Account filtering policy updating method, device, equipment and storage medium

Abstract

The present application discloses an update method, device, equipment and storage medium for an account filtering strategy, which belongs to the field of data processing. Through the technical solution provided by the embodiment of the present application, the process of pushing content items is divided into multiple statistical periods, and the expected number of pushes in each statistical period and the number of accounts filtered out by the account filtering level are monitored in real time, so as to determine the impact parameters of different account filtering levels in real time. The impact parameters can quantify the degree of influence of the account filtering level on the difference in the expected number of pushes in different statistical periods. Dynamically adjusting the account filtering strategy of each account filtering level according to the impact parameters can ensure to the greatest extent that the number of pushed content items reaches the agreed number of pushes, thereby improving the effect of content item pushes.

Description

Technical Field

[0001] The present application relates to the field of data processing, and in particular to a method, apparatus, device, and storage medium for updating an account filtering policy. Background Art

[0002] With the development of computer technology, content providers will push content items to users through content delivery platforms to achieve the purpose of content dissemination. For example, businesses will push advertisements to users through advertising delivery platforms.

[0003] In related technologies, content providers agree on a push quantity with content delivery platforms, which are then required to push content items to users according to the agreed push quantity. During the content delivery platform's push process, it screens accounts based on pre-set rules and ultimately pushes content items to these accounts.

[0004] However, after the accounts are screened, the number of screened accounts is often less than the agreed push quantity, resulting in failure to achieve the goal of content item push and poor content item push effect. Summary of the Invention

[0005] The embodiments of the present application provide a method, apparatus, device, and storage medium for updating an account filtering policy, which can improve the effect of content item push. The technical solution is as follows:

[0006] In one aspect, a method for updating an account filtering policy is provided, the method comprising:

[0007] Obtaining a first estimated push quantity of the content item to be pushed in a previous statistical period, where the first estimated push quantity is the number of accounts that are estimated to push the content item obtained in the previous statistical period;

[0008] Recalling multiple target push accounts for the content item, where the target push accounts are accounts that meet the push conditions for the content item;

[0009] filtering the multiple target push accounts using the account filtering policies of the multiple account filtering levels to obtain a second estimated push quantity for the content item and a first quantity of accounts filtered out by each of the account filtering levels;

[0010] Obtaining, based on the first number of accounts and the number of second accounts filtered out by each account filtering level in the previous statistical period, an impact parameter for each account filtering level, the impact parameter being used to indicate the degree of impact of the corresponding account filtering level on the change between the second estimated number of push notifications and the first estimated number of push notifications;

[0011] Based on the impact parameter of each account filtering level, the account filtering policies of the multiple account filtering levels are updated.

[0012] In one aspect, a device for updating an account filtering policy is provided, the device comprising:

[0013] A push quantity acquisition module is used to obtain a first estimated push quantity of the content item to be pushed in the last statistical period, where the first estimated push quantity is the number of accounts that are expected to push the content item obtained in the last statistical period;

[0014] A recall module, configured to recall multiple target push accounts for the content item, wherein the target push accounts are accounts that meet the push conditions for the content item;

[0015] a filtering module, configured to filter the plurality of target push accounts using account filtering policies of a plurality of account filtering levels, to obtain a second estimated push quantity for the content item and a first number of accounts filtered out by each of the account filtering levels;

[0016] an impact parameter acquisition module, configured to acquire an impact parameter for each account filtering layer based on the number of the first accounts and the number of second accounts filtered out by each account filtering layer in the previous statistical period, the impact parameter being used to indicate the degree of impact of the corresponding account filtering layer on the change between the second estimated number of push notifications and the first estimated number of push notifications;

[0017] A policy updating module is configured to update the account filtering policies of the plurality of account filtering levels based on an impact parameter of each account filtering level.

[0018] In one possible implementation, the policy update module is configured to update the account filtering policy of any account filtering level in response to an impact parameter of any account filtering level being greater than or equal to an impact parameter threshold, so as to reduce the number of accounts filtered out by any account filtering level.

[0019] In a possible implementation manner, the current statistical period is the last statistical period, and the apparatus further includes:

[0020] The push module is used to push the content item to the filtered multiple target push accounts.

[0021] In one possible implementation, the account filtering policies of the multiple account filtering levels include at least one of the following:

[0022] Priority filtering, push frequency filtering, freshness filtering, and redistribution filtering;

[0023] The priority filtering refers to filtering accounts that have been assigned to other content items with a higher priority than the content item;

[0024] The push frequency filtering refers to filtering accounts whose push times exceed the first number threshold within the first target duration;

[0025] The freshness filtering refers to filtering out accounts that have pushed the content item more than a second number threshold number of times within the second target duration;

[0026] The reallocation filtering refers to filtering accounts that have been allocated to other types of content item push services.

[0027] On the one hand, a computer device is provided, which includes one or more processors and one or more memories, wherein at least one computer program is stored in the one or more memories, and the computer program is loaded and executed by the one or more processors to implement the account filtering policy update method.

[0028] In one aspect, a computer-readable storage medium is provided, wherein at least one computer program is stored in the computer-readable storage medium, and the computer program is loaded and executed by a processor to implement the method for updating the account filtering policy.

[0029] On the one hand, a computer program product or computer program is provided, which includes a program code, which is stored in a computer-readable storage medium. The processor of a computer device reads the program code from the computer-readable storage medium, and the processor executes the program code, so that the computer device executes the above-mentioned account filtering policy update method.

[0030] Through the technical solution provided in the embodiments of this application, the content item push process is divided into multiple statistical periods. The expected number of pushes in each statistical period and the number of accounts filtered out by the account filtering layer are monitored in real time, thereby determining the impact parameters of different account filtering layers in real time. The impact parameters can quantify the degree of influence of the account filtering layer on the difference in the expected number of pushes in different statistical periods. Dynamically adjusting the account filtering strategy of each account filtering layer based on the impact parameters can maximize the guarantee that the number of pushed content items reaches the agreed push number, thereby improving the effectiveness of content item push. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0032] Figure 1 This is a schematic diagram of the interaction between various platforms in an advertising transaction provided by an embodiment of the present application;

[0033] Figure 2 This is a schematic diagram of an implementation environment of a method for updating an account filtering policy provided in an embodiment of the present application;

[0034] Figure 3 This is a flow chart of a method for updating an account filtering policy provided in an embodiment of the present application;

[0035] Figure 4 This is a flow chart of a method for updating an account filtering policy provided in an embodiment of the present application;

[0036] Figure 5 This is a schematic diagram of a statistical table provided in an embodiment of the present application;

[0037] Figure 6 is a schematic diagram of a flow distribution funnel provided in an embodiment of the present application;

[0038] Figure 7 Schematic diagram of a difference statistics table provided in an embodiment of the present application;

[0039] Figure 8 This is a schematic diagram of the results of a phase relationship analysis provided in an embodiment of the present application;

[0040] Figure 9 This is a schematic diagram of the structure of an account filtering policy update device provided in an embodiment of the present application;

[0041] Figure 10 This is a structural diagram of a server provided in an embodiment of the present application. DETAILED DESCRIPTION

[0042] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0043] In this application, the terms "first", "second", etc. are used to distinguish identical or similar items with substantially the same effects and functions. It should be understood that there is no logical or temporal dependency between "first", "second", and "nth", nor is there any limitation on the quantity and execution order.

