Resource recommendation method, device, electronic device and storage medium

Abstract

The present disclosure provides a resource recommendation method, device, electronic device and storage medium, which relate to the field of computer technology, and specifically to the technical fields of big data, information flow, intelligent recommendation, etc. The specific implementation scheme is: obtaining the resources to be adjusted and the resource set corresponding to the resources to be adjusted; determining the first proportion of the resources of the target category in the resource set; based on the first proportion and the expected proportion, determining the adjustment factor of the resources to be adjusted by the PID method, so that the first proportion approaches the expected proportion; correcting the recommended score of the resources to be adjusted based on the adjustment factor, and obtaining the corrected score of the resources to be adjusted; and recommending resources to the target object based on the corrected score of the resources to be adjusted. In the embodiment of the present disclosure, the PID method is used to correct the recommended score of the resources to be adjusted, which is conducive to the reasonable reordering of resources to meet the user's demand for diversified resources, thereby facilitating the reasonable use of processing resources to improve user retention rate.

Description

Technical Field

[0001] The present disclosure relates to the field of computer technology, and in particular to technical fields such as big data, information flow, and intelligent recommendation. Background Art

[0002] For information flow recommendations, the diversity of target objects is particularly important. Among a large number of resources, it is necessary to filter out the resources that need to be recommended to users based on recommendation strategies. Although some recommendation strategies are given in related technologies, how to recommend diverse content to users is still an issue that needs attention. Summary of the invention

[0003] The present disclosure provides a resource recommendation method, device, electronic device and storage medium.

[0004] According to one aspect of the present disclosure, a resource recommendation method is provided, comprising:

[0005] Obtain the resource to be adjusted and the resource set corresponding to the resource to be adjusted; the resource to be adjusted is any resource in the sequence of resources to be recommended, and the resource set includes the resource to be adjusted and the resources before the resource to be adjusted in the sequence of resources to be recommended;

[0006] Determine the first proportion of resources of the target category in the resource set;

[0007] Based on the first proportion and the expected proportion, the adjustment factor of the resource to be adjusted is determined by the PID (Proportional, Integral, Differential) method, so that the first proportion approaches the expected proportion;

[0008] Modifying the recommended score of the resource to be adjusted based on the adjustment factor to obtain a modified score of the resource to be adjusted;

[0009] Based on the revised scores of the resources to be adjusted, resources are recommended to the target object.

[0010] According to another aspect of the present disclosure, there is provided a resource recommendation device, comprising:

[0011] A first acquisition module is used to acquire a resource to be adjusted and a resource set corresponding to the resource to be adjusted; the resource to be adjusted is any resource in the sequence of resources to be recommended, and the resource set includes the resource to be adjusted and the resources before the resource to be adjusted in the sequence of resources to be recommended;

[0012] A first proportion determination module, used to determine a first proportion of resources of a target category in a resource set;

[0013] An adjustment factor determination module, configured to determine an adjustment factor of the resource to be adjusted by a PID method based on the first proportion and the expected proportion, so that the first proportion approaches the expected proportion;

[0014] A first correction module, configured to correct the recommended score of the resource to be adjusted based on the adjustment factor to obtain a corrected score of the resource to be adjusted;

[0015] The recommendation module is used to recommend resources to the target object based on the corrected scores of the resources to be adjusted.

[0016] According to another aspect of the present disclosure, there is provided an electronic device, comprising:

[0017] at least one processor; and

[0018] a memory communicatively connected to the at least one processor; wherein,

[0019] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute the method of any embodiment of the present disclosure.

[0020] According to another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable the computer to execute the method according to any embodiment of the present disclosure.

[0021] According to another aspect of the present disclosure, a computer program product is provided, including a computer program. When the computer program is executed by a processor, the method according to any one of the embodiments of the present disclosure is implemented.

[0022] In the disclosed embodiment, the PID method is used to adjust the recommendation ranking of the resources to be adjusted, which is conducive to reasonably enriching the content of the resources to be recommended, making the resources to be recommended more diverse, thereby facilitating more effective resource recommendations and increasing user retention rates.

