Methods, devices, and electronic equipment for configuring quantity-protected quotas in coarse sorting

Abstract

This application provides a method, apparatus, and electronic device for configuring guaranteed quotas in coarse ranking. The method includes: obtaining the number of coarse rankings for each recall channel and the target parameters for each recall channel during the business recommendation process; for each recall channel, determining the business conversion rate of the recall channel as the ratio of the number of coarse rankings to the target parameters of the recall channel; and, if the difference between the business conversion rates of two recall channels is greater than a preset threshold, increasing the guaranteed quota for a first type of recall channel and decreasing the guaranteed quota for a second type of recall channel. The first type of recall channel is one whose business conversion rate is greater than the overall conversion rate, and the second type of recall channel is one whose business conversion rate is less than the overall conversion rate. This application adjusts the guaranteed quota for each recall channel based on the current business conversion rate of the recall channel, which can improve the ranking and filtering effect of the recommendation system in the coarse ranking stage.

Description

Technical Field

[0001] This application relates to the field of recommendation technology, and in particular to a method, apparatus and electronic device for configuring quantity quotas in coarse sorting. Background Technology

[0002] Recommendation systems utilize e-commerce websites to provide customers with product information and suggestions, helping users decide what products to buy. The core of a recommendation system is the recommendation algorithm, which determines which content is recommended to the user.

[0003] Currently, the recommendation process of recommendation systems typically involves three stages: recall, coarse ranking, and fine ranking. In the recall stage, multiple recall channels are often used for processing. Because the scoring in the coarse ranking stage is much higher than in the fine ranking stage, its model structure is usually relatively simple and has a single prediction target for computational performance reasons. The fine ranking stage, however, considers more prediction targets and performs multi-target fusion scoring and ranking based on the final business-oriented indicators. Therefore, the scoring results in the coarse ranking stage often differ from the final business indicators, resulting in relatively poor ranking and filtering performance. When the ranking and filtering performance in the coarse ranking stage is unsatisfactory, the current practice is to set a quota for each of the multiple recall channels during the coarse ranking stage, limiting the number of recalled content items that have passed the coarse ranking stage in each channel.

[0004] However, the above approach requires experienced business personnel to determine the guaranteed volume quota for each recall channel based on their own experience. Even with rich experience, the guaranteed volume quota cannot be guaranteed, which can improve the sorting and screening effect in the coarse-sorting stage. Summary of the Invention

[0005] This application provides a method, apparatus, and electronic device for configuring quantity-preserving quotas in coarse-sorting, so as to at least solve the technical problem in the prior art that configuring quantity-preserving quotas cannot guarantee the improvement of sorting and filtering effects in the coarse-sorting stage.

[0006] According to one aspect of the embodiments of this application, a method for configuring quantity-preserving quotas in coarse sorting is provided, the method comprising:

[0007] In the process of obtaining business recommendations, the number of coarse rankings for each recall channel and the target parameters for each recall channel are as follows: the number of coarse rankings is the number of target businesses that pass the coarse ranking among the businesses recalled by the recall channel, and the target parameters are parameters determined based on the user's target operation on the target businesses that pass the fine ranking.

[0008] For each of the aforementioned recall channels, the ratio of the number of coarse-ranked results for the recall channel to the target parameter of the recall channel is determined as the business conversion rate of the recall channel.

[0009] If the difference between the business conversion rates of two recall channels in a plurality of recall channels exceeds a preset threshold, the guaranteed quota of the first type of recall channel is increased and the guaranteed quota of the second type of recall channel is decreased. The first type of recall channel is a recall channel with a business conversion rate greater than the overall conversion rate, and the second type of recall channel is a recall channel with a business conversion rate less than the overall conversion rate. The overall conversion rate is the ratio of the sum of the coarse-ranked numbers of each recall channel to the sum of the target parameters of each recall channel. The guaranteed quota is used to indicate the number of businesses that pass the coarse-ranking.

[0010] Optionally, after increasing the guaranteed volume quota for the first type of recall channel and decreasing the guaranteed volume quota for the second type of recall channel, the method further includes:

[0011] At preset intervals, the business conversion rate of each recall channel under the current guaranteed quota is determined once;

[0012] If, in any instance, the difference between the business conversion rates of two of the recall channels under the current guaranteed quota is greater than a preset threshold, the guaranteed quota of the first type of recall channel is increased, and the guaranteed quota of the second type of recall channel is decreased, until the difference between the business conversion rates of any two of the recall channels under the current guaranteed quota is less than or equal to the preset threshold.

