target object of the content push task from the full set of objects based on the tag prediction information and its confidence level, so as to push content through the target object.

8. A storage medium, characterized in that, The storage medium stores a computer program adapted for loading by a processor to perform the method according to any one of claims 1-6.

9. A computer device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method of any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes computer instructions stored in a computer-readable storage medium; the processor of the computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the method described in any one of the modified claims 1-6.

Images