[0023] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0025] Figure 1 is a flowchart of a resource recommendation method according to another embodiment of the present disclosure;

[0026] Figure 2is a schematic diagram of a mapping method of impact factors according to another embodiment of the present disclosure;

[0027] Figure 3a is a process schematic diagram of a resource recommendation method according to another embodiment of the present disclosure;

[0028] Figure 3b is a process schematic diagram of another resource recommendation method according to another embodiment of the present disclosure;

[0029] Figure 3c is a flowchart of a resource recommendation method according to another embodiment of the present disclosure;

[0030] Figure 4 is a schematic diagram of the effect of a resource recommendation method according to another embodiment of the present disclosure;

[0031] Figure 5 is a structural diagram of a resource recommendation device according to another embodiment of the present disclosure;

[0032] Figure 6 It is a block diagram of an electronic device used to implement the resource recommendation method of the embodiment of the present disclosure. DETAILED DESCRIPTION

[0033] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope of the present disclosure. Similarly, for the sake of clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0034] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features. In the description of the present disclosure, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0035] In the technical solution disclosed herein, the acquisition, storage and application of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0036] With the development of computer technology, the amount of resources has increased. Especially with the emergence of short video platforms, every user can create resources and publish them on the Internet. Faced with a huge amount of resources, it is necessary to classify resources when performing resource recommendations. The division of resources can include multiple levels, such as major categories and minor categories, or general categories and detailed categories, depending on the granularity of the division.

[0037] Resource recommendation strategies in related technologies generally include resource recommendation based on information entropy, MMR (Maximal Marginal Relevance), LCN (Longest Continuous no-click Num), and other diverse strategies. These resource recommendation strategies are based on fine-grained categories, and sometimes the recommended content is too single.

[0038] In view of this, if we can reasonably enrich the recommended resources by improving the recommendation strategy, it will be helpful to meet the users' demand for rich resources.

[0039] Based on this, the present disclosure provides a resource recommendation method in which the PID method can be used to enrich the resources recommended to the user. Figure 1 FIG. 1 is a flow chart of a resource recommendation method in an embodiment of the present disclosure, including:

[0040] S101, obtaining a resource to be adjusted and a resource set corresponding to the resource to be adjusted; the resource to be adjusted is any resource in a sequence of resources to be recommended, and the resource set includes the resource to be adjusted and resources before the resource to be adjusted in the sequence of resources to be recommended.

[0041] Among them, the resource to be adjusted already has a recommendation score, and the recommendation score is used to determine the recommendation ranking of the resource to be recommended in the sequence of resources to be recommended. The higher the recommendation ranking, the higher the priority of recommendation to the target object. The recommendation score can be obtained based on any resource scoring method in the field of resource recommendation technology. For example, the recommendation score is determined based on resource scoring methods such as ICF (item-based collaborative filtering) or UCF (user-based collaborative filtering), thereby determining the recommendation ranking. Therefore, during implementation, the sequence of resources to be recommended can be a resource set screened based on the resource scoring method, which will be referred to as the candidate resource set hereinafter.

[0042] In some other possible implementations, when N resources need to be screened out from the candidate resource set and sent to the target object, if the N resources are selected one by one, the resource sequence to be recommended may be a resource sequence constructed by the selected resources and the resources to be adjusted.

[0043] S102: Determine a first proportion of resources of the target category in the resource set.

[0044] The resource set may include resources of various categories, such as news resources, entertainment resources, etc. The target category is any one of the various resource categories of the resource set.

[0045] S103: Based on the first proportion and the expected proportion, determine the adjustment factor of the resource to be adjusted by using the PID method, so that the first proportion approaches the expected proportion.