[0013] Optionally, before obtaining the coarse-ranking count for each recall channel and the target parameters for each recall channel during the business recommendation process, the method further includes:

[0014] Set the guaranteed quota for each recall channel to a target value so that there is no limit to the number of coarse rankings in the business recommendation process.

[0015] Optionally, the sum of the increases in the guaranteed quota for each recall channel of the first type is equal to the sum of the decreases in the guaranteed quota for each recall channel of the second type.

[0016] Optionally, for a target recall channel, the increase or decrease in the guaranteed quota of the target recall channel is equal to the product of the coarse ranking quantity of the target recall channel multiplied by a preset step size value and a target absolute value, wherein the target absolute value is the absolute value of the difference between the business conversion rate of the target recall channel and the overall conversion rate, and the target recall channel is any of the recall channels.

[0017] Optionally, the target parameters include: number of seedings, number of purchases, or content consumption time.

[0018] Optionally, in the process of obtaining service recommendations, the number of coarse-ranked results for each recall channel and the target parameters for each recall channel include:

[0019] Obtain recall and exposure logs during the business recommendation process;

[0020] Based on the recall log and exposure log, the number of coarse-ranking results for each recall channel and the target parameters for each recall channel are determined.

[0021] According to another aspect of the embodiments of this application, a configuration device for quantity-preserving quotas in coarse sorting is provided, the device comprising:

[0022] The acquisition module is used to acquire the number of coarse-ranked target services for each recall channel and the target parameters for each recall channel during the business recommendation process. The number of coarse-ranked target services is the number of target services that pass the coarse-ranking among the services recalled by the recall channel, and the target parameters are parameters determined based on the user's target operation on the target services that pass the fine-ranking.

[0023] The first determining module is used to determine the business conversion rate of each recall channel by the ratio of the number of coarse-sorted results of the recall channel to the target parameter of the recall channel.

[0024] The first configuration module is used to increase the guaranteed quota of a first type of recall channel and decrease the guaranteed quota of a second type of recall channel when the difference between the business conversion rates of two recall channels in the plurality of recall channels is greater than a preset threshold. The first type of recall channel is a recall channel with a business conversion rate greater than the overall conversion rate, and the second type of recall channel is a recall channel with a business conversion rate less than the overall conversion rate. The overall conversion rate is the ratio of the sum of the coarse-ranked numbers of each recall channel to the sum of the target parameters of each recall channel. The guaranteed quota is used to indicate the number of services that pass the coarse-ranking.

[0025] Optionally, the device further includes:

[0026] The second determining module is used to determine the business conversion rate of each of the recall channels under the current guaranteed quota every preset time interval;

[0027] The second configuration module is used to increase the guaranteed quota of the first type of recall channel and decrease the guaranteed quota of the second type of recall channel when the difference between the business conversion rates of two recall channels under the current guaranteed quota is greater than a preset threshold, until the difference between the business conversion rates of any two recall channels under the current guaranteed quota is less than or equal to the preset threshold.

[0028] Optionally, the device further includes:

[0029] The third configuration module is used to set the guaranteed quota for each recall channel to the target value so that there is no limit to the number of coarse rankings in the business recommendation process.

[0030] Optionally, the sum of the increases in the guaranteed quota for each recall channel of the first type is equal to the sum of the decreases in the guaranteed quota for each recall channel of the second type.

[0031] Optionally, for a target recall channel, the increase or decrease in the guaranteed quota of the target recall channel is equal to the product of the coarse ranking quantity of the target recall channel multiplied by a preset step size value and a target absolute value, wherein the target absolute value is the absolute value of the difference between the business conversion rate of the target recall channel and the overall conversion rate, and the target recall channel is any of the recall channels.

[0032] Optionally, the target parameters include: number of seedings, number of purchases, or content consumption time.

[0033] Optionally, the acquisition module includes:

[0034] The acquisition unit is used to acquire recall logs and exposure logs during the business recommendation process;

[0035] The determining unit is used to determine the number of coarse-ranking results for each recall channel and the target parameters for each recall channel based on the recall log and the exposure log.

[0036] According to another aspect of the embodiments of this application, an electronic device is provided, including: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the quantity-guaranteed quota configuration method as described above in coarse sorting.

[0037] According to another aspect of the embodiments of this application, a readable storage medium is provided, which, when the instructions in the readable storage medium are executed by a processor of an electronic device, enables the electronic device to perform the quantity-preserving quota configuration method in coarse sorting as described above.

