CN114765624B

CN114765624B - Information recommendation method, device, server and storage medium

Info

Publication number: CN114765624B
Application number: CN202011638439.9A
Authority: CN
Inventors: 金雅然; 曾立群; 翟思楠; 马奕潇
Original assignee: Beijing Dajia Internet Information Technology Co Ltd
Current assignee: Beijing Dajia Internet Information Technology Co Ltd
Priority date: 2020-12-31
Filing date: 2020-12-31
Publication date: 2024-04-30
Anticipated expiration: 2040-12-31
Also published as: CN114765624A

Abstract

The disclosure discloses an information recommendation method, an information recommendation device, a server and a storage medium, and belongs to the technical field of communication. The information recommendation method comprises the following steps: receiving a recommendation request sent by a first account of a login application program, wherein the recommendation request is used for requesting to acquire multimedia information; determining a target account group corresponding to the first account in response to the recommendation request, wherein the target account group comprises a second account, the similarity between the first account information of the first account and the second account information of the second account is greater than or equal to a preset similarity, and the first similarity is determined by the social relationship information of the first account and the second account; acquiring the category of recommendation information corresponding to the state information according to the state information of the target account group, wherein the state information is used for representing and predicting the state of the application program used by the second account; multimedia information corresponding to the category of the recommended information is recommended to the first account. The information recommending method, the device, the server and the storage medium are adopted to at least solve the problem of inaccurate information recommendation at present.

Description

Information recommendation method, device, server and storage medium

Technical Field

The disclosure relates to the technical field of communication, and in particular relates to an information recommendation method, an information recommendation device, a server and a storage medium.

Background

With the continuous development of internet technology, electronic devices rely on application programs to provide services to users. Some applications can conduct targeted recommended services to the account according to the current use condition of the account.

However, because the current information recommendation method is based on a relatively fixed account group such as an old account group, inaccurate information recommendation for new accounts is caused.

Disclosure of Invention

An objective of the disclosed embodiments is to provide an information recommendation method, an information recommendation device, a server and a storage medium, so as to at least solve the problem of inaccurate information recommendation at present.

The technical scheme of the present disclosure is as follows:

according to a first aspect of embodiments of the present disclosure, there is provided an information recommendation method, which may include:

receiving a recommendation request sent by a first account of a login application program, wherein the recommendation request is used for requesting to acquire multimedia information;

Determining a target account group corresponding to the first account in response to the recommendation request, wherein the target account group comprises a second account, the similarity between the first account information of the first account and the second account information of the second account is greater than or equal to a preset similarity, and the first similarity is determined by the social relationship information of the first account and the second account;

Acquiring the category of recommendation information corresponding to the state information according to the state information of the target account group, wherein the state information is used for representing and predicting the state of the application program used by the second account;

multimedia information corresponding to the category of the recommended information is recommended to the first account.

In one possible embodiment, before the step of determining the target account group corresponding to the first account, the method may further include:

acquiring account information of a plurality of accounts corresponding to the application program;

Grouping the accounts according to the similarity of the account information to obtain a plurality of account groups; wherein,

The plurality of account groups includes a target account group.

In another possible embodiment, the step of determining the target account group corresponding to the first account may include:

Matching the first account information with account information of accounts in each account group to obtain similarity between the first account and each account group;

if the similarity between the first account and the account group is greater than the preset similarity, the account group is determined to be a target account group.

In still another possible embodiment, before the step of obtaining the category of the recommended information corresponding to the status information according to the status information of the target account group, the method may further include: when the second account comprises a fourth account, and the fourth account is the second account with the frequency of accessing the application program in the target account group being greater than or equal to the preset frequency, according to the access information of each second account in the target account group for accessing the application program, historical access information of the target account group is summarized;

Calculating the similarity of the first behavior information of the first account access application program and the second behavior information of the fourth account access application program to obtain the target similarity of the first account and the target account group;

obtaining proportion information of a target account group, wherein the proportion information is used for representing the proportion of a second account in the target account group in a platform account in a preset time period;

status information of account access applications in the target account group is determined based on the historical access information, the target similarity, and the scale information.

In still another possible embodiment, the step of determining the state information of the account access application in the target account group based on the history access information, the target similarity and the scale information may include:

Extracting target historical access information in a preset time period from the historical access information;

obtaining forecast proportion information of a target account group, wherein the forecast proportion information is used for representing the proportion of a second account in the target account group in a platform account in a forecast preset time period;

and determining state information according to the target historical access information, the predicted proportion information, the target similarity and the proportion information.

Based on this, the step of determining the state information according to the target historical access information, the predicted access proportion information, the target similarity and the proportion information may include:

Z＝1+1/B1*D*(1-T)*B2

Wherein z is state information, B1 is proportion information, D is target similarity, T is target history access information, and B2 is prediction proportion information.

In still another possible embodiment, the step of obtaining the category of the recommended information corresponding to the status information according to the status information of the target account group may include:

Determining a target account group category to which the target account group belongs according to the state information of the target account group;

and acquiring the category of the recommendation information corresponding to the target account group category from the preset associated information of the account group category and the category of the recommendation information.

In still another possible embodiment, the step of determining the target account group category to which the target account group belongs according to the state information of the target account group may include:

And under the condition that the state information of the target account group comprises the average duration of the second account access application program in the target account group, if the average duration corresponding to the target account group is within the duration range of the account group category, determining that the account group category is the target account group category.