[0046] Among them, the role of the PID method is to use the expected proportion as a reference value, calculate the difference between the first proportion of the target category of resources in the resource set and the reference value, so as to adjust the first proportion to quickly approach the reference value. Ideally, the closer the adjusted first proportion is to the expected proportion, the better, and it is best to be able to stably maintain it near the expected proportion. However, in the embodiment of the present disclosure, the PID method is used as a way to adjust the sorting, as long as the goal of the adjustment is to make the first proportion approach the expected proportion, and the degree of closeness to the expected proportion is not strictly required.

[0047] The PID method includes proportional term, integral term and differential term. The proportional term is used as the basis, and the integral term and differential term are used as auxiliary terms. The difference between the first proportion and the expected proportion can be described from different levels and dimensions, thereby obtaining a reasonable adjustment factor.

[0048] During implementation, the expression of the PID method can be designed according to actual needs, as long as the goal of making the first proportion approach the expected proportion can be met. The embodiments of the present disclosure do not limit the specific expression.

[0049] S104: Modify the recommended score of the resource to be adjusted based on the adjustment factor to obtain a modified score of the resource to be adjusted.

[0050] S105: Recommend resources to the target object based on the corrected scores of the resources to be adjusted.

[0051] In the disclosed embodiment, the PID method is used to adjust the recommended order of the resources to be adjusted in the sequence of resources to be recommended, which is conducive to reasonably enriching the content of the resources to be recommended and making the resources to be recommended more diversified. This is conducive to achieving more effective resource recommendation and improving the retention rate of the target object. Since the PID method itself relies on the calculation of proportional terms, integral terms and differential terms, the PID method is used for resource recommendation to obtain the adjustment factor of the resources to be adjusted. The calculation items based on multiple different dimensions are conducive to making the adjustment factor of the resources to be adjusted more accurate, and then the resources can be reasonably reordered to meet the user's demand for diversified resources, and the utilization rate of processing resources can also be improved.

[0052] In some embodiments, as described above, the PID method for determining the adjustment factor of the resource to be adjusted includes a proportional term, an integral term, and a differential term.

[0053] Among them, the differential term is used to represent the change between the first gap and the second gap. The first gap is the gap between the first proportion and the expected proportion. The second gap is the gap between the second proportion of the target category of resources in the resource set corresponding to the reference resources and the expected proportion when the adjacent resources of the resources to be adjusted in the recommended resource sequence are used as the resources to be adjusted.

[0054] It should be noted that the resource to be adjusted is any resource, and therefore the PID method is used to adjust the recommended score for each resource to obtain a corrected score corresponding to each resource.

[0055] In the disclosed embodiment, it is assumed that the candidate resource set includes m resources, and N resources need to be selected from the m candidate resources and sent to the target object. Then the resource sequence to be recommended can be a resource sequence constructed by the m candidate resources, or a resource sequence constructed by the selected resources and the resources to be adjusted from the N resources to be selected.

[0056] Taking the m candidate resources as an example, assuming that there are 15 resources in the resource sequence to be recommended (i.e., m=15), when the resource to be adjusted is the 10th resource, the first 10 resources in the resource sequence to be recommended are used as the resource set corresponding to the 10th resource to be adjusted, and the proportion of the target category resources in the resource set is determined as the first proportion. The difference between the first proportion and the expected proportion is determined, thereby obtaining the first gap. The resource to be adjusted, which is an adjacent resource to the resource to be adjusted, may be the 9th resource in the resource sequence to be recommended. Then the first 9 resources in the resource sequence to be recommended are used as the resource set corresponding to the 9th resource, and the proportion of the target category resources in the resource set is determined as the second proportion. The difference between the second proportion and the expected proportion is determined, thereby obtaining the second gap. Among them, the change of the first gap and the second gap can be understood as a vector, which represents the change rate and change direction of the difference between the actual proportion and the expected proportion from the adjacent resource to the resource to be adjusted.

[0057] The integral term and the differential term are used together with the proportional term to determine the adjustment factor of the resource to be adjusted.