[0038] In this embodiment, the business conversion rate of each recall channel can be determined by the ratio of the number of coarse-ranked services to the target parameter during the business recommendation process. The number of coarse-ranked services refers to the number of target businesses recalled by the recall channel and passing the coarse-ranking stage. The target parameter is a parameter determined based on the user's target operation on the target businesses that passed the fine-ranking stage. Therefore, the business conversion rate of the recall channel can, to a certain extent, characterize the ranking and filtering effect of the coarse-ranking stage. When the difference between the business conversion rates of two recall channels is greater than a preset threshold, it indicates that some recall channels (i.e., the first type of recall channels) have a better ranking and filtering effect in the coarse-ranking stage, while some recall channels (i.e., the second type of recall channels) have a poorer ranking and filtering effect. In this case, by increasing the guaranteed quota of the first type of recall channel, the number of businesses that meet user needs passing the coarse-ranking stage can be increased; by decreasing the guaranteed quota of the second type of recall channel, the number of businesses that do not meet user needs passing the coarse-ranking stage can be reduced, allowing more businesses that meet user needs to pass through the coarse-ranking stage, thereby improving the ranking and filtering effect of the recommendation system in the coarse-ranking stage. This application embodiment dynamically adjusts the guaranteed quota of each recall channel based on the current business conversion rate of the recall channel. This not only ensures and improves the sorting and filtering effect of the recommendation system in the coarse ranking stage, but also improves the final recommendation effect of the recommendation system, recommending services that are more likely to meet the user's needs to the user. Attached Figure Description

[0039] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0040] Figure 1 A flowchart illustrating the steps of the method for configuring quantity-preserving quotas in coarse sorting provided in this application embodiment;

[0041] Figure 2 A schematic diagram illustrating the practical application of the method for configuring quantity-preserving quotas in coarse sorting provided in this application embodiment;

[0042] Figure 3 A structural block diagram of the configuration device for maintaining quantity quotas in coarse sorting provided in the embodiments of this application. Detailed Implementation

[0043] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0044] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments.

[0045] In the various embodiments of this application, it should be understood that the sequence number of each process described below does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0046] See Figure 1 This application provides a method for configuring quantity-preserving quotas in coarse sorting, the method comprising:

[0047] Step 101: Obtain the number of coarse rankings for each recall channel and the target parameters for each recall channel during the business recommendation process.

[0048] In this step, multiple recall channels exist during the business recommendation process. Different channels recall a large number of identical or different businesses based on different targets. The recalled businesses then undergo coarse and fine ranking processes, and the remaining businesses are recommended to the user. Here, "business" can be understood as a business entry point. For example, in the advertising business, the recommended advertising business to the user can be a hyperlink within the advertising business, allowing the user to click on the hyperlink to purchase the corresponding product. Of course, "business" here is not limited to advertising businesses; it can be any form of data pushed to the user for user interaction.

[0049] Regarding the coarse ranking process in the recommendation process, we won't go into too much detail here. It can be understood that the target business refers to the business that has passed the coarse ranking stage, from input to fine ranking. In other words, the target business is the business recalled by the recall channel and has passed the coarse ranking. Therefore, the coarse ranking count is the number of target businesses that passed the coarse ranking among the businesses recalled by the recall channel.

[0050] The target parameters are determined based on the user's targeted actions in the target business segment, which are then refined. The specific values ​​of the target parameters vary depending on the recommendation scenario. Continuing with advertising as an example, target parameters could include impressions, clicks, and favorites.

[0051] Step 102: For each recall channel, the ratio of the number of coarse-ranked results for the recall channel to the target parameter of the recall channel is determined as the business conversion rate of the recall channel.

[0052] In this step, each recall channel corresponds to a business conversion rate, which is equal to the quotient obtained by dividing the target parameter by the number of coarse-ranked results.

[0053] Step 103: If the difference between the business conversion rates of two recall channels in multiple recall channels is greater than a preset threshold, increase the guaranteed quota of the first type of recall channel and decrease the guaranteed quota of the second type of recall channel.

[0054] In this step, the first type of recall channel is the one with a business conversion rate greater than the overall conversion rate, and the second type of recall channel is the one with a business conversion rate less than the overall conversion rate. The overall conversion rate is the ratio of the sum of the coarse-ranked results for each recall channel to the sum of the target parameters for each recall channel. The volume guarantee quota is used to indicate the number of businesses that pass the coarse-ranking. In the coarse-ranking stage, if the volume guarantee quota is the first quantity, and a second quantity of businesses is obtained through screening, the first quantity of businesses should be retained from the second quantity of businesses.