According to a second aspect of embodiments of the present disclosure, there is provided an information recommendation apparatus, which may include:

The receiving module is configured to execute a recommendation request sent by a first account of the login application program, wherein the recommendation request is used for requesting to acquire the multimedia information;

The first determining module is configured to execute a determination of a target account group corresponding to the first account in response to the recommendation request, wherein the target account group comprises a second account, the similarity of first account information of the first account and second account information of the second account is greater than or equal to a preset similarity, and the first similarity is determined by social relationship information of the first account and the second account;

The first acquisition module is configured to acquire the category of the recommendation information corresponding to the state information according to the state information of the target account group, wherein the state information is used for representing and predicting the state of the second account application program;

And a recommending module configured to recommend the multimedia information corresponding to the category of the recommendation information to the first account.

In a possible embodiment, the information recommendation device further includes:

a second acquisition module configured to perform acquisition of account information of a plurality of accounts corresponding to the application program;

a processing module configured to perform grouping of the plurality of accounts according to the similarity of the account information to obtain a plurality of account groups; wherein,

The plurality of account groups includes a target account group.

In another possible embodiment, the first determining module related to the foregoing includes:

The matching module is configured to perform matching of the first account information with account information of accounts in each account group to obtain similarity between the first account and each account group;

The second determining module is configured to determine the account group as a target account group if the similarity between the first account and the account group is greater than a preset similarity.

In still another possible embodiment, the information recommendation device related to the foregoing further includes:

The third acquisition module is configured to execute, when the second account comprises a fourth account, and the fourth account is a second account with the frequency of accessing the application program being greater than or equal to the preset frequency in the target account group, summarizing historical access information of the target account group according to the access information of each second account in the target account group for accessing the application program;

The computing module is configured to execute computing of similarity of first behavior information of the first account access application program and second behavior information of the fourth account access application program to obtain target similarity of the first account and the target account group;

a fourth obtaining module configured to perform obtaining ratio information of the target account group, where the ratio information is used to characterize a ratio of the second account in the target account group to the platform account in a predetermined period of time;

And a third determination module configured to perform determining state information of the account access application in the target account group based on the history access information, the target similarity, and the scale information.

In a further possible embodiment, the third determining module related to the foregoing includes:

An extraction module configured to perform extraction of target history access information within a preset period of time from the history access information;

a fifth obtaining module configured to obtain predicted proportion information of the target account group, where the predicted proportion information is used to characterize a proportion of the second account in the target account group in the platform account in a preset time period;

and a fourth determining module configured to perform determining the state information based on the target history access information, the predicted scale information, the target similarity, and the scale information.

In a further possible embodiment, the fourth determination module referred to above is configured to perform,

Z＝1+1/B1*D*(1-T)*B2

In a further possible embodiment, the first acquisition module related to the foregoing includes:

A fifth determining module configured to determine a target account group class to which the target account group belongs according to the state information of the target account group;

and a sixth acquisition module configured to perform acquisition of a category of recommendation information corresponding to the target account group category from the association information of the preset account group category and the category of recommendation information.

In yet another possible embodiment, the fifth determining module is configured to determine, in a case where the state information of the target account group includes an average duration of the second account access application in the target account group, that the account group category is the target account group category if the average duration corresponding to the target account group falls within a duration range of the account group category.

According to a third aspect of embodiments of the present disclosure, there is provided a server, which may include:

A processor;

a memory configured to store processor-executable instructions;

wherein the processor is configured to execute instructions to implement the information recommendation method as shown in any of the embodiments of the first aspect.

According to a fourth aspect of embodiments of the present disclosure, there is provided a computer readable storage medium, which when executed by a processor of a server, causes the server to implement the information recommendation method as shown in any one of the embodiments of the first aspect.

According to a fifth aspect of embodiments of the present disclosure, there is provided a computer program product comprising a computer program stored in a readable storage medium, the computer program being read from the storage medium and executed by at least one processor of the device, causing the device to perform the information recommendation method shown in any one of the embodiments of the first aspect.

The technical scheme provided by the embodiment of the disclosure at least brings the following beneficial effects:

according to the embodiment of the disclosure, the first account of the login application program is divided into groups, a target account group corresponding to the first account is determined, and then the category of recommendation information corresponding to the state information is acquired according to the state information of the target account group, so that multimedia information corresponding to the category of recommendation information is recommended to the first account. Therefore, the information can be respectively recommended to the accounts in the corresponding account groups according to the state information of the different account groups, so that different groups correspond to different recommended contents, the current mode of recommending the accounts on the application program platform based on the relatively fixed account groups is eliminated, the targeted recommendation service to the accounts of multiple groups is facilitated, and the accuracy of information recommendation is improved.

It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the disclosure.

Drawings

The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the disclosure and together with the description, serve to explain the principles of the disclosure and do not constitute an undue limitation on the disclosure.

FIG. 1 is a diagram illustrating an architecture for information recommendation, according to an exemplary embodiment;

FIG. 2 is a schematic illustration of an application scenario illustrating an information recommendation, according to an example embodiment;

FIG. 3 is a flowchart illustrating a method of information recommendation, according to an example embodiment;

FIG. 4 is a block diagram illustrating a structure of an information recommendation device according to an exemplary embodiment;

FIG. 5 is a block diagram illustrating a computing device according to an example embodiment.

Detailed Description

In order to enable those skilled in the art to better understand the technical solutions of the present disclosure, the technical solutions of the embodiments of the present disclosure will be clearly and completely described below with reference to the accompanying drawings.

It should be noted that the terms "first," "second," and the like in the description and claims of the present disclosure and in the foregoing figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the disclosure described herein may be capable of operation in sequences other than those illustrated or described herein. The implementations described in the following exemplary examples are not representative of all implementations consistent with the present disclosure. Rather, they are merely examples of apparatus and methods consistent with some aspects of the present disclosure as detailed in the accompanying claims.