[0058] In the disclosed embodiment, the differential term is represented by the difference between the actual proportion (i.e., the first proportion and the second proportion) and the expected proportion corresponding to two adjacent resources. This representation method can accurately describe the change between the actual value and the reference value, and can determine the appropriate adjustment factor through the PID method principle to facilitate the reasonable adjustment of the sorting order. Finally, on this basis, the resources of the target category can be reasonably configured to increase the exposure of the target category resources, so as to guide users to pay attention to the resources of the target category and meet the user's needs for different resources.

[0059] In some embodiments, the expression of the PID method for determining the adjustment factor of the resource to be adjusted is shown in formula (1):

[0060]

[0061] In formula (1), U(K) represents the influence factor of the resource to be adjusted based on the PID method, that is, the result directly calculated based on the PID expression. In order to reasonably adjust the order of the resources to be adjusted, the embodiment of the present disclosure further adopts the influence factor to determine the adjustment factor, and the specific method will be described later.

[0062] In formula (1), K represents the resource to be adjusted, and represents the recommended ranking of the resource to be adjusted in the sequence of recommended resources. For example, in a sequence of recommended resources containing 15 resources, the resource to be adjusted is the 12th resource, then K is 12, and K also represents the resource to be adjusted with the recommended ranking at 12.

[0063] In formula (1), (K-1) represents the reference resource with the recommended ranking (K-1) in the sequence of recommended resources; e(K) represents the difference between the first proportion and the expected proportion, and e(K-1) represents the difference between the second proportion and the expected proportion. The second proportion is the proportion of the target category resources in the resource set corresponding to the reference resource; e(n) represents the difference between the third proportion and the expected proportion. The third proportion is the proportion of the target category resources in the resource set corresponding to the resource with the recommended ranking n; K p , K i , K d are weights, among which K p represents the weight of the proportional term, K i represents the weight of the integral term, K d Represents the weight of the derivative term.

[0064] In some embodiments, the weight K p , K i , K d It can be dynamically adjusted according to the specific needs of the actual business scenario. For business scenarios belonging to the same category, in order to save computing power, the weight K p , K i , K d It can also be set to a fixed value. p , K i , K d The method is not limited.

[0065] In the embodiment of the present disclosure, in the calculation formula of the PID method, the difference between the actual proportion and the expected proportion is directly used as the input, and the change in the gap between the two adjacent resources and the expected proportion is taken into account. Not only is the calculation method simple and easy, it can also support processing resources of various computing powers, but it can also describe the gap between the actual proportion and the expected proportion from multiple dimensions, thereby accurately determining the adjustment factor and obtaining a reasonable resource ranking.

[0066] In some embodiments, determining the weight of any one of the proportional term, the integral term, and the derivative term in the PID method may be implemented as follows:

[0067] A1, determine the difference between the first proportion and the expected proportion.

[0068] A2, based on the mapping relationship between the gap and the weight, determine the weight of any item.

[0069] Among them, the weight K p , K i , K d It changes with the change of e(K) in the expression of the PID method, and the two have a certain mapping relationship. The mapping relationship can be a linear relationship or a nonlinear relationship. When implemented, it can be determined according to the actual situation. Since e(K) represents the difference between the first proportion and the expected proportion, and e(K) is a value that may be positive or negative, when e(K) is negative, it is necessary to correct the negative trend of e(K), that is, to correct the trend of e(K) to a positive trend. Therefore, the mapping relationship can be set based on this principle.

[0070] In some embodiments, the purpose of determining the above weights includes two requirements. The first requirement is to ensure as much as possible that the impact factor calculated based on the determined weights can make the first proportion closer to the expected proportion. Moreover, when the first proportion does not exceed the expected proportion, it is hoped that the impact factor is positive; when the first proportion exceeds the expected proportion, it is hoped that the impact factor is negative, which is equivalent to downgrading. The second requirement is that when the gap between the first proportion and the expected proportion is large, it is necessary to adjust the above weights as much as possible so that the first proportion can quickly approach the expected proportion. According to the above two requirements, the weight K can be finally determined through experimental test data. p , K i , K d The value of .