[0055] When there are two recall channels, it is only necessary to determine whether the difference between the conversion rates of the two recall channels is greater than a preset threshold. When there are more than two recall channels, it is necessary to determine whether the difference between the conversion rates of each recall channel and the remaining recall channels is greater than a preset threshold. For example, when there are three recall channels, it is necessary to determine whether the difference between the conversion rates of the first and second recall channels is greater than a preset threshold, whether the difference between the conversion rates of the first and third recall channels is greater than a preset threshold, and whether the difference between the conversion rates of the second and third recall channels is greater than a preset threshold. Assuming that the difference between the conversion rates of the second and third recall channels is greater than a preset threshold, and the conversion rate of the second recall channel is greater than the overall conversion rate, and the conversion rate of the third recall channel is less than the overall conversion rate, then the guaranteed quota for the second recall channel is increased, and the guaranteed quota for the third recall channel is decreased.

[0056] In this embodiment, the business conversion rate of each recall channel can be determined by the ratio of the number of coarse-ranked services to the target parameters during the business recommendation process. The number of coarse-ranked services refers to the number of target businesses recalled by the recall channel and passing the coarse-ranking process. The target parameters are parameters determined based on the user's target actions on the target businesses that passed the fine-ranking process. Therefore, the business conversion rate of the recall channel can, to a certain extent, characterize the ranking and filtering effect of the coarse-ranking stage. When the difference between the business conversion rates of two recall channels exceeds a preset threshold, it indicates that some recall channels (i.e., the first type of recall channels) have a better ranking and filtering effect in the coarse-ranking stage, while some recall channels (i.e., the second type of recall channels) have a poorer ranking and filtering effect. In this case, by increasing the guaranteed quota of the first type of recall channels, the number of businesses meeting user needs that pass the coarse-ranking stage can be increased; by decreasing the guaranteed quota of the second type of recall channels, the number of businesses that do not meet user needs that pass the coarse-ranking stage can be reduced, allowing more businesses that meet user needs to pass through the coarse-ranking stage, thereby improving the ranking and filtering effect of the recommendation system in the coarse-ranking stage. This application embodiment dynamically adjusts the guaranteed quota of each recall channel based on the current business conversion rate of the recall channel. This not only ensures and improves the sorting and filtering effect of the recommendation system in the coarse ranking stage, but also improves the final recommendation effect of the recommendation system, recommending services that are more likely to meet the user's needs to the user.

[0057] Optionally, after increasing the guaranteed volume quota for the first type of recall channel and decreasing the guaranteed volume quota for the second type of recall channel, the method further includes:

[0058] At preset intervals, determine the business conversion rate of each recall channel under the current guaranteed quota.

[0059] It should be noted that the preset duration can be any duration set by the user, such as 3 hours, but is not limited to this. Understandably, when determining the conversion rate of the recall channel under the current guaranteed quota, the number of coarse-ranked results and target parameters of the recall channel during the recommendation process under the current guaranteed quota are used for determination, which will not be elaborated upon here.

[0060] If the difference between the business conversion rates of two recall channels under the current guaranteed quota is greater than a preset threshold, the guaranteed quota of the first type of recall channel will be increased and the guaranteed quota of the second type of recall channel will be decreased, until the difference between the business conversion rates of any two recall channels under the current guaranteed quota is less than or equal to the preset threshold.

[0061] It should be noted that when the difference in business conversion rates between two recall channels exceeds a preset threshold, it indicates that some recall channels (type 1) perform well in the coarse-sorting stage, while others (type 2) perform poorly. In other words, the ranking and filtering effects of each recall channel are unbalanced. In this case, by dynamically adjusting the guaranteed quotas for different types of recall channels, the ranking and filtering effects of each channel can be balanced, meaning their performance is not significantly different, thereby maximizing the ranking and filtering effect in the coarse-sorting stage.

[0062] In this embodiment, by adjusting the guaranteed quota of each recall channel multiple times, the guaranteed quota of each recall channel is balanced, thereby maximizing the sorting and filtering effect in the coarse sorting stage.

[0063] Optionally, before obtaining the coarse-ranked number of target services for each recall channel and the target parameters for each recall channel during the business recommendation process, the method further includes:

[0064] Set the guaranteed quota for each recall channel to a target value so that there is no limit to the number of coarse rankings in the business recommendation process.