Currently, service platforms corresponding to some applications often face the dilemma that the user population is relatively fixed. For example, the old user in the service platform uses the middle-hard force of the service platform, and the maintenance cost is low. But a large number of old users can tilt the recommended content and the designed display interface morphology in the application towards the old users. The reason is that in order to refine or upgrade these recommended contents and designs, it is necessary to acquire feedback information of the user and evaluate the recommended contents and designs before the improvement of the recommended contents and designs can be achieved. If only the feedback information of the old user is obtained, the improved recommended content and design are also prone to the old user, so that accurate information recommendation for most or even all users corresponding to the application program cannot be realized.

The following may be illustrated by way of an example: the service platform is to evaluate a certain recommendation information A, and the service platform can select a plurality of related specific users, implement the recommendation information A to recommend information to the specific users, acquire feedback information of the specific users, and evaluate whether the recommendation information meets the requirement of the service platform. However, this way of obtaining feedback information requires knowledge of the domain of the specific information a to construct the user list, and if the user only uses the feedback information to evaluate the quality of the recommended information a, the evaluation is prone to be subjective and error. In addition, the above manner is difficult to generalize, and different recommendation strategies often need to reselect specific users, so that the calculation amount for selecting the specific users is increased, and the application range is small. In addition, the above evaluation method is only performed alone, and cannot give the optimization direction of the recommended information, that is, the above method cannot be organically combined with upgrading and optimizing of the recommendation system. Thus, there is a need for an evaluation method that is organically combined with information recommendation.

In view of this, the embodiment of the disclosure determines, by using the optimization of the number of active accounts for a long period of time in the service platform as a core traction, a state of the application program used for characterizing the predicted account in the predicted target account group, so as to implement an evaluation manner of combining the proportion information occupied by the target account group in the platform account, the access information of each second account in the target account group to access the application program, and the target similarity of the account and the target account group, to obtain the state information of the target account group, and select the recommendation information in combination with the state information.

Therefore, the information recommendation method provided in the embodiments of the present disclosure may be applied to the architecture as shown in fig. 1, and specifically described with reference to fig. 1.

Fig. 1 is a diagram illustrating an architecture for information recommendation, according to an exemplary embodiment.

As shown in fig. 1, the information recommendation system includes a client 10 and a server 20.

Wherein the client 10 comprises an application installed on an electronic device. The application program may include various types of applications, such as: a plurality of types of applications such as a (long or short) video playing type application, a music playing type application, an instant messaging type application and the like.

The server 20 may include a service platform corresponding to the application program, and the function of implementing the service platform may be implemented by a server or a cloud server. In the embodiment of the present disclosure, a server is described as an example.

Based on this, the steps in fig. 2 are described, as step 210 in fig. 2: the server may obtain a plurality of accounts registered on the application program, and group the plurality of accounts according to similarity of account information of the accounts, to obtain a plurality of account groups. Wherein the group attributes of each of the plurality of account groups are different. The account information includes at least one of the following: account portrait information, geographic location, priority information for geographic location.

Therefore, different contents are conveniently recommended to the accounts of the unnecessary groups, and of course, the recommendation information is also conveniently evaluated through the state information of the different account groups, so that the service platform can better acquire the data of the application program used by the users of different types, and different multimedia information is recommended according to the state information of the users of different types.

Thus, upon receiving a recommendation request sent by the first account of the login application, the recommendation request is used to request acquisition of multimedia information. As in step 220 of fig. 2: and responding to the recommendation request, determining a target account group corresponding to the first account, wherein the target account group comprises a second account, the similarity of the first account information of the first account and the second account information of the second account is greater than or equal to a preset similarity, and the first similarity is determined by the social relationship information of the first account and the second account.

Then, as in step 230 of fig. 2: acquiring the category of recommendation information corresponding to the state information according to the state information of the target account group, wherein the state information is used for representing and predicting the state of the application program used by the second account; multimedia information corresponding to the category of the recommended information is recommended to the first account. Therefore, the information can be respectively recommended to the accounts in the corresponding account groups according to the state information of the different account groups, so that different groups correspond to different recommended contents, the current mode of recommending the accounts on the application program platform based on the relatively fixed account groups is eliminated, the targeted recommendation service to the accounts of multiple groups is facilitated, and the accuracy of information recommendation is improved.

Based on the above information recommendation architecture, the state information in the embodiments of the present disclosure may be used to obtain a category of recommendation information corresponding to the state information, and recommend multimedia information to a user based on the category of recommendation information, and may also be applied to other scenarios, as specifically shown below:

As shown in fig. 2, as in step 240 and step 250 in fig. 2: first, the status information may be used as an evaluation index for evaluating the type of products such as display interfaces and/or recommendation information that are pushed by the service platform.

The state information of each account group of the plurality of account groups under different products can be calculated through an A/B experiment and the like. The higher the indicator value characterized by the status information, the higher the number of predicted overall accounts for the application over a preset period of time and the liveness of the users in the application. In this way, new products such as display interfaces and/or application functions may be generated based on the status information of the account group.

Second, as in step 260 of fig. 2: and optimizing the current information recommendation model by using the state information of the account group.

Wherein the status information of the account group may be used as a factor in adjusting the loss function in the information recommendation model. For example, in user click through rate model training of recommended content, the state information of account groups is used to redefine the loss function and the samples are re-weighted by the state information of each account group. In this way, the recommendation model may be more biased to the accuracy of predicting whether to click or not by the user with high attribute of the state information. Similarly, weighting the state information of the group of accounts may also be applied to a series of models, such as generating training against a neural network model, such as a CF-GAN network model, or a relevance ranking model, i.e., an L2R model.