[0071] In an embodiment of the present disclosure, a method is provided for dynamically determining the weight of any one of the proportional term, integral term and differential term in the PID method, which is beneficial for adjusting the weight so that the PID method can obtain a suitable adjustment factor to reasonably arrange the recommended ranking.

[0072] In some embodiments, determining the adjustment factor of the resource to be adjusted by the PID method may be implemented as follows:

[0073] B1, determine the influencing factor through the expression of PID method.

[0074] B2, based on the mapping rules, maps the impact factor to the target value range to obtain the adjustment factor.

[0075] The mapping rules meet the following conditions:

[0076] When the impact factor is a negative number, the impact factor is mapped to the first sub-interval of the target value interval, and the first sub-interval is used to reduce the recommended ranking of the resources to be adjusted;

[0077] When the impact factor is a positive number, the impact factor is mapped to a second sub-interval of the target value interval, and the second sub-interval is used to improve the recommended ranking of the resources to be adjusted.

[0078] Among them, no matter whether the impact factor is positive or negative, when mapping the impact factor to the target value range, these impact factors need to be converted into positive numbers for mapping. When the impact factor is a negative number, it means that the first proportion of the target category's resources in the resource set exceeds the expected proportion, that is, the target category has too many resources, and it is hoped to reduce the relevant resources under the target category; similarly, when the impact factor is a positive number, it means that the first proportion of the target category's resources in the resource set does not reach the expected proportion, that is, the target category's resources are insufficient, and it is hoped to increase the target category's resources.

[0079] During implementation, when the value of the impact factor is negative, the impact factor is mapped to the first sub-interval to reduce the recommended ranking of the resources to be adjusted under the target category corresponding to the impact factor, and the smaller the value of the impact factor, the smaller the value mapped to the first sub-interval. When the value of the impact factor is positive, the impact factor is mapped to the second sub-interval to improve the recommended ranking of the resources to be adjusted under the target category corresponding to the impact factor, and the larger the value of the impact factor, the larger the value mapped to the second sub-interval.

[0080] In the disclosed embodiment, according to the mapping rules, the impact factor is mapped to the target value range to obtain the adjustment factor, which is conducive to adjusting the recommended ranking of the resources to be adjusted based on the impact factor, making the recommended ranking of the resources to be adjusted more reasonable and able to more easily meet the needs of the target object.

[0081] In some embodiments, the mapping rule may be implemented using a neural network or an activation function.

[0082] In the embodiment of the present disclosure, in order to simplify the operation and improve the efficiency, the influence factor is mapped to the target value range based on the mapping rule, which can be implemented as follows: based on the sigmoid activation function that satisfies the mapping rule, the influence factor is mapped to the target value range.

[0083] Among them, due to the monotonically increasing property of the sigmoid activation function and the monotonically increasing property of the inverse function of the sigmoid activation function, the sigmoid activation function can usually map the variable to between (0,1). In the embodiment of the present disclosure, in order to reasonably adjust the recommended ranking, the value interval (0,1) is the first sub-interval, that is, the interval used to reduce the ranking. In the embodiment of the present disclosure, the sigmoid activation function also needs to map the impact factor to the second sub-interval, that is, the interval used to improve the ranking. In order to reasonably map the impact factor to these two sub-areas, the expression of the sigmoid activation function is shown in the following formula (2), where final_factor represents the value of the sigmoid activation function, and a, b, and c are constants.

[0084]

[0085] Based on expression (2), in the embodiment of the present disclosure, the impact factor can be mapped to the first sub-interval and the second sub-interval. The overall target value interval can be set to (0, 3) or (0, 5), etc. The target value interval can be determined according to actual business needs. Figure 2 As shown in the figure, when the influence factor calculated based on the PID expression is -A1, it is mapped to point B1 of the first subinterval (0,1). When the influence factor is -A2, since -A1 is less than -A2, -A2 is mapped to point B2 of the first subinterval (0,1), and B2 is greater than B1. Similarly, when the influence factor is a positive number, it is mapped to the second subinterval (1,3). Figure 2 As shown, when the impact factor calculated based on the PID expression is C1, it is mapped to the D1 point of the second subinterval of (1,3). When the impact factor is C2, since C1 is less than C2, C2 is mapped to the D2 point of the second subinterval of (1,3), and D2 is greater than D1.