[0065] It should be noted that the target value is the initial value before adjusting the guaranteed quota for the recall channel. When the initial guaranteed quota for a particular recall channel is the target value, there will be no limit on the number of services that pass the coarse ranking for that recall channel; that is, there is no limit to the number of services that pass the coarse ranking during the service recommendation process. For example, the target value can be 0, and the initial guaranteed quota can be 0, meaning that no more than a certain number of services are guaranteed to pass the coarse ranking for each recall channel. Here, the sum of the guaranteed quotas for all recall channels should be less than or equal to the total number of services that pass the coarse ranking in the recommendation system (the sum of the number of services that pass the coarse ranking for all recall channels).

[0066] Understandably, it's also possible to not set a guaranteed quota for each recall channel at the initial stage of the recommendation system, thus achieving an unlimited number of services in the coarse ranking process. Here, without setting a guaranteed quota for each recall channel at the initial stage, it's not guaranteed that any number of services will pass the coarse ranking for each recall channel. In other words, services naturally recalled by each recall channel are scored during the coarse ranking stage, and services with scores exceeding a threshold will proceed to the fine ranking stage.

[0067] In this embodiment of the application, before adjusting the guaranteed quota of the recall channel, the number of services that pass the coarse ranking in the service recommendation process is not limited. Without manually setting the guaranteed quota, the scoring results of the coarse ranking stage can be used to avoid the poor sorting and filtering effect caused by the guaranteed quota.

[0068] Optionally, the sum of the increases in the guaranteed quota for each recall channel of the first type is equal to the sum of the decreases in the guaranteed quota for each recall channel of the second type.

[0069] It should be noted that if the sum of the increases in the guaranteed quota for all recall channels in the first type equals the sum of the decreases in the guaranteed quota for all recall channels in the second type, the number of businesses that enter the fine-tuning stage after the guaranteed quota adjustment will not change. Therefore, it will not affect the fine-tuning stage. For example, before the guaranteed quota adjustment, the sum of the guaranteed quotas for all recall channels in the first type is 100, and the sum of the guaranteed quotas for all recall channels in the second type is 100; after the guaranteed quota adjustment, if the sum of the guaranteed quotas for all recall channels in the first type is 150, then the sum of the guaranteed quotas for all recall channels in the second type is 50.

[0070] In this embodiment of the application, the number of target services input to the fine-ranking stage will not change before and after the adjustment of the quota for the recall channel, thus eliminating the need to adjust the fine-ranking stage and reducing workload.

[0071] Optionally, for the target recall channel, the increase or decrease in the guaranteed quota of the target recall channel is equal to the product of the coarse ranking quantity of the target recall channel multiplied by the preset step size value and the target absolute value, where the target absolute value is the absolute value of the difference between the business conversion rate of the target recall channel and the overall conversion rate, and the target recall channel is any recall channel.

[0072] It should be noted that the preset step size can be understood as a hyperparameter, which can be set to a fixed value according to the parameter magnitude in the actual scenario. Specifically, the increase or decrease in the target recall channel's guaranteed quota can be calculated using Formula 1, where Formula 1 is:

[0073]

[0074] in, This indicates the increase or decrease in the guaranteed quota for the target recall channel; λ represents the preset step size; T i Indicates the target parameters of the target recall channel; α i This represents the number of coarse-ranked entries in the target recall channel; T represents the sum of the target parameters for each recall channel; Λ represents the sum of the number of coarse-ranked entries in each recall channel.

[0075] In this embodiment, the adjustment value of the guaranteed quota of the recall channel is proportional to the difference between the business conversion rate and the overall conversion rate, which can accelerate the iteration efficiency.

[0076] Optionally, target parameters include: number of seedings, number of purchases, or content consumption time.

[0077] It should be noted that "grass planting" refers to the act of sharing and recommending the excellent qualities of a product to stimulate others' desire to buy it, or the process of developing a desire to experience or own something based on external information; it also refers to the act of sharing and recommending something to another person, making that person like it. "Grass planting volume" is a statistical measure of the behavior indicated by "grass planting." "Purchase volume" refers to the number of products purchased by users from the target business, and "content consumption time" refers to the time people spend consuming internet information content, which is not limited to media types such as articles, pictures, and videos. For example, content consumption time can be the viewing time of video files shown in the target business, the reading time of business content from the target business, or the dwell time on web pages displaying the target business. It can be understood that "grass planting volume," "purchase volume," and "content consumption time" are target parameters in different scenarios.