According to the above architecture and application scenario, the information recommendation method provided by the embodiment of the present disclosure is described in detail below with reference to fig. 3, where the information recommendation method may be executed by a server in the server shown in fig. 1, and the embodiment of the present disclosure is not limited to this.

Fig. 3 is a flowchart illustrating an information recommendation method according to an exemplary embodiment.

As shown in fig. 3, the information recommendation method specifically includes the following steps:

First, in step 310, a recommendation request sent by a first account of a login application is received, where the recommendation request is used to request acquisition of multimedia information.

Next, in step 320, in response to the recommendation request, a target account group corresponding to the first account is determined, where the target account group includes a second account, and a similarity between the first account information of the first account and the second account information of the second account is greater than or equal to a preset similarity, and the first similarity is determined by social relationship information of the first account and the second account.

In addition, in step 330, according to the status information of the target account group, a category of recommended information corresponding to the status information is obtained, where the status information is used to characterize and predict the status of the second account application.

Then, step 340 recommends multimedia information corresponding to the category of the recommendation information to the first account.

Therefore, the embodiment of the disclosure determines the target account group corresponding to the first account by classifying the first account of the login application program, and then obtains the category of the recommendation information corresponding to the state information according to the state information of the target account group, and recommends the multimedia information corresponding to the category of the recommendation information to the first account. Therefore, the information can be respectively recommended to the accounts in the corresponding account groups according to the state information of the different account groups, so that different groups correspond to different recommended contents, the current mode of recommending the accounts on the application program platform based on the relatively fixed account groups is eliminated, the targeted recommendation service to the accounts of multiple groups is facilitated, and the accuracy of information recommendation is improved.

The following describes the above steps in detail, as follows:

first, in one possible embodiment, before step 320, the information recommendation method may further include:

Grouping the accounts according to the similarity of the account information to obtain a plurality of account groups; wherein the plurality of account groups includes a target account group.

Here, the account information includes at least one of the following information: account portrait information, zone priority. The account portrait information comprises portrait information such as age, gender, contact information and the like of the corresponding user of the account.

Based on this, a plurality of accounts may be grouped according to the similarity of territories and ages, e.g., where the territories are all shown at A and the ages differ by no more than 5 years, such accounts may be grouped into one account group.

Based on this, this step 320 may specifically include:

Matching the first account information with account information of accounts in each account group to obtain similarity between the first account and each account group; if the similarity between the first account and the account group is greater than the preset similarity, the account group is determined to be a target account group.

The embodiments of the present disclosure provide two ways of determining a target account group, specifically as follows:

mode one: and obtaining the similarity of the behavior information of the first account and the behavior information of each account in each account group, and weighting the similarities to obtain the similarity of the first account and each account group. And then, selecting an account group corresponding to the similarity larger than the preset similarity from the similarities of the first account and each account group, and determining the account group corresponding to the similarity larger than the preset similarity as a target account group.

Mode two: and obtaining the similarity of the behavior information of the first account and the behavior information of each account in each account group, and weighting the similarities to obtain the similarity of the first account and each account in each account group. And then, selecting the first N second accounts with the similarity larger than a preset threshold value from the similarity between the first accounts and each account in the plurality of first accounts, and counting the account group with the largest number of the N second accounts in each account group as a target account group.

Here, in the embodiment of the present disclosure, the accounts may be divided into account groups with the finest granularity according to the dimensions of age, gender, region priority, and the like related in the account information, and the inter-group difference may be maximized according to the decision tree model in the decision tree analysis method. Here, the decision tree analysis method is a risk type decision method that uses probabilities to compare different schemes in decisions with trees in graph theory, thereby obtaining an optimal scheme. Then, after grouping suggestions are made by the decision tree, a final target account group is determined using a least cadier's score group that may contain the final all classifications of the decision tree.

Second, in one possible embodiment, the second account includes a fourth account, the fourth account being a second account in the target account group having a frequency of accessing the application greater than or equal to a preset frequency. Thus, prior to step 330, the information recommendation method may further include:

summarizing historical access information of the target account group according to the access information of each second account access application program in the target account group;

Here, the reason for selecting the 3 parameters of the historical access information, the target similarity and the proportion information in the embodiment of the disclosure is as follows, calculating the proportion information of each account group obtained by grouping according to census data, using the historical access information of each account group of the application program, constructing a quadratic regression model of the historical access information of each account group, the target similarity and the proportion information, and adding a new account and the number of return accounts to control the application program to obtain the influence of the new account (wherein, the return user is a user who is silent for more than 7 days and is involved in the re-regression of the month in a preset period, the silencing can refer to unloading the application program or not using a record, and similarly, the re-regression can refer to re-downloading the application program or having a use record), and predicting the proportion information of each account group in the future, for example, after half a year. Based on this quadratic regression model an estimate of this parameter, the status information of the account group, can be obtained. In addition, the predictive effect of the quadratic regression model was evaluated using the preset test set, and the increment of the calculated proportion information and the proportion information of each account group in the future, for example, after half a year, was 2.7%. This illustrates that the quadratic regression model has small errors and can react more accurately to changes in the scale information, and thus, embodiments of the present disclosure determine state information for account access applications in an account group based on historical access information, target similarity, and scale information.

Here, the embodiment of the present invention provides a manner of calculating history access information, target similarity, and scale information, which is specifically shown below.

(1) Computing historical access information

The information recommendation method may further include: and obtaining historical access information according to the access frequency of each account in the target account group to the application program.