[0044] In this application, the term "at least one" means one or more, and the term "plurality" means two or more. For example, a plurality of reference facial images refers to two or more reference facial images.

[0045] Contract advertising: This includes ad space contracts and impression contracts. Ad space contracts involve an agreement between an advertiser and a publisher to place their ads on certain ad spaces within a specific timeframe. The corresponding settlement method is CPT (Cost Per Time). In this model, the advertiser is paid for ad space. Impression contracts involve a predetermined number of impressions under certain audience conditions, with settlement based on a pre-agreed price per impression. This settlement method is CPM (Cost Per Mille), also known as Guarantee Delivery (GD), which guarantees that the advertiser's requested ad impressions will be met. In this model, advertisers are paid for both ad space and audience.

[0046] PDB (Programmatic Direct Buying) means that before an ad is released, an order is placed with the media based on the advertiser's delivery needs, with a fixed CPM price, fixed resource position, and fixed reservation quantity. During the ad delivery process, when a user generates an exposure opportunity by accessing the media, the media will send the ad request to a single demander based on the advertiser's reservation quantity. The demander can selectively select and return traffic according to the rules of the N-fold push agreement without bidding.

[0047] CPC (Cost Per Click) is a form of advertising that charges per click.

[0048] Adx (Ad Exchange) is a platform that provides advertising transactions. Adx can integrate information from multiple SSPs for DSP to select.

[0049] DSP (Demand Side Platform) refers to an advertising delivery platform that provides advertisers with cross-media, cross-platform, and cross-terminal capabilities. Through data integration and analysis, it achieves precise advertising delivery based on audiences. DSP can interact with Adx and integrate media information provided by different Adxs. Advertisers can quickly deliver advertisements through DSP.

[0050] SSP (Supply Side Platform) refers to a platform provided for traffic providers (media). Traffic providers use SSP to display the traffic they can provide. SSP interacts with Adx and DSP to communicate with advertisers and media, thereby reaching advertising transactions.

[0051] The relationship between DSP, Adx and SSP can be found in Figure 1Advertisers place advertising orders through the DSP platform, and the advertising delivery platform (media) uploads the traffic (number of accounts) available for sale through the SSP. Adx matches transactions based on the advertiser's needs and the traffic provided by the media, and matches the most suitable media for advertisers to deliver advertisements.

[0052] Figure 2 This is a schematic diagram of an implementation environment of an account filtering policy update method provided in an embodiment of the present application, see Figure 2 The implementation environment may include a terminal 210 and a server 240, wherein the server 240 is an advertising exchange platform (Adx) or an advertising delivery platform (media).

[0053] The terminal 210 is connected to the server 240 via a wireless network or a wired network. Optionally, the terminal 210 is a smart phone, a tablet computer, a laptop computer, a desktop computer, a smart speaker, a smart watch, etc., but is not limited thereto. The terminal 210 installs and runs an application that supports account filtering policy updates.

[0054] Optionally, the server is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, distribution networks (Content Delivery Network, CDN), and big data and artificial intelligence platforms.

[0055] Optionally, the terminal 210 generally refers to one of multiple terminals, and the embodiment of the present application only takes the terminal 210 as an example.

[0056] Those skilled in the art will appreciate that the number of terminals may be greater or less. For example, there may be only one terminal, or there may be dozens, hundreds, or even more terminals, in which case the implementation environment may also include other terminals. The embodiments of this application do not limit the number or device types of terminals.

[0057] After introducing the implementation environment of the technical solution provided in the embodiment of the present application, the application scenario provided in the embodiment of the present application is described below.

[0058] 1. The technical solutions provided in the embodiments of this application can be applied to the push scenario of contract advertising. The business process of contract advertising is mainly divided into four steps: resource inquiry, order locking, order delivery, and monitoring and settlement. Resource inquiry refers to the available traffic displayed by the advertising exchange (AdX) to advertisers before placing an order. In this embodiment, traffic refers to the number of accounts available for advertising delivery. Order locking refers to the process in which, after an advertiser fills in an advertising delivery requirement and places an order on the advertising exchange, the advertising exchange determines the advertising delivery platform (media) that meets the advertiser's requirements and the number of ads the advertiser needs to deliver. In some embodiments, the delivery requirement includes the number of ads delivered and the target audience for the delivery. In contract advertising, the advertising delivery platform will only receive the advertiser's full commission if the number of ads delivered by the advertising delivery platform meets the advertiser's requirements. If the number of ads delivered by the advertising delivery platform does not meet the advertiser's requirements, the advertiser may file a claim with the advertising delivery platform. Order delivery refers to the process in which the advertising delivery platform delivers ads to the corresponding accounts based on the advertiser's requirements. Monitoring and settlement refers to the process in which the advertising delivery platform determines the actual number of ads delivered after placing an order. Resource inquiries and order locking are part of the resource sales phase, while order placement and monitoring and settlement are part of the advertising delivery phase. Because contract advertising involves locking in quantity before delivery, it's not uncommon for ad delivery platforms to deliver fewer ads than advertisers request. In some embodiments, this phenomenon of an ad delivery platform delivering fewer ads than advertisers request is called "shortage."

[0059] The problem of contract advertising shortages manifests itself in two forms:

[0060] a. During the resource selling stage, the advertising exchange platform estimates that the remaining resources of the advertising delivery platform cannot meet the needs of advertisers. In this case, the technical solution provided in the embodiment of the present application can be applied to the advertising exchange platform (Adx).

[0061] b. During the advertising delivery stage, the traffic allocated to the advertiser by the advertising delivery platform fails to meet the advertiser's needs. In this case, the technical solution provided in the embodiment of the present application can be applied to the advertising delivery platform (media).

[0062] To sum up, the technical solution provided in the embodiments of the present application can be used in the event of "shortage" in the business process of contract advertising to reduce the "shortage" of contract advertising, or can be used to adjust the advertising push strategy in time after the "shortage" occurs to maximize the number of advertisers' requirements.

[0063] 2. The technical solution provided by the embodiment of the present application can be applied in the scenario of video push, that is, after the video producer has completed the video production, the completed video is uploaded to the video push platform, and the video push platform pushes the video to different users. If the video producer signs a contract with the video push platform, the contract requires the video push platform to push the video to an agreed number of users. Only when the number of users to which the video push platform pushes the video reaches the agreed number, the video producer will pay the full remuneration to the video push platform. When the video push platform fails to push the video to the agreed number of users, the video producer will not pay the full remuneration and may even file a claim with the video push platform. The technical solution provided by the embodiment of the present application can be applied in the process of video push platform pushing videos, to reduce the difference between the actual number of pushes and the agreed number of the video push platform, or to adjust the push strategy in time when a difference occurs, so as to complete the push target of the agreed number to the greatest extent.

[0064] 3. The technical solution provided by the embodiment of the present application can be applied in the scenario of article push, that is, after the author finishes writing the article, he uploads the completed article to the article push platform, and the article push platform pushes the article to different users. If the author signs a contract with the article push platform, the contract requires the article push platform to push the article to an agreed number of users. Only when the number of users to which the article push platform pushes the article reaches an agreed number, will the author pay the full remuneration to the article push platform. When the article push platform does not push the article to the agreed number of users, the author will not pay the full remuneration and may even file a claim with the article push platform. The technical solution provided by the embodiment of the present application can be applied in the process of article push platform pushing articles, to reduce the difference between the actual number of pushes and the agreed number of the article push platform, or to adjust the push strategy in time when a difference occurs, so as to complete the push target of the agreed number to the greatest extent.