[0078] In this embodiment of the application, the target parameter can be determined as the amount of seeding, the amount of purchase, or the duration of content consumption based on different application scenarios, so that the quota can be adjusted in different application scenarios.

[0079] Optionally, during the business recommendation process, the number of coarse-ranked results for each recall channel and the target parameters for each recall channel are obtained, including:

[0080] Obtain recall and exposure logs during the business recommendation process;

[0081] Based on the recall logs and exposure logs, the number of coarse-ranking tests for each recall channel and the target parameters for each recall channel are determined.

[0082] In this embodiment of the application, based on the recall log and the exposure log, the number of coarse-ranked results for each recall channel and the target parameters for each recall channel can be determined quickly and accurately.

[0083] like Figure 2 The diagram shown illustrates a practical application of the method for configuring quantity-preserving quotas in coarse sorting provided in this application, including:

[0084] Step 201: Calculate the conversion rate of different recall channels. Ideally, initially, no guaranteed quota is set for each recall channel; that is, both natural recall and non-guaranteed coarse ranking are performed. Based on the recall logs and exposure logs within a set time window (e.g., 3 hours), calculate the conversion rate of each recall channel. Specifically, Formula 2 can be used to determine the conversion rate of each recall channel. Formula 2:

[0085]

[0086] Where i is a positive integer, τ i α represents the business conversion rate of recall channel i; iT represents the number of coarse sorts for recall channel i. i Let α represent the target parameter of recall channel i. i and T i All of these can be obtained directly from the recall logs and exposure logs. The definition of the target parameter depends on the specific business scenario and requirements. Taking the review information flow recommendation as an example, the target parameter is the number of product seedings (such as clicks on interest points, collections, etc.); in other recommendation scenarios, it can correspond to factors such as user purchase volume, content consumption time, etc.

[0087] Step 202: Set the guaranteed quota for each recall channel. Here, the number of services passing the coarse-sorting stage is generally a fixed value, denoted by Λ. Assuming there are n recall channels, then the following condition is met:

[0088] Where, α i This indicates the number of coarse rows in recall channel i.

[0089] And let T represent the sum of the target parameters of each recall channel, that is:

[0090] Among them, T i This represents the target parameter of recall channel i.

[0091] The guaranteed volume quota for each recall channel will then be adjusted to α′. i :

[0092]

[0093] Where λ represents the preset step size value; T i Indicates the target parameters of the target recall channel; α i Λ represents the number of coarse-ranked results for the target recall channel; T represents the sum of the target parameters for each recall channel; Λ represents the sum of the number of coarse-ranked results for each recall channel.

[0094] Understandably, the above formula derivation can determine the sum of the coarse-ranked numbers for each recall channel after adjusting the guaranteed quota. Therefore, the number of businesses that enter the fine-ranking stage from the coarse-ranking stage will not change before and after the quota adjustment.

[0095] Step 203: After setting the guaranteed quota for each recall channel, calculate the business conversion rate of different recall channels again. See step 201 for details.

[0096] Step 204: Determine if the recall channels have reached a balance. If yes, proceed to step 205; otherwise, proceed to step 202. The standard for balance among the recall channels is that the difference in conversion rates between any two recall channels does not exceed a preset threshold. That is:

[0097] |τi -τ j |≤∈,1≤i ≤n,1≤j≤n;

[0098] Where, τ i τ represents the business conversion rate of recall channel i. j Let represent the conversion rate of recall channel j, and ∈ represent a preset threshold. The conversion rates of all recall channels are compared pairwise to determine if there are two recall channels that do not meet the above conditions. If so, the channel is not considered balanced; otherwise, it is considered balanced.

[0099] Step 205: Wait for subsequent triggers, that is, temporarily stop adjusting the guaranteed quota until the multi-channel recall changes or the timer ends, then execute step 203 again.

[0100] In this embodiment, the quota of each recall channel is adjusted dynamically based on the current business conversion rate of the recall channel. This not only ensures the improvement of the ranking and filtering effect of the recommendation system in the coarse ranking stage, but also improves the final recommendation effect of the recommendation system, recommending services that are more likely to meet the user's needs to the user.

[0101] See Figure 3 This application also provides a configuration device for quantity-preserving quotas in coarse sorting, the device comprising:

[0102] The acquisition module 31 is used to acquire the number of coarse-ranked target services of each recall channel and the target parameters of each recall channel during the business recommendation process. The number of coarse-ranked target services is the number of target services that pass the coarse-ranking among the services recalled by the recall channel, and the target parameters are the parameters determined based on the user's target operation on the target services that pass the fine-ranking.