For example, short-term sensitivity index fitting is used to access information for medium-term users through machine learning and causal analysis methods. Wherein the historical access information includes retention, i.e., the frequency with which the application is currently accessed, and predicting future continued access to the application. User access information (e.g., 30 days after account access, i.e., long retention) is important but difficult to observe in the short term and has large volatility, whereby medium-long term retention is mapped into short term index retention using methods of machine learning and causal analysis. This aspect may fit long-term access information in the application in the future through the a/B index system, and may also fit short-term historical access information of the user, and define the overall index of account group k as Σ _kθ_kf(α_1k,α_2k,...,α_nk), where ajk is the j-th short-term index of account group k, Σ _kθ_kf(α_1k,α_2k,...,α_nk) is larger, the higher the historical access information of the account group, such as long-term retention, where n and j are both positive integers.

(2) Calculating the similarity of targets

Acquiring first behavior information of a first account access application program and second behavior information of a fourth account access application program;

and obtaining target similarity according to the Euclidean distance between the first behavior information and the second behavior information, wherein the target similarity is used for representing the similarity degree of the first account and the fourth account access application program.

The method for obtaining the target similarity is specifically described by the following method:

Here, the method of calculating the target similarity is to select the account with higher activity in the target account group as the representative of all accounts in the entire target account group. And secondly, measuring the similarity of the first account and the target account group by using the behavior similarity between the account with higher activity and the first account. For example, the target similarity may be calculated using all new users from the pre-installed channel and sampling a portion of the IOS system as the first account, and sampling the IOS system's accounts as the more active accounts in the target account group.

Here, the reason for employing pre-installed channels is that the new users will not have selective bias in channels like information flow advertisements when selecting to open an application, but pre-installed channels are typically android system users, so in the embodiments of the present disclosure, a portion of users of the random IOS system are supplemented as a sample. In principle the sample can also be built by a richer approach, e.g. by joining a new first account pulled by some other operating activity of the application.

The target similarity calculation method may then comprise the step of, for an active account in the application referred to in a target account group i, finding the intersection of this active account with all new first account exposed videos from the target account group i. This allows to find the euclidean distance between the video completion rate of this active account and the video completion rate of the generally new first account on these videos. Finally, a decreasing function may be defined for which the similarity is i 2 distance smoothed to between 0 and 1.

In order to ensure the rationality of the target similarity manner determined by the above, verification can be performed as follows. In this case, as shown in table 1, the calculation of the target similarity is exemplified.

TABLE 1

Based on table 1, the euclidean distance between the new first account and the target account group k where the active account is located is dk= [ (S1-O1) 2+ (S2-O2) 2+ (S3-O3) 2 ]/number of video IDs, and the larger the value of Dk is, the larger the difference between new and old users is, and the lower the similarity is. Then, in order to reduce the calculation amount, all 1/Dk are normalized to be in the range of 0-1, and the target similarity is obtained.

(3) Determining scale information

The ratio information is determined based on the number of second accounts in the target account group and the total number of accounts in the application. Here, the ratio information may include user permeability.

Based on this, the state information of the application program can be determined in the following manner, specifically as follows:

obtaining forecast proportion information of a target account group, wherein the forecast proportion information is used for representing the proportion of a second account in the target account group in a platform account in a forecast preset time period; here, the above prediction proportion information may be calculated by the target similarity and the history access information.

Wherein the status information may be calculated according to the following formula (1):

Z＝1+1/B1*D*(1-T)*B2 (1)

Based on this, the formula is described by way of example below, where B1 is 0.55, which indicates that the second account in the target account group occupies about half of the platform account, i.e., 55%, within 30 days; at D of 0.8, the first behavior information representing the first account accessing application is very similar to the second behavior information of the fourth account accessing application in the target account group by 80%, where a higher target similarity means that the first account is a popular account in the population of target accounts; when T is 0.6, the total value of the access information of each account access application program of the target account group in 30 days reaches 60%, namely the access information of the second account access application program of the target account group in each day accounts for more than half; at B2 of 0.5, it may be indicated that the proportion of the second account in the target account group in the platform account is about half, i.e. 50%, within 30 days after the next half year, meaning that the target account group has a high retention in the platform, and it is highly likely that the application will continue to be used for at least 30 days after the half year. Thus, the above data is brought into formula (1), i.e., z=1+1/0.55 x 0.8 (1-0.6) x 0.5, and Z is 12.36 greater than 10, which means that the second account in the target account group is more likely to be continuously used by the application program within 30 days after the next half year, and of course, the new product form pushed by the application program will also be resource-tilted toward the product direction of accessing the application program by the target account group.

Because the embodiment of the disclosure adopts a new mode to determine the target similarity and determines the prediction proportion information through the target similarity, the embodiment of the disclosure also provides a process for calculating the reason of the prediction proportion information and verifying the prediction proportion information, which is specifically as follows:

First, the objective similarity may largely account for the marginal effect of historical access information on future scale information. This gain should come from two aspects. First, a high target similarity characterizes an application with a high acceptance. Then each account in the target account population can be quickly assimilated once it is contacted by the application, provided that the historical access information from this first account is optimized for higher values. Therefore, the effect of improving access information is greater with high target similarity. And secondly, if the target similarity is higher, the social attribute is stronger. A high target similarity means that the first account is a larger group of accounts under the target account group, and may also indicate that the accounts in the target account group should be a larger group of accounts under the target account group. Thus, the prediction ratio information can be calculated by the target similarity and the history access information.

Second, in one possible embodiment, the step 330 may specifically include:

And under the condition that the state information of the target account group comprises the average time length of the second account access application program in the target account group, if the average time length corresponding to the target account group is within the time length range of the account group category, determining that the account group category is the target account group category.