[0065] It should be noted that in the above description, three application scenarios of advertising push, video push and article push are listed. In other possible implementations, the embodiments of the present application can also be applied to other push scenarios, and the embodiments of the present application do not limit this.

[0066] It should be noted that in the following description of the technical solution provided by this application, a server is used as an example of an execution subject. In other possible implementations, the technical solution provided by this application can also be executed by cooperation between a terminal and a server, that is, the server performs background processing and the terminal displays the results of data processing. The embodiments of this application do not limit the type of execution subject.

[0067] Figure 3This is a flowchart of a method for updating an account filtering policy provided by an embodiment of the present application. Figure 3 , methods include:

[0068] 301. The server obtains a first estimated push quantity of a content item to be pushed in a previous statistical period. The first estimated push quantity is the number of accounts that are expected to push the content item obtained in the previous statistical period.

[0069] The content item may be an advertisement, video, or article, which is not limited in this embodiment of the present application. If the content item is an advertisement, then the first estimated number of pushes is the number of accounts that the advertisement push platform estimates can push advertisements based on actual conditions; if the content item is a video, then the first estimated number of pushes is the number of accounts that the video push platform estimates can push videos based on actual conditions; if the content item is an article, then the first estimated number of pushes is the number of accounts that the article push platform estimates can push articles based on actual conditions.

[0070] 302. The server recalls multiple target push accounts for the content item, where the target push accounts are accounts that meet the push conditions of the content item.

[0071] In some embodiments, the push conditions of the content item are conditions for limiting the push accounts. For example, the push conditions of the content item are "Region A" and "Male". Then, in the process of pushing the content item, the server needs to filter the accounts according to the push conditions of the content item and push the content item to the accounts that meet the two conditions of "Shanghai" and "Male".

[0072] 303. The server filters multiple target push accounts using the account filtering policies of multiple account filtering levels to obtain a second estimated push quantity of the content item and a first number of accounts filtered out by each account filtering level.

[0073] The account filtering policy is also a policy for selecting accounts. The server can use the account filtering policy to filter out accounts that cannot push content items from the recalled multiple target push accounts, thereby determining the actual number of accounts that can push content items. In some embodiments, different account filtering levels correspond to different account filtering policies.

[0074] 304. The server obtains an impact parameter of each account filtering layer based on the number of first accounts and the number of second accounts filtered out by each account filtering layer in the previous statistical period. The impact parameter is used to indicate the degree of influence of the corresponding account filtering layer on the change between the second estimated push number and the first estimated push number.

[0075] During each statistical cycle, the server filters the recalled target push accounts using the account filtering layer to determine the estimated push volume for each cycle. Since the estimated push volume is closely related to the number of accounts filtered out by each account filtering layer, the server can determine the reasons for changes in the estimated push volume by measuring the differences in the number of accounts filtered out by the same account filtering layer across different cycles, allowing for targeted adjustments.

[0076] 305. The server updates the account filtering policies of multiple account filtering levels based on the impact parameters of each account filtering level.

[0077] Among them, the impact parameter is a quantitative description of the reason for the change in the expected push quantity in different periods. By updating the filtering strategy of the account filtering level through the impact parameter, the account filtering strategy can be updated quantitatively, and the update effect is better.

[0078] Through the technical solution provided in the embodiments of this application, the content item push process is divided into multiple statistical periods. The expected number of pushes in each statistical period and the number of accounts filtered out by the account filtering layer are monitored in real time, thereby determining the impact parameters of different account filtering layers in real time. The impact parameters can quantify the degree of influence of the account filtering layer on the difference in the expected number of pushes in different statistical periods. Dynamically adjusting the account filtering strategy of each account filtering layer based on the impact parameters can maximize the guarantee that the number of pushed content items reaches the agreed push number, thereby improving the effectiveness of content item push.

[0079] The above steps 301-305 are a brief description of the technical solution provided by the embodiment of the present application. The technical solution provided by the embodiment of the present application will be further described below with reference to some examples. Figure 4 This is a flowchart of a method for updating an account filtering policy provided by an embodiment of the present application. Figure 4 , methods include:

[0080] 401. The server obtains a first estimated push quantity of the content item to be pushed in the last statistical period. The first estimated push quantity is the number of accounts that are expected to push the content item obtained in the last statistical period.

[0081] The description of the content items and the first expected number of pushes is described in the relevant description of step 301 and will not be repeated here. The statistical period is the period during which the server expects the number of accounts to push content items.

[0082] In order to more clearly illustrate the technical solution provided by the embodiments of the present application, the statistical period in the present application will be explained in conjunction with the application scenario of advertising push. In the advertising push scenario, there is still a period of time between the advertiser placing an order and the advertising push platform pushing the advertisement. For example, the advertiser may place an order on January 1, and the advertising push platform will push the advertisement on January 4. Then there is a time difference of 3 days between January 1 and January 4, and the statistical period exists in this 3-day time difference. In some embodiments, the statistical period is one day, that is, the server will count the number of accounts that are expected to push content items every other day. Through multiple statistical periods, the server can predict the results of content item push in advance, so that when the goal of content item push cannot be achieved, corresponding measures can be taken in time to recover losses.

[0083] In a possible implementation, the server can use a statistical table to store the estimated push quantity in different statistical periods. The server obtains the first estimated push quantity from the corresponding statistical table based on an identifier of a previous statistical period.

[0084] For example, see Figure 5 , the server can use Figure 5 The statistical table shown stores the expected push quantity in different statistical periods. D0, D1 and D2 in the statistical table represent the first statistical period, the second statistical period and the third statistical period respectively.

[0085] 402. The server recalls multiple target push accounts for the content item, where the target push accounts are accounts that meet the push conditions of the content item.

[0086] In one possible implementation, the server obtains a push account tag that meets the push conditions of the content item, where the push account tag is used to represent an attribute of the account, and obtains multiple target push accounts whose account tags match the push account tags from multiple accounts.

[0087] In this implementation, the server can match based on push account tags, thereby quickly recalling multiple target push accounts with high efficiency. The server can then filter based on the multiple target push accounts to determine accounts that can push content items.

[0088] The above implementation is described below through several examples.

[0089] Example 1: Taking the content item as an advertisement, when an advertiser places an advertisement push order, they often set push conditions for the advertisement, such as pushing the advertisement to men in Region A. The server obtains the push conditions set by the advertiser, and obtains the push account tags "Region A" and "Male" that meet the push conditions from the push conditions. The server queries the account database based on the push account tags "Region A" and "Male", and the account database stores account tags corresponding to multiple accounts. The server obtains multiple target push accounts whose account tags match the push account tags "Region A" and "Male". For example, there is an account stored in the account database, and the account tags of the account are "Region A", "Male", "25 years old" and "Office Worker", then the server can obtain the account as a target push account.

[0090] Example 2: Taking the content item of a video as an example, when the video author places a video push order, they often set push conditions for the video, such as pushing the video to men who like electronic products. The server obtains the push conditions set by the video author and obtains the push account tags "electronic products" and "male" that meet the push conditions from the push conditions. The server queries the account database based on the push account tags "electronic products" and "male". The account database stores account tags corresponding to multiple accounts. The server obtains multiple target push accounts whose account tags match the push account tags "electronic products" and "male". For example, if there is an account stored in the account database with the tags "A region", "male", "25 years old", and "electronic products", the server can obtain this account as a target push account.