[0103] The first determining module 32 is used to determine the business conversion rate of each recall channel by the ratio of the number of coarse-sorted results of the recall channel to the target parameter of the recall channel.

[0104] The first configuration module 33 is used to increase the guaranteed quota of the first type of recall channel and decrease the guaranteed quota of the second type of recall channel when the difference between the business conversion rates of two recall channels in multiple recall channels is greater than a preset threshold. The first type of recall channel is the recall channel with a business conversion rate greater than the overall conversion rate, and the second type of recall channel is the recall channel with a business conversion rate less than the overall conversion rate. The overall conversion rate is the ratio of the sum of the coarse rankings of each recall channel to the sum of the target parameters of each recall channel. The guaranteed quota is used to indicate the number of businesses that pass the coarse ranking.

[0105] Optionally, the device further includes:

[0106] The second determination module is used to determine the business conversion rate of each recall channel under the current guaranteed quota every preset time interval;

[0107] The second configuration module is used to increase the guaranteed quota of the first type of recall channel and decrease the guaranteed quota of the second type of recall channel when the difference between the business conversion rates of two recall channels under the current guaranteed quota is greater than a preset threshold, until the difference between the business conversion rates of any two recall channels under the current guaranteed quota is less than or equal to the preset threshold.

[0108] Optionally, the device further includes:

[0109] The third configuration module is used to set the guaranteed quota for each recall channel to the target value so that there is no limit to the number of coarse rankings in the business recommendation process.

[0110] Optionally, the sum of the increases in the guaranteed quota for each recall channel of the first type is equal to the sum of the decreases in the guaranteed quota for each recall channel of the second type.

[0111] Optionally, for the target recall channel, the increase or decrease in the guaranteed quota of the target recall channel is equal to the product of the coarse ranking quantity of the target recall channel multiplied by the preset step size value and the target absolute value, where the target absolute value is the absolute value of the difference between the business conversion rate of the target recall channel and the overall conversion rate, and the target recall channel is any recall channel.

[0112] Optionally, target parameters include: number of seedings, number of purchases, or content consumption time.

[0113] Optionally, module 31 includes:

[0114] The acquisition unit is used to acquire recall logs and exposure logs during the business recommendation process;

[0115] The determination unit is used to determine the number of coarse-ranking criteria for each recall channel and the target parameters for each recall channel based on the recall log and the exposure log.

[0116] The configuration device for maintaining quantity quotas in the coarse-grained sorting provided in this application embodiment can achieve... Figures 1 to 2 The various processes implemented in the coarse sorting quota configuration method in the method embodiment are not described in detail here to avoid repetition.

[0117] In the embodiments of this application, the business conversion rate of each recall channel can be determined by the ratio of the number of coarse-ranked services to the target parameters during the business recommendation process. The number of coarse-ranked services refers to the number of target businesses recalled by the recall channel and passing the coarse-ranking stage. The target parameters are parameters determined based on the user's target operations on the target businesses that have passed the fine-ranking stage. Therefore, the business conversion rate of the recall channel can, to a certain extent, characterize the ranking and filtering effect of the coarse-ranking stage. When the difference between the business conversion rates of two recall channels is greater than a preset threshold, it indicates that some recall channels (i.e., the first type of recall channels) have a better ranking and filtering effect in the coarse-ranking stage, while some recall channels (i.e., the second type of recall channels) have a poorer ranking and filtering effect. In this case, by increasing the guaranteed quota of the first type of recall channel, the number of businesses that meet user needs passing the coarse-ranking stage can be increased; by decreasing the guaranteed quota of the second type of recall channel, the number of businesses that do not meet user needs passing the coarse-ranking stage can be reduced, allowing more businesses that meet user needs to pass through the coarse-ranking stage, thereby improving the ranking and filtering effect of the recommendation system in the coarse-ranking stage. This application embodiment dynamically adjusts the guaranteed quota of each recall channel based on the current business conversion rate of the recall channel. This not only ensures and improves the sorting and filtering effect of the recommendation system in the coarse ranking stage, but also improves the final recommendation effect of the recommendation system, recommending services that are more likely to meet the user's needs to the user.

[0118] On the other hand, embodiments of this application also provide an electronic device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the configuration method for maintaining quotas in coarse sorting provided in the above-described embodiments.

[0119] Furthermore, embodiments of this application also provide a readable storage medium, which, when the instructions in the readable storage medium are executed by the processor of an electronic device, enables the electronic device to execute the coarse-sorting quota configuration method provided in the above embodiments.