For example, the duration occupied by the state of each content of the application program used by the target account group can be obtained according to the state information of the target account group, and the category of the recommendation information corresponding to the category of the target account group corresponding to the duration occupied can be found. Wherein the content includes product information and/or recommendation information corresponding to the product information. If the target account group uses the application program content 1 for 20 minutes, the category of the recommendation information corresponding to the 20 minutes for 20 minutes is category a. The target account group uses the application program content 1 for 10 hours, and the category of the recommendation information corresponding to the 10 hours is category B. Similarly, if the duration of using the application program content 2 by the target account group is 50 minutes, the category of the recommended information corresponding to the 50 minutes is category a. The target account group uses the application content 2 for 10 hours, and the category of the recommendation information corresponding to the 10 hours is category B.

In this way, the state information of the target account group is used as an evaluation index to generate an index system for evaluating products pushed by the service platform, such as a display interface and/or a recommendation strategy, and in addition, the state information of the target account group is input into a multimedia recommendation model of an application program to obtain a category of recommendation information corresponding to the state information; wherein the multimedia recommendation model is determined by behavioral trend information for a plurality of account groups.

Wherein after the state information of the target account group is input into the multimedia recommendation model of the application program to obtain the category of the recommendation information corresponding to the state information, the method further comprises:

And training the multimedia recommendation model by taking the state information of the target account group as a sample until the preset training condition is met, so as to obtain the optimized multimedia recommendation model.

Here, the state information of the target account group may be used to optimize the current information recommendation model. Wherein the status information of the target account group may be used as a factor in adjusting the loss function in the information recommendation model. For example, in user click through rate model training of recommended content, the loss function is redefined with and samples are re-weighted by the state information of the target account group. In this way, the recommendation model may be more biased to the accuracy of click prediction for users with high attributes of the state information of the target account group. Similarly, weighting the state information of the target account group may also be applied to a series of models, such as generating training against neural network models, such as a CF-GAN network model, or a relevance ranking model, i.e., an L2R model.

In summary, the embodiments of the present disclosure determine a target account group corresponding to a first account by classifying the first account of a login application, and then obtain a category of recommendation information corresponding to the status information according to the status information of the target account group, and recommend multimedia information corresponding to the category of recommendation information to the first account. Therefore, the information can be respectively recommended to the accounts in the corresponding account groups according to the state information of the different account groups, so that different groups correspond to different recommended contents, the current mode of recommending the accounts on the application program platform based on the relatively fixed account groups is eliminated, the targeted recommendation service to the accounts of multiple groups is facilitated, and the accuracy of information recommendation is improved.

In addition, embodiments of the present disclosure propose a state information system that is reusable, generalizable, and comprehensive for account groups. For existing schemes, which tend to be temporary and non-generalizable, the state information derivation of account groups in embodiments of the present disclosure is from a long-term retention optimization perspective, and from inherent attributes of account groups, such as historical access information, target similarity, and scale information. Therefore, the state information of the account group can be repeatedly and widely applied to various products and recommendation system optimization without manual participation.

It should be noted that, the application scenario described in the foregoing embodiments of the present disclosure is for more clearly describing the technical solution of the embodiments of the present disclosure, and does not constitute a limitation on the technical solution provided by the embodiments of the present disclosure, and as a person of ordinary skill in the art can know that, with the appearance of a new application scenario, the technical solution provided by the embodiments of the present disclosure is equally applicable to similar technical problems.

Based on the same inventive concept, the present disclosure also provides an information recommendation apparatus. This is described in detail with reference to fig. 4.

Fig. 4 is a block diagram showing a structure of an information recommending apparatus according to an exemplary embodiment.

As shown in fig. 4, the information recommendation device 40 may specifically include:

a receiving module 401 configured to execute a recommendation request sent by a first account of the login application, where the recommendation request is used for requesting to acquire multimedia information;

a first determining module 402 configured to perform determining, in response to the recommendation request, a target account group corresponding to the first account, where the target account group includes a second account, and a similarity between first account information of the first account and second account information of the second account is greater than or equal to a preset similarity, where the first similarity is determined by social relationship information of the first account and the second account;

A first obtaining module 403 configured to obtain a category of recommended information corresponding to the status information according to the status information of the target account group, the status information being used for characterizing and predicting a status of the second account usage application;

The recommendation module 404 is configured to perform recommendation of the multimedia information corresponding to the category of the recommendation information to the first account.

The plurality of account groups includes a target account group.

In another possible embodiment, the first determining module 402 related to the foregoing includes:

Z＝1+1/B1*D*(1-T)*B2

In yet another possible embodiment, the first acquisition module 403 referred to above includes:

Based on the same inventive concept, the embodiment of the present disclosure further provides a computing device, which is specifically described in detail with reference to fig. 5.

As shown in fig. 5, the computing device 50 is capable of implementing a structural diagram of an exemplary hardware architecture of a computing device according to an information recommendation method and an information recommendation apparatus in an embodiment of the present disclosure. The computing device may refer to a server in embodiments of the present disclosure.

The computing device 50 may include a processor 501 and a memory 502 storing computer program instructions.

In particular, the processor 501 may include a Central Processing Unit (CPU), or an Application SPECIFIC INTEGRATED Circuit (ASIC), or may be configured as one or more integrated circuits that implement embodiments of the present disclosure.