[0091] Example 3: Taking an article as an example, when the author of the article places an order to push the article, they often set push conditions for the article, such as pushing the article to women who love traveling. The server obtains the push conditions set by the article author and obtains the push account tags "travel" and "female" that meet the push conditions from the push conditions. The server queries the account database based on the push account tags "travel" and "female". The account database stores account tags corresponding to multiple accounts. The server obtains multiple target push accounts whose account tags match the push account tags "travel" and "female". For example, if the account database stores an account with the tags "Area A", "Female", "28 years old", and "Travel", the server can obtain this account as a target push account.

[0092] In one possible implementation, the server obtains a target account that meets push conditions for the content item and extracts target account features of the target account, where the target account features are used to represent attributes of the target account. The server then obtains, from multiple accounts, multiple target push accounts whose account features have similarities with the target account features that meet target similarity conditions.

[0093] Under this implementation, the server can recall target push accounts based on account features. Since the recall based on account features is based on the similarity between features, the server can recall a large number of target push accounts, thereby expanding the scope of content item push.

[0094] The above implementation is described below through several examples.

[0095] Example 1: Taking the content item as an advertisement, when the advertiser places an advertisement push order, he or she will often set the push conditions for the advertisement, such as the advertisement needs to be pushed to men in area A. Based on the push conditions set by the advertiser, the server obtains a target account that meets the push conditions, and the target account is also the account of a male in area A. The server obtains the target account feature vector of the target account, and the target account feature vector is used to represent the target account feature. The server determines the cosine similarity between the account feature vectors of multiple accounts and the target account feature vector, and determines the account whose similarity meets the target similarity condition as the target push account. For example, the server obtains a target account A that meets the push conditions and extracts the target account feature vector (1, 0, 1) of the target account A. Taking three accounts as an example, the server obtains the account feature vectors (1, 1, 1), (0, 1, 0) and (1, 1, 0) of the three accounts. The server determines the cosine similarities between the account feature vectors (1, 1, 1), (0, 1, 0), and (1, 1, 0) of the three accounts and the target account feature vector (1, 0, 1) as 0.816, 0, and 0.5, respectively. If the target similarity threshold is 0.8, the server can select the account corresponding to the account feature vector (1, 1, 1) as the target push account.

[0096] Example 2: Taking the content item as a video, when the video author issues a video push order, he or she often sets the push conditions for the video, such as pushing the video to men who like electronic products. Based on the push conditions set by the video author, the server obtains a target account that meets the push conditions. The target account is also the account of a man who likes electronic products. The server obtains the target account feature vector of the target account, which is used to represent the target account characteristics. The server determines the cosine similarity between the account feature vectors of multiple accounts and the target account feature vector, and determines the account whose similarity meets the target similarity condition as the target push account. For example, the server obtains a target account B that meets the push conditions and extracts the target account feature vector (0, 0, 1) of the target account B. Taking three accounts as an example, the server obtains the account feature vectors (1, 0, 1), (0, 1, 0) and (1, 1, 1) of the three accounts. The server determines the cosine similarities between the three account feature vectors (1, 0, 1), (0, 1, 0), and (1, 1, 1) and the target account feature vector (0, 0, 1): 0.707, 0, and 0.577. If the target similarity threshold is 0.7, the server can select the account corresponding to the account feature vector (1, 0, 1) as the target push account.

[0097] Example 3: Taking the content item as an article, when the author of the article places an order to push the article, he or she will often set the push conditions for the article, such as pushing the article to women who love traveling. Based on the push conditions set by the author of the article, the server obtains a target account that meets the push conditions. The target account is also the account of a woman who loves traveling. The server obtains the target account feature vector of the target account, which is used to represent the target account characteristics. The server determines the cosine similarity between the account feature vectors of multiple accounts and the target account feature vector, and determines the account whose similarity meets the target similarity condition as the target push account. For example, the server obtains a target account C that meets the push conditions and extracts the target account feature vector (1, 1, 1) of the target account C. Taking three accounts as an example, the server obtains the account feature vectors (1, 1, 0), (0, 1, 0), and (0, 1, 1) of the three accounts. The server determines the cosine similarities between the three account feature vectors (1, 1, 0), (0, 1, 0), and (0, 1, 1) and the target account feature vector (1, 1, 1) to be 0.816, 0.577, and 0.816, respectively. If the target similarity threshold is 0.8, the server can select the accounts corresponding to the account feature vectors (1, 1, 0) and (0, 1, 1) as the two target push accounts.

[0098] 403. The server filters multiple target push accounts using multiple account filtering strategies at multiple account levels to obtain a second estimated push quantity for the content item and a first number of accounts filtered out by each account filtering level.

[0099] In some embodiments, the account filtering strategies of multiple account filtering levels include at least one of the following: priority filtering, push frequency filtering, freshness filtering, and reallocation filtering, wherein priority filtering refers to filtering accounts that have been assigned to other content items with a higher priority than the content item. Push frequency filtering refers to filtering accounts that have pushed more than a first number threshold within a first target duration. Freshness filtering refers to filtering accounts that have pushed content items more than a second number threshold within a second target duration. Reallocation filtering refers to filtering accounts that have been assigned to other types of content item push services.

[0100] In one possible implementation, the server filters multiple target push accounts using account filtering policies at multiple account filtering levels, obtaining multiple first filtered accounts that are not filtered out and multiple second filtered accounts that are filtered out by each account filtering level. The server determines the number of the multiple first filtered accounts as the second estimated number of pushes, and determines the number of the multiple second filtered accounts as the number of the first accounts filtered out by each account filtering level.

[0101] The following will take an account filtering level corresponding to different account filtering policies as an example to illustrate how the server filters the target push account.

[0102] 1. Regarding priority filtering, in one possible implementation, the server determines the push priority of a content item to be pushed. The server uses an account filtering hierarchy to filter multiple target push accounts based on the push priority of the content item. In response to any target push account being assigned to other content items with a higher push priority than the content item, the server uses the account filtering hierarchy to filter the target push account from the multiple target push accounts.

[0103] For example, in an advertising push scenario, the server or the advertising push platform is set with multiple priorities for advertising push. The higher the priority of the advertisement, the more push resources the advertising push platform will occupy. In some embodiments, under the premise of pushing the same number of advertisements, the higher the priority of the advertisement push, the more fees will be charged. When issuing an advertising push order, the advertiser can select the priority of the advertising push by himself. After the advertiser selects the priority of the advertising push, the server sets the corresponding priority for the advertising push order issued by the advertiser. After obtaining multiple target push accounts based on the advertising push conditions set by the advertiser, the server can filter the multiple target push accounts according to the priority of the advertising push order, and obtain multiple first filtered accounts that are not filtered out and multiple second filtered accounts that are filtered out by the account filtering level. The server determines the number of the multiple first filtered accounts as the second expected push number, and determines the number of the multiple second filtered accounts as the number of first accounts filtered out by the account filtering level.