[0120] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0121] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0122] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0123] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for allocating quantity-preserving quotas in coarse sorting, characterized in that, The method includes: In the process of obtaining business recommendations, the number of coarse rankings for each recall channel and the target parameters for each recall channel are as follows: the number of coarse rankings is the number of target businesses that pass the coarse ranking among the businesses recalled by the recall channel, and the target parameters are parameters determined based on the user's target operation on the target businesses that pass the fine ranking. For each of the aforementioned recall channels, the ratio of the number of coarse-ranked results for the recall channel to the target parameter of the recall channel is determined as the business conversion rate of the recall channel. If the difference between the business conversion rates of two recall channels in a plurality of recall channels exceeds a preset threshold, the guaranteed quota of the first type of recall channel is increased and the guaranteed quota of the second type of recall channel is decreased. The first type of recall channel is a recall channel with a business conversion rate greater than the overall conversion rate, and the second type of recall channel is a recall channel with a business conversion rate less than the overall conversion rate. The overall conversion rate is the ratio of the sum of the coarse-ranked numbers of each recall channel to the sum of the target parameters of each recall channel. The guaranteed quota is used to indicate the number of businesses that pass the coarse-ranking.

2. The method according to claim 1, characterized in that, After increasing the guaranteed quantity quota for the first type of recall channel and decreasing the guaranteed quantity quota for the second type of recall channel, the method further includes: At preset intervals, the business conversion rate of each recall channel under the current guaranteed quota is determined once; If, in any instance, the difference between the business conversion rates of two of the recall channels under the current guaranteed quota is greater than a preset threshold, the guaranteed quota of the first type of recall channel is increased, and the guaranteed quota of the second type of recall channel is decreased, until the difference between the business conversion rates of any two of the recall channels under the current guaranteed quota is less than or equal to the preset threshold.

3. The method according to claim 1, characterized in that, Before determining the number of coarse-ranked results for each recall channel and the target parameters for each recall channel during the business recommendation process, the method further includes: Set the guaranteed quota for each recall channel to a target value so that there is no limit to the number of coarse rankings in the business recommendation process.

4. The method according to claim 1, characterized in that, The sum of the increases in the guaranteed quota for each recall channel of the first type is equal to the sum of the decreases in the guaranteed quota for each recall channel of the second type.

5. The method according to claim 4, characterized in that, For a target recall channel, the increase or decrease in the guaranteed quota of the target recall channel is equal to the product of the number of coarse rankings of the target recall channel multiplied by a preset step size and a target absolute value, wherein the target absolute value is the absolute value of the difference between the business conversion rate of the target recall channel and the overall conversion rate, and the target recall channel is any of the recall channels.

6. The method according to claim 1, characterized in that, The target parameters include: number of product seedings, number of purchases, or content consumption time.

7. The method according to claim 1, characterized in that, In the process of obtaining service recommendations, the number of coarse-ranked results for each recall channel and the target parameters for each recall channel include: Obtain recall and exposure logs during the business recommendation process; Based on the recall log and exposure log, the number of coarse-ranking results for each recall channel and the target parameters for each recall channel are determined.

8. A configuration device for maintaining quantity quotas in coarse sorting, characterized in that, The device includes: The acquisition module is used to acquire the number of coarse rankings for each recall channel and the target parameters for each recall channel during the business recommendation process. The number of coarse rankings is the number of target businesses that have passed the coarse ranking among the businesses recalled by the recall channel, and the target parameters are parameters determined based on the user's target operation on the target businesses that have passed the fine ranking. The first determining module is used to determine the business conversion rate of each recall channel by the ratio of the number of coarse-sorted results of the recall channel to the target parameter of the recall channel. The first configuration module is used to increase the guaranteed quota of a first type of recall channel and decrease the guaranteed quota of a second type of recall channel when the difference between the business conversion rates of two recall channels in the plurality of recall channels is greater than a preset threshold. The first type of recall channel is a recall channel with a business conversion rate greater than the overall conversion rate, and the second type of recall channel is a recall channel with a business conversion rate less than the overall conversion rate. The overall conversion rate is the ratio of the sum of the coarse-ranked numbers of each recall channel to the sum of the target parameters of each recall channel. The guaranteed quota is used to indicate the number of services that pass the coarse-ranking.

9. An electronic device, characterized in that, include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the processor, when executing the program, implements the method for configuring quantity-preserving quotas in coarse sorting as described in any one of claims 1-7.

Images