Memory 502 may include mass storage configured as information or instructions. By way of example, and not limitation, memory 1202 may include a hard disk drive (HARD DISK DRIVE, HDD), floppy disk drive, flash memory, optical disk, magneto-optical disk, magnetic tape, or universal serial bus (universal serial bus, USB) drive, or a combination of two or more of these. Memory 1202 may include removable or non-removable (or fixed) media where appropriate. The memory 1202 may be internal or external to the integrated gateway device, where appropriate. In a particular embodiment, the memory 1202 is a non-volatile solid-state memory. In particular embodiments, memory 1202 includes Read Only Memory (ROM). The ROM may be mask programmed ROM, programmable ROM (PROM), erasable PROM (EPROM), electrically Erasable PROM (EEPROM), electrically rewritable ROM (EAROM), or flash memory, or a combination of two or more of these, where appropriate.

The processor 501 reads and executes the computer program instructions stored in the memory 502 to perform the steps of:

A processor 501 configured to execute a recommendation request sent by a first account receiving a login application, the recommendation request being used for requesting acquisition of multimedia information; determining a target account group corresponding to the first account in response to the recommendation request, wherein the target account group comprises a second account, the similarity between the first account information of the first account and the second account information of the second account is greater than or equal to a preset similarity, and the first similarity is determined by the social relationship information of the first account and the second account; acquiring the category of recommendation information corresponding to the state information according to the state information of the target account group, wherein the state information is used for representing and predicting the state of the application program used by the second account; and recommending the multimedia information corresponding to the category of the recommended information to the first account.

In one possible embodiment, the processor 501 is configured to perform obtaining account information for a plurality of accounts corresponding to an application; grouping the accounts according to the similarity of the account information to obtain a plurality of account groups; wherein the plurality of account groups includes a target account group.

In another possible embodiment, the processor 501 is configured to perform matching the first account information with account information of accounts in each account group to obtain a similarity between the first account and each account group; and a second determining module configured to determine the account group as a target account group if the similarity between the first account and the account group is greater than a preset similarity.

In yet another possible embodiment, the processor 501 is configured to perform summarizing, when the second accounts include fourth accounts, and the fourth accounts are second accounts with a frequency of accessing the application program in the target account group greater than or equal to a preset frequency, historical access information of the target account group according to access information of each second account in the target account group; calculating the similarity of the first behavior information of the first account access application program and the second behavior information of the fourth account access application program to obtain the target similarity of the first account and the target account group; obtaining proportion information of a target account group, wherein the proportion information is used for representing the proportion of a second account in the target account group in a platform account in a preset time period; and determining status information of the account access application in the target account group based on the historical access information, the target similarity, and the scale information.

In yet another possible embodiment, the processor 501 is configured to extract the target historical access information within a preset time period from the historical access information; obtaining forecast proportion information of a target account group, wherein the forecast proportion information is used for representing the proportion of a second account in the target account group in a platform account in a forecast preset time period; and determining state information according to the target historical access information, the prediction proportion information, the target similarity and the proportion information.

In yet another possible embodiment, the processor 501 referred to above is configured to execute,

Z＝1+1/B1*D*(1-T)*B2

In yet another possible embodiment, the processor 501 is configured to determine, according to the status information of the target account group, a target account group class to which the target account group belongs; and acquiring the category of the recommendation information corresponding to the target account group category from the preset associated information of the account group category and the category of the recommendation information.

In still another possible embodiment, the processor 501 is configured to determine that the account group class is the target account group class if the state information of the target account group includes an average duration of the second account access application in the target account group, and if the average duration corresponding to the target account group is within a duration range of the account group class.

In one example, the computing device 50 may also include a transceiver 503 and a bus 504. As shown in fig. 5, the processor 501, the memory 502, and the transceiver 503 are connected to each other via the bus 504 and perform communication with each other.

Bus 504 includes hardware, software, or both. By way of example, and not limitation, the buses may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), a HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an infiniband interconnect, a Low Pin Count (LPC) bus, a memory bus, a micro channel architecture (MCa) bus, a Peripheral Control Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a video electronics standards association local (VLB) bus, or other suitable bus, or a combination of two or more of the above. Bus 503 may include one or more buses, where appropriate. Although embodiments of the disclosure describe and illustrate a particular bus, the disclosure contemplates any suitable bus or interconnect.

The embodiment of the disclosure also provides a computer storage medium, in which computer executable instructions are stored, configured to implement the information recommendation method described in the embodiment of the disclosure.

In some possible implementations, aspects of the methods provided by the present disclosure may also be implemented in the form of a program product comprising program code configured to cause a computer device to perform the steps of the methods described in the present specification according to the various exemplary implementations of the present disclosure, when the program product is run on a computer device, for example, the computer device may perform the information recommendation method described in the embodiments of the present disclosure.

The program product may employ any combination of one or more readable media. The readable medium may be a readable signal medium or a readable storage medium. The readable storage medium may be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a combination of any of the foregoing. More specific examples (a non-exhaustive list) of the readable storage medium would include the following: an electrical connection having one or more wires, a portable disk, a hard disk, random Access Memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

The present disclosure is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus and computer program products according to the disclosure. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable control display device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable control display device, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable control display device to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable control display device to cause a series of operational steps to be performed on the computer or other programmable device to produce a computer implemented process such that the instructions which execute on the computer or other programmable device provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

It will be apparent to those skilled in the art that various modifications and variations can be made to the present disclosure without departing from the spirit or scope of the disclosure. Thus, the present disclosure is intended to include such modifications and alterations insofar as they come within the scope of the appended claims or the equivalents thereof.