[0104] For example, the server pushes ads for two advertisers simultaneously, and the time for pushing ads for both advertisers is the same. Advertiser A sets the ad push condition as "males in region A who like electronic products" and selects the ad push priority level 1 (the highest priority). Advertiser B sets the ad push condition as "males in region A who like smartphones" and selects the ad push priority level 3 (the lowest priority). Then, when the server pushes Advertiser A's ads, it can recall multiple first target push accounts based on the ad push condition set by Advertiser A, "males in region A who like electronic products," where the first target account is the account of "males in region A who like smartphones." Based on the number of ad pushes set by Advertiser A, the server determines the corresponding number of first push accounts from the first target accounts and assigns them to Advertiser A. In some embodiments, since Advertiser A has selected the highest ad push priority, Advertiser A's first filtered account is the number of accounts assigned to Advertiser A's ads, and the second expected push number is the number of ad pushes set by Advertiser A. Advertiser A does not have a second filtered account, which means that the number of first accounts for Advertiser A's ads is 0.

[0105] When the server pushes the advertisement of advertiser B, it can recall multiple second target push accounts based on the advertisement push condition set by advertiser B, "males in area A who like smartphones". The second target push account is also the account of "males in area A who like smartphones". Since smartphones are also electronic products, the multiple second target push accounts may include the first push account assigned to advertiser A. For these first push accounts, since the advertisement push priority selected by advertiser A is higher than the advertisement push priority selected by advertiser B, the advertisement push platform will give priority to pushing advertiser A's advertisements to these first push accounts. For advertiser B, the server will filter out these first push accounts from the recalled multiple second target push accounts. According to the number of advertisement pushes set by advertiser B, the server allocates a corresponding number of second push accounts from the filtered second target push accounts to advertiser B. In some embodiments, the first filtered account of advertiser B is also the second target push account after filtering, the second expected push quantity is also the number of the second target push accounts after filtering, the second filtered account through which advertiser B places advertisements is also the second target push account that is filtered out, and the number of the first accounts through which advertiser B places advertisements is also the number of the second target push accounts that are filtered out.

[0106] In addition, if in the statistical period after the current statistical period, advertiser C places an advertising push order, and the advertising push condition set by advertiser C is also "males in region A who like electronic products", the selected advertising push priority is level two. During the advertising push process, the server can recall multiple third target push accounts based on the advertising push condition "males in region A" set by advertiser C. The third target push account is also the account of "males in region A who like electronic products". Of course, since the advertising push priority selected by advertiser C is lower than the advertising push priority selected by advertiser A, after recalling multiple third target push accounts, the server can filter out the accounts that are the same as the first push account from the multiple third target push accounts. Based on the number of advertising pushes set by advertiser C, the server allocates a corresponding number of third push accounts from the filtered third target push accounts to advertiser C. In some embodiments, advertiser C's first filtered account is the filtered third target push account, the second expected number of pushes is the number of filtered third target push accounts, the second filtered account for advertiser C's advertising placement is the filtered third target push account, and the number of first accounts for advertiser C's advertising placement is the number of filtered third target push accounts. Because advertiser C's selected advertising placement priority is higher than advertiser B's selected advertising placement priority, if the previously determined second target push account and the third push account have the same account number, the server will filter the same account number from the second target push account to ensure advertiser C's advertising placement needs are met first. In this case, advertiser B's first filtered account becomes the second target push account after filtering out the first and third push accounts, the second expected number of pushes becomes the number of second target push accounts after filtering out the first and third push accounts, and the second filtered account for advertiser B's advertising placement becomes the first and third push accounts filtered out of the second target push account. The number of first accounts for advertiser B's advertising placement is the sum of the first and third push accounts filtered out of the second target push account.

[0107] 2. Regarding push frequency filtering, in one possible implementation, the server determines the number of content item pushes by multiple target push accounts within a first target duration. The server uses an account filtering layer to filter the multiple target push accounts based on the number of content item pushes by the multiple target push accounts within the first target duration. In response to any target push account having a content item push count greater than or equal to a first count threshold within the first target duration, the server uses the account filtering layer to filter the target push account from the multiple target push accounts.

[0108] For example, in an ad push scenario, to avoid frequent ad pushes that degrade user experience, ad push platforms or servers often set a target duration and a first push threshold. The first push threshold indicates the maximum number of times the server can push ads to a user within the target duration. For example, if the target duration is 7 days and the first push threshold is 3, the server can push ads to the same user a maximum of three times within those 7 days.

[0109] For example, if the ad push condition set by advertiser A is "males in region A who like electronic products," when the server pushes Advertiser A's ad, it can recall multiple target push accounts based on the ad push condition set by advertiser A, "males in region A who like electronic products." The target push account is also the account of "males in region A who like electronic products." The server can determine the number of ad pushes to the multiple target push accounts within the first target duration. In response to the number of ad pushes for any target push account reaching the first number threshold within the first target duration, the server filters out the target push account from the multiple target push accounts to ensure that the filtered target push accounts are all accounts whose number of ad pushes within the first target duration is less than the first number threshold.

[0110] 3. Regarding freshness filtering, in one possible implementation, the server determines the number of times multiple target push accounts have pushed the content item within a second target duration. The server uses the account filtering layer to filter the multiple target push accounts based on the number of times the multiple target push accounts have pushed the content item within the second target duration. In response to any target push account pushing the content item greater than or equal to a second threshold number within the first target duration, the server uses the account filtering layer to filter the target push account from the multiple target push accounts.

[0111] For example, in ad push scenarios, to avoid frequent pushes of the same ad to users, which could negatively impact the advertiser's impression, ad push platforms or servers often set a second target duration and a second count threshold for each ad. The second count threshold indicates the maximum number of times the server can push the same ad to a user within the second target duration. For example, if the second target duration is 7 days and the second count threshold is 2, this means the server can push the same ad to the same user a maximum of twice within 7 days.

[0112] For example, if the ad push condition set by advertiser A is "males in region A who like electronic products", when the server pushes the ad of advertiser A, it can recall multiple target push accounts based on the ad push condition set by advertiser A, "males in region A who like electronic products", and the target push account is also the account of "males in region A who like electronic products". The server can determine the number of times the ad of advertiser A is pushed to multiple target push accounts within the second target duration. In response to the number of times any target push account is pushed the ad of advertiser A within the second target duration reaching the second number threshold, the server filters out any target push account from the multiple target push accounts to ensure that the filtered target push accounts are all accounts that push the ad of advertiser A less than the second number threshold within the second target duration.

[0113] 4. Regarding reallocation filtering, in one possible implementation, the server determines service type information for multiple target push accounts, where the service type information indicates the service type to which the corresponding target push account is assigned. In response to any target push account being assigned to a different type of content item push service, the server filters the target push account from the multiple target push accounts using the account filtering hierarchy.

[0114] For example, in an advertising push scenario, in order to provide advertisers with a richer range of advertising push methods, an advertising push platform or server often sets up multiple types of advertising push services, such as a pay-per-view (CPT) advertising push service, a pay-per-click (CPC) advertising push service, or an advertising push service based on the amount of advertising displayed. Different types of advertising push services also have different priorities. An advertising push service classified as a high priority can occupy the number of accounts of an advertising push service with a low priority. The priorities of different types of advertising push services are set by technical personnel based on actual conditions, and this embodiment of the application does not limit this.

[0115] For example, if Advertiser A selects a pay-per-click advertising service, and the pay-per-click advertising service has a higher priority than the pay-per-impression advertising service, the server can determine the advertising services to which the multiple target push accounts are assigned after recalling the multiple target push accounts based on the advertising push conditions set by Advertiser A. In response to any target push account being assigned to the pay-per-click advertising service, the server can filter out the target push account from the multiple target push accounts to ensure that the higher-priority advertising service is executed first.

[0116] It should be noted that, in the above description, an account filtering level corresponding to different account filtering policies is used as an example to illustrate the method by which the server filters the target push account. In an embodiment of the present application, the server can use multiple account filtering levels corresponding to different account filtering policies to filter the target push account, and the method is as follows.

[0117] In one possible implementation, a server determines a push priority for a content item to be pushed. The server filters multiple target push accounts based on the push priority of the content item using a first account filtering level. In response to any target push account being assigned to a content item with a higher push priority than the content item, the server filters the target push account from the multiple target push accounts using the first account filtering level. The server determines the number of content item pushes by the multiple target push accounts after the first filtering within a first target duration. The server filters the multiple target push accounts after the first filtering based on the number of content item pushes by the multiple target push accounts after the first filtering within the first target duration using a second account filtering level. In response to any content item push count by any target push account within the first target duration being greater than or equal to a first count threshold, the server filters the target push account from the multiple target push accounts after the first filtering using the second account filtering level. The server determines the number of content item pushes by the multiple target push accounts after the second filtering within a second target duration. The server filters the multiple target push accounts after the second filtering based on the number of content item pushes by the multiple target push accounts after the second filtering within the second target duration. In response to any target push account pushing the content item a number of times within the first target duration being greater than or equal to a second number threshold, the server filters out the target push account from the multiple target push accounts after the second filtering, using a third account filtering level. The server determines business type information for the multiple target push accounts after the third filtering, where the business type information indicates the business type to which the corresponding target push account is classified. In response to any target push account being classified as another type of content item push service, the server filters out the target push account from the multiple target push accounts after the third filtering, using a fourth account filtering level.

[0118] Under this implementation, the server can filter the target push accounts through multiple account filtering levels, and finally obtain the target push accounts that meet the content item push requirements, avoiding content item push exceptions due to business conflicts during subsequent content item push processes, and improving the efficiency of content item push.

[0119] For example, see Figure 6The embodiment of the present application provides a traffic distribution funnel 601, which includes multiple account filtering levels 602. The area of ​​the account filtering level 602 can represent the number of target push accounts after filtering. It can be seen that as the account filtering level 602 increases, the number of target push accounts after filtering continues to decrease.

[0120] Optionally, if the current statistical period is the last statistical period, the server can execute the following step 404 ; if the current statistical period is not the last statistical period, the server can execute the following step 405 .

[0121] 404. The server pushes the content item to the filtered multiple target push accounts.

[0122] Among them, if the content item is an advertisement, the server will push the advertisement to the multiple filtered target push accounts; if the content item is a video, the server will push the video to the multiple filtered target push accounts; if the content item is an article, the server will push the article to the multiple filtered target push accounts.

[0123] 405. The server obtains an impact parameter of each account filtering layer based on the number of first accounts and the number of second accounts filtered out by each account filtering layer in the previous statistical period. The impact parameter is used to indicate the degree of influence of the corresponding account filtering layer on the change between the second estimated push number and the first estimated push number.

[0124] In one possible implementation, the server obtains the difference between the first number of accounts and the corresponding second number of accounts at each account filtering level. Based on the change value and the difference between the number of accounts at each account filtering level, the server obtains the impact parameter of each account filtering level. In some embodiments, if the server adopts Figure 5 The form shown above stores the expected number of pushes in different statistical periods and the number of accounts filtered out by each account filtering level in different statistical periods. Then the server can make a difference between the values ​​of the corresponding entries in the statistical table of different periods and obtain the following: Figure 7 The difference statistics table shown stores the change values ​​between the expected push numbers in different statistical periods and the difference values ​​between the number of accounts filtered out by each account filtering level in different statistical periods. In some embodiments, Δ01 is used to represent the difference value in the number of accounts between the first statistical period and the second statistical period, and Δ12 is used to represent the difference value in the number of accounts between the second statistical period and the third statistical period. The change value between the expected push numbers can be called "shortage", and the difference value between the number of accounts filtered out by each account filtering level in different statistical periods can be called "reason for shortage".

[0125] The following uses an advertisement push scenario as an example to illustrate the principle of the above implementation method.

[0126] In the ad push scenario, the lifecycle of an ad order is defined as the period from when an advertiser places an order to when an ad is placed on an ad delivery platform. Within this lifecycle, the difference in the expected number of pushes between any two statistical periods is related to the number of accounts filtered out by the account filtering layer. The impact parameter of the account consideration layer quantifies the degree to which the account filtering layer affects the difference in the expected number of pushes. This is illustrated using the first and last statistical periods, that is, when the advertiser places an order and when the ad delivery platform places the ad.

[0127] When an advertiser places an order, if the server recalls 100 target push accounts based on the push conditions set by the advertiser, and 10 of these target push accounts are classified into higher-priority advertising orders, then these 10 (10%) target push accounts are filtered out by the first account filtering level, and the remaining 90 target push accounts are also the second filtered accounts filtered out by the first account filtering level. If, among these 90 target push accounts, there are 5 target push accounts whose number of ad pushes within the first target duration is greater than or equal to the first count threshold, then these 5 (5%) target push accounts are also filtered out by the second account filtering level, and the remaining 85 target push accounts are also the second filtered accounts filtered out by the second account filtering level. If, among these 85 target push accounts, there are 3 target push accounts whose number of times the content item is pushed to them within the second target duration is greater than or equal to the second number threshold, the server will filter out these 3 (3%) target push accounts through the third account filtering level, leaving 82 target push accounts. These 3 target push accounts are also the second filtered accounts filtered out by the third account filtering level. If, among these 82 target push accounts, there are two target push accounts that are classified as other types of advertising push services, then the server can filter out these two (2%) target push accounts through the fourth account filtering level, ultimately obtaining 80 target push accounts. These 80 target push accounts are also the first filtered accounts that have not been expanded and filtered out. These 80 target push accounts are also the quantity required by advertisers when placing orders. These two target push accounts are also the second filtered accounts filtered out by the fourth account filtering level.

[0128] When an ad delivery platform delivers ads, if the server recalls 100 target push accounts based on the push conditions set by the advertiser, and 40 of these target push accounts are classified into higher-priority advertising orders, then these 40 (40%) target push accounts are filtered out by the first account filtering level, and the remaining 60 target push accounts are also the second filtered accounts filtered out by the first account filtering level. If, among these 60 target push accounts, there are 6 target push accounts whose number of ad pushes within the first target duration is greater than or equal to the first count threshold, then these 6 (6%) target push accounts are also filtered out by the second account filtering level, and the remaining 54 target push accounts are also the second filtered accounts filtered out by the second account filtering level. If, among these 54 target push accounts, there are 4 target push accounts whose number of times the content item is pushed to them within the second target duration is greater than or equal to the second number threshold, the server will filter out these 4 (4%) target push accounts through the third account filtering level, and the remaining 50 target push accounts, these 4 target push accounts are also the second filtered accounts filtered out by the third account filtering level. If, among these 50 target push accounts, there are two target push accounts that are classified as other types of advertising push services, then the server can filter out these two (2%) target push accounts through the fourth account filtering level, and finally obtain 48 target push accounts. These 48 target push accounts are also the first filtered accounts that have not been expanded and filtered out. These 48 target push accounts are also the actual number of advertisements delivered by the advertising delivery platform. These 2 target push accounts are also the second filtered accounts filtered out by the fourth account filtering level.