Claims

1. An information recommendation method, comprising:

Determining a target account group corresponding to the first account in response to the recommendation request, wherein the target account group comprises a second account, the similarity between the first account information of the first account and the second account information of the second account is greater than or equal to a preset similarity, and the first similarity is determined by social relationship information of the first account and the second account; the second account comprises a fourth account, and the fourth account is a second account with the frequency of accessing the application program being greater than or equal to a preset frequency in the target account group;

Summarizing historical access information of the target account group according to the access information of each second account in the target account group for accessing the application program;

calculating the similarity of the first behavior information of the first account accessing the application program and the second behavior information of the fourth account accessing the application program to obtain the target similarity of the first account and the target account group;

Obtaining proportion information of the target account group, wherein the proportion information is used for representing the proportion of a second account in the target account group in a platform account in a preset time period;

Determining state information of the application program accessed by the accounts in the target account group based on the historical access information, the target similarity and the proportion information;

Accessing state information of the application program according to the account in the target account group, and acquiring a category of recommended information corresponding to the state information, wherein the state information is used for representing and predicting the state of the second account using the application program;

and recommending the multimedia information corresponding to the category of the recommendation information to the first account.

2. The method of claim 1, wherein prior to the determining the group of target accounts corresponding to the first account, the method further comprises:

The plurality of account groups includes the target account group.

3. The method of claim 2, wherein the determining the group of target accounts corresponding to the first account comprises:

And if the similarity between the first account and the account group is greater than the preset similarity, determining the account group as the target account group.

4. The method of claim 1, wherein the determining status information of the application for account access in the target account group based on the historical access information, the target similarity, and the scale information comprises:

Obtaining forecast proportion information of the target account group, wherein the forecast proportion information is used for representing the proportion of a second account in the target account group in the platform account in a forecast preset time period;

and determining the state information according to the target historical access information, the prediction proportion information, the target similarity and the proportion information.

5. The method of claim 4, wherein the determining the status information based on the target historical access information, the predicted access proportion information, the target similarity, and the proportion information comprises:

Z＝1+1/B1*D*(1-T)*B2

wherein z is the state information, B1 is the proportion information, D is the target similarity, T is the target history access information, and B2 is the predicted proportion information.

6. The method according to claim 1, wherein the obtaining the category of the recommended information corresponding to the status information according to the status information of the target account group includes:

Determining a target account group class to which the target account group belongs according to the state information of the target account group;

7. The method of claim 6, wherein the status information for the target account group includes an average duration of a second account access application in the target account group; the determining, according to the status information of the target account group, a target account group class to which the target account group belongs includes:

and if the average duration corresponding to the target account group belongs to the duration range of the account group category, determining the account group category as the target account group category.

8. An information recommendation device, characterized by comprising:

A receiving module configured to execute a recommendation request sent by a first account of a login application, wherein the recommendation request is used for requesting to acquire multimedia information;

A first determining module configured to execute determining a target account group corresponding to the first account in response to the recommendation request, wherein the target account group comprises a second account, and the similarity between the first account information of the first account and the second account information of the second account is greater than or equal to a preset similarity, and the first similarity is determined by social relationship information of the first account and the second account; the second account comprises a fourth account, and the fourth account is a second account with the frequency of accessing the application program being greater than or equal to a preset frequency in the target account group;

A third acquisition module configured to perform access information for accessing the application according to each second account in the target account group, summarizing historical access information of the target account group;

A calculating module configured to perform calculation of similarity between first behavior information of the first account accessing the application program and second behavior information of the fourth account accessing the application program, so as to obtain target similarity between the first account and the target account group;

a fourth obtaining module configured to perform obtaining proportion information of the target account group, where the proportion information is used to characterize a proportion of the second account in the target account group in the platform account in a predetermined period of time;

a third determination module configured to perform determining status information of accounts in the target account group accessing the application based on the historical access information, the target similarity, and the scale information;

A first obtaining module configured to perform accessing state information of the application program according to an account in the target account group, and obtain a category of recommended information corresponding to the state information, where the state information is used for characterizing and predicting a state of the application program used by the second account;

and a recommending module configured to execute recommending the multimedia information corresponding to the category of the recommending information to the first account.

9. The apparatus of claim 8, wherein the information recommendation apparatus further comprises:

A processing module configured to perform grouping of the plurality of accounts according to the similarity of account information to obtain the plurality of account groups; wherein,

The plurality of account groups includes the target account group.

10. The apparatus of claim 9, wherein the first determining module comprises:

The matching module is configured to perform matching of the first account information and account information of accounts in each account group to obtain similarity between the first account and each account group;

and the second determining module is configured to determine the account group as the target account group if the similarity between the first account and the account group is greater than a preset similarity.

11. The apparatus of claim 8, wherein the third determination module comprises:

an extraction module configured to perform extraction of target history access information within a preset period of time in the history access information;

A fifth obtaining module configured to obtain predicted proportion information of the target account group, where the predicted proportion information is used to characterize a proportion of the second account in the target account group in the platform account predicted in a preset time period;

And a fourth determination module configured to perform determination of the state information based on the target history access information, the predicted scale information, the target similarity, and the scale information.

12. The apparatus of claim 11, wherein the fourth determination module is configured to perform,

Z＝1+1/B1*D*(1-T)*B2

13. The apparatus of claim 8, wherein the first acquisition module comprises:

And a sixth acquisition module configured to acquire a category of recommendation information corresponding to the target account group category from the association information of the preset account group category and the category of recommendation information.

14. The apparatus of claim 13, wherein the fifth determination module is configured to perform, if the status information of the target account group includes an average duration of a second account access application in the target account group, determine that the account group category is a target account group category if the average duration corresponding to the target account group falls within a duration range of an account group category.

15. A server, comprising:

A processor;

a memory configured to store the processor-executable instructions;

wherein the processor is configured to execute the instructions to implement the information recommendation method of any of claims 1-7.

16. A computer readable storage medium, characterized in that instructions in the computer readable storage medium, when executed by a processor of a server, cause the processor of the server to implement the information recommendation method according to any one of claims 1 to 7.