[0129] Based on the above information, we can know that when the advertiser places an order and when the ad delivery platform delivers the ad, under the premise of recalling the same 100 target push accounts, the number of accounts filtered out by the first account filtering level (priority filtering) increases from 10 to 40, and the proportion of the 100 target push accounts increases from 10% to 40%. The fluctuation is relatively high, exceeding the fluctuation of the number of accounts filtered out by other account filtering levels. Therefore, the server can determine that the first account filtering level (priority filtering) is the main reason why the actual number of ads delivered by the ad delivery platform is 48 less than the required number of 80 when the advertiser places an order. During the experiment, see Figure 8 A correlation analysis of the account filtering strategies at multiple account filtering levels and the shortage (the difference between the actual quantity and the required quantity) shows that there is a high correlation between the number of accounts filtered out by multiple account filtering levels and the shortage, which further confirms the above explanation.

[0130] Under this implementation, the server can determine the reasons for the differences in the expected number of pushes in different statistical periods based on the difference in the number of accounts filtered out by each account filtering level in different statistical periods, thereby facilitating subsequent updates to the account filtering policies of each account filtering level.

[0131] For example, the server uses a linear attribution model to perform a linear fit on the change value and the difference in the number of accounts at each account filtering level to obtain the impact parameter of each account filtering level. For example, the server can obtain the impact parameter of each account filtering level using formula (1).

[0132] Δalloc=ω1Δx1+ω2Δx2+…+ωnΔxn(1)

[0133] Among them, Δ alloc is the change value, ω1-ω n are the influencing parameters of different account filtering levels. In some embodiments, ω1+ω2+…+ω n =1, Δx1-Δx n The number of accounts filtered out for different account filtering levels.

[0134] 406. The server updates the account filtering policies of multiple account filtering levels based on the impact parameters of each account filtering level.

[0135] In a possible implementation, in response to an impact parameter of any account filtering level being greater than or equal to an impact parameter threshold, the server updates the account filtering policy of any account filtering level to reduce the number of accounts filtered out by the any account filtering level.

[0136] Under this implementation, the server can update the account filtering policy of the account filtering level according to the influencing parameters of different account filtering levels to ensure that the number of target push accounts after filtering meets the requirements to the greatest extent possible and improve the effect of subsequent content item push.

[0137] For example, if the server determines that the impact parameter of the first account filtering level (priority filtering) is 0.9, then the server can determine that the first account filtering level has the greatest impact on the change between the second expected number of pushes and the first expected number of pushes. The server can then adjust the account filtering policy of the first account filtering level. For example, the original account filtering policy is that target push accounts assigned to the second level (medium priority) will be filtered, and the adjusted account filtering policy is that target push accounts classified as the first level (high priority) will be filtered. Through such adjustments, the number of accounts filtered out by the first account filtering level can be reduced, thereby reducing the change between the second expected number of pushes and the first expected number of pushes. Optionally, in this case, the server can also adjust the priority of the content item instead of adjusting the account filtering policy of the first account filtering level. For example, if the original priority of the content item is the second level (medium priority), the server can adjust the priority of the content item to the first level (high priority). This embodiment of the present application is not limited to this.

[0138] All of the above optional technical solutions can be combined in any way to form optional embodiments of the present application, and will not be described in detail here.

[0139] Through the technical solution provided in the embodiments of this application, the content item push process is divided into multiple statistical periods. The expected number of pushes in each statistical period and the number of accounts filtered out by the account filtering layer are monitored in real time, thereby determining the impact parameters of different account filtering layers in real time. The impact parameters can quantify the degree of influence of the account filtering layer on the difference in the expected number of pushes in different statistical periods. Dynamically adjusting the account filtering strategy of each account filtering layer based on the impact parameters can maximize the guarantee that the number of pushed content items reaches the agreed push number, thereby improving the effectiveness of content item push.

[0140] Figure 9 This is a schematic diagram of the structure of an update device for an account filtering policy provided by an embodiment of the present application, see Figure 9 The device includes: a push quantity acquisition module 901, a recall module 902, a filtering module 903, an impact parameter acquisition module 904 and a strategy update module 905.

[0141] The push quantity acquisition module 901 is used to acquire a first estimated push quantity of the content item to be pushed in the last statistical period. The first estimated push quantity is the number of accounts that are expected to push the content item obtained in the last statistical period.

[0142] The recall module 902 is used to recall multiple target push accounts of the content item, where the target push accounts are accounts that meet the push conditions of the content item.

[0143] The filtering module 903 is configured to filter multiple target push accounts using the account filtering policies of multiple account filtering levels to obtain a second estimated push quantity of the content item and a first number of accounts filtered out by each account filtering level.

[0144] The impact parameter acquisition module 904 is used to obtain the impact parameter of each account filtering level based on the number of first accounts and the number of second accounts filtered out by each account filtering level in the previous statistical period. The impact parameter is used to indicate the degree of influence of the corresponding account filtering level on the change value between the second expected push number and the first expected push number.

[0145] The policy updating module 905 is configured to update the account filtering policies of multiple account filtering levels based on the impact parameters of each account filtering level.

[0146] In one possible implementation, the recall module is configured to obtain a push account tag that meets the push conditions of the content item, where the push account tag represents an attribute of the account, and obtain multiple target push accounts whose account tags match the push account tags from multiple accounts.

[0147] In one possible implementation, the filtering module is configured to filter multiple target push accounts using account filtering policies at multiple account filtering levels to obtain multiple first filtered accounts that are not filtered out and multiple second filtered accounts that are filtered out by each account filtering level. The number of the multiple first filtered accounts is determined as the second expected number of pushes. The number of the multiple second filtered accounts is determined as the number of the first accounts that are filtered out by each account filtering level.

[0148] In one possible implementation, the impact parameter acquisition module is configured to acquire the difference between the first number of accounts and the corresponding second number of accounts at each account filtering level, and acquire the impact parameter for each account filtering level based on the change value and the difference in the number of accounts at each account filtering level.

[0149] In a possible implementation, the impact parameter acquisition module is used to perform linear fitting on the change value and the difference in the number of accounts at each account filtering level through a linear attribution model to obtain the impact parameter of each account filtering level.

[0150] In one possible implementation, the policy update module is configured to update the account filtering policy of any account filtering level in response to an impact parameter of any account filtering level being greater than or equal to an impact parameter threshold, so as to reduce the number of accounts filtered out by any account filtering level.

[0151] In a possible implementation, the current statistical period is the last statistical period, and the apparatus further includes:

[0152] The push module is used to push content items to multiple filtered target push accounts.

[0153] In one possible implementation, the account filtering policies at multiple account filtering levels include at least one of the following:

[0154] Priority filtering, push frequency filtering, freshness filtering, and redistribution filtering.

[0155] Priority filtering refers to filtering accounts that have been assigned to other content items with a higher priority than the content item.

[0156] Push frequency filtering means filtering out accounts whose push count exceeds the first count threshold within the first target duration.

[0157] Freshness filtering refers to filtering out accounts that push content items more than a second threshold number of times within the second target duration.

[0158] Reassignment filtering refers to filtering accounts that have been assigned to other types of content item push services.

[0159] Through the technical solution provided in the embodiments of this application, the content item push process is divided into multiple statistical periods. The expected number of pushes in each statistical period and the number of accounts filtered out by the account filtering layer are monitored in real time, thereby determining the impact parameters of different account filtering layers in real time. The impact parameters can quantify the degree of influence of the account filtering layer on the difference in the expected number of pushes in different statistical periods. Dynamically adjusting the account filtering strategy of each account filtering layer based on the impact parameters can maximize the guarantee that the number of pushed content items reaches the agreed push number, thereby improving the effectiveness of content item push.

[0160] The above-mentioned computer device can also be implemented as a server. The structure of the server is introduced below:

[0161] Figure 10 This is a structural diagram of a server provided in an embodiment of the present application. The server 1000 may have relatively large differences due to different configurations or performances, and may include one or more processors (Central Processing Units, CPU) 1001 and one or more memories 1002, wherein the one or more memories 1002 store at least one computer program, and the at least one computer program is loaded and executed by the one or more processors 1001 to implement the methods provided in the above-mentioned various method embodiments. Of course, the server 1000 may also have components such as a wired or wireless network interface, a keyboard, and an input and output interface for input and output. The server 1000 may also include other components for implementing device functions, which will not be described in detail here.

[0162] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory device including a computer program. The computer program can be executed by a processor to implement the account filtering policy update method described in the above embodiment. For example, the computer-readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a compact disc (CD-ROM), a magnetic tape, a floppy disk, or an optical data storage device.

[0163] In an exemplary embodiment, a computer program product or computer program is also provided, which includes a program code, which is stored in a computer-readable storage medium. The processor of a computer device reads the program code from the computer-readable storage medium, and the processor executes the program code, so that the computer device executes the above-mentioned account filtering policy updating method.

[0164] Those skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware, or may be accomplished by a program to instruct the relevant hardware, and the program may be stored in a computer-readable storage medium, and the above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc.

[0165] The above are only optional embodiments of the present application and are not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should be included in the scope of protection of the present application.

Claims

1. A method for updating an account filtering policy, characterized in that: The method comprises: Obtaining a first estimated push quantity of the content item to be pushed in a previous statistical period, where the first estimated push quantity is the number of accounts that are estimated to push the content item obtained in the previous statistical period; Recalling multiple target push accounts for the content item, where the target push accounts are accounts that meet the push conditions for the content item; filtering the multiple target push accounts using the account filtering policies of the multiple account filtering levels to obtain a second estimated push quantity for the content item and a first quantity of accounts filtered out by each of the account filtering levels; Obtaining the difference in the number of accounts between the first number of accounts at each of the account filtering levels and the second number of accounts filtered out by each of the account filtering levels in the previous statistical period; Using a linear attribution model, linearly fit the change between the second estimated number of push notifications and the first estimated number of push notifications, as well as the difference in the number of accounts at each account filtering level, to obtain an impact parameter for each account filtering level, where the impact parameter represents the degree of impact of the corresponding account filtering level on the change between the second estimated number of push notifications and the first estimated number of push notifications. In response to an impact parameter of any account filtering level being greater than or equal to an impact parameter threshold, the account filtering policy of any account filtering level is updated to reduce the number of accounts filtered out by the any account filtering level.

2. The method according to claim 1, characterized in that The multiple target push accounts for recalling the content item include: Obtaining a push account tag that meets the push conditions of the content item, wherein the push account tag is used to represent an attribute of the account; The multiple target push accounts whose account tags match the push account tags are obtained from the multiple accounts.

3. The method according to claim 1, characterized in that The filtering of the multiple target push accounts by the account filtering strategies of the multiple account filtering levels to obtain the second estimated push quantity of the content item and the first number of accounts filtered out by each of the account filtering levels includes: Filtering the multiple target push accounts using the account filtering policies of the multiple account filtering levels to obtain multiple first filtered accounts that are not filtered out and multiple second filtered accounts that are filtered out by each of the account filtering levels; determining the number of the plurality of first filtering accounts as the second estimated push number; The number of the plurality of second filtering accounts is determined as the number of the first accounts filtered out by each of the account filtering levels.

4. The method according to claim 1, wherein The current statistical period is the last statistical period. After filtering the multiple target push accounts using the account filtering policies of the multiple account filtering levels to obtain a second estimated push quantity for the content item and a first number of accounts filtered out by each of the account filtering levels, the method further includes: The content item is pushed to the filtered multiple target push accounts.

5. The method according to claim 1, characterized in that The account filtering policies of the multiple account filtering levels include at least one of the following: Priority filtering, push frequency filtering, freshness filtering, and redistribution filtering; The priority filtering refers to filtering accounts that have been assigned to other content items with a higher priority than the content item; The push frequency filtering refers to filtering accounts whose push times exceed the first number threshold within the first target duration; The freshness filtering refers to filtering out accounts that have pushed the content item more than a second number threshold number of times within the second target duration; The reallocation filtering refers to filtering accounts that have been allocated to other types of content item push services.

6. An account filtering policy updating device, characterized in that: The device comprises: A push quantity acquisition module is used to obtain a first estimated push quantity of the content item to be pushed in the last statistical period, where the first estimated push quantity is the number of accounts that are expected to push the content item obtained in the last statistical period; A recall module, configured to recall multiple target push accounts for the content item, wherein the target push accounts are accounts that meet the push conditions for the content item; a filtering module, configured to filter the plurality of target push accounts using account filtering policies of a plurality of account filtering levels, to obtain a second estimated push quantity for the content item and a first number of accounts filtered out by each of the account filtering levels; an impact parameter acquisition module, configured to obtain the difference in account number between the first number of accounts at each account filtering level and the second number of accounts filtered out by each account filtering level in the previous statistical period; and to perform a linear fit on the change between the second estimated number of push notifications and the first estimated number of push notifications, as well as the difference in account number at each account filtering level, using a linear attribution model to obtain an impact parameter for each account filtering level, the impact parameter being used to indicate the degree of influence of the corresponding account filtering level on the change between the second estimated number of push notifications and the first estimated number of push notifications. The policy updating module is configured to update the account filtering policy of any account filtering layer in response to an influence parameter of any account filtering layer being greater than or equal to an influence parameter threshold, so as to reduce the number of accounts filtered out by the any account filtering layer.

7. The device according to claim 6, characterized in that The recall module is used to obtain a push account tag that meets the push conditions of the content item, where the push account tag is used to represent the attributes of the account; and obtain the multiple target push accounts whose account tags match the push account tags from multiple accounts.

8. The device according to claim 6, characterized in that The filtering module is configured to filter the multiple target push accounts using the account filtering policies of the multiple account filtering levels to obtain multiple first filtered accounts that are not filtered out and multiple second filtered accounts that are filtered out by each of the account filtering levels; determining the number of the plurality of first filtering accounts as the second estimated push number; The number of the plurality of second filtering accounts is determined as the number of the first accounts filtered out by each of the account filtering levels.

9. A computer device, characterized in that: The computer device includes one or more processors and one or more memories, and at least one computer program is stored in the one or more memories. The computer program is loaded and executed by the one or more processors to implement the account filtering policy updating method according to any one of claims 1 to 5.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores at least one computer program, and the computer program is loaded and executed by the processor to implement the method for updating the account filtering policy according to any one of claims 1 to 5.

Images