[0086] In the disclosed embodiment, the influencing factor is mapped to the target value range based on the sigmoid activation function that satisfies the mapping rule, which is beneficial to utilize the monotonically increasing property of the sigmoid activation function and its inverse function to simply and efficiently complete the mapping from the influencing factor to the target value range, so as to ultimately obtain a reasonable adjustment factor.

[0087] In some embodiments, the expected share may be determined based on the following method:

[0088] C1, determine the target object's interest in the target category's resources within a specified period of time.

[0089] In some embodiments, the specified duration can be any time length, such as 1 hour, 12 hours, or 1 day. The specified duration can also be determined based on the refresh of the resource by the target object. For example, the time period between two adjacent refresh behaviors of the target object is used as the specified duration. In the embodiments of the present disclosure, there is no limitation on the acquisition of the specified duration.

[0090] During implementation, the interest level can be determined based on the target operation of the target object on the target category of resources. The target operation can include at least one of the behaviors indicating the target object's interest in the resource, such as click behavior, forwarding behavior, purchase behavior, and like behavior. During implementation, if the click behavior is emphasized, the target operation can only select the click behavior. In other embodiments, when the target operation includes multiple behaviors, different weights can be set for different behaviors to determine the interest level. For example, the number of each behavior in the target operation is weighted and summed to determine the interest level.

[0091] C2, adjusts the interest level based on the preset bias to obtain the expected share.

[0092] The expected proportion can be determined based on the interest level and the preset bias according to: expected proportion = interest level + preset bias. For example, if the interest level is 0.3 and the preset bias is 0.4, 0.3+0.4=0.7, then the expected proportion is 0.7.

[0093] In the disclosed embodiment, adding a preset bias to adjust the interest of the target object is helpful to avoid the situation where the target object generates too single or too few behaviors on the target category resources, and the resources recommended to the target object based on the determined expected proportion are also too single. This is helpful to enhance the diversity of the resources to be recommended.

[0094] In some embodiments, determining the interest of the target object in resources of the target category within a specified period of time may also be implemented as follows:

[0095] D1, obtain the recommended resource set for the target object within the specified time period.

[0096] D2, obtaining the target object's operation records on the resources of the target category in the recommended resource set.

[0097] The operation records may include operations such as clicking, liking, collecting, purchasing, and forwarding.

[0098] D3, based on the operation records, filters out resources with target operations from resources of the target category in the recommended resource set.

[0099] For example, the operation record contains three operation behaviors of the target object: click, favorite, and purchase. And each operation behavior corresponds to a corresponding resource, which may be repeated or different. However, if the target operation is set to click, only the resources corresponding to the click operation of the target object are filtered out.

[0100] D4, determine the proportion of resources with target operations in the recommended resources and obtain the interest level.

[0101] In the disclosed embodiment, by screening resources with target operations, the target object's interest in the recommended resources with target operations is determined, which is beneficial for adjusting the recommended resources based on the target object's interest so that the resources to be adjusted can more easily meet the target object's interest needs.

[0102] In some embodiments, if the expected proportion is too large, it may result in a relatively single recommended resource. Therefore, in the embodiment of the present disclosure, when the expected proportion is greater than the expected upper limit, the expected proportion needs to be corrected to be less than or equal to the expected upper limit.

[0103] Among them, the expected upper limit can be dynamically adjusted according to the specific business scenario.

[0104] In the disclosed embodiment, by setting an expected upper limit, it is helpful to avoid the situation where the expected proportion is too large, resulting in too single resources recommended to the target object, due to the expected proportion being greater than the expected upper limit. It is also helpful to timely adjust the situation where the expected proportion is too large to effectively reduce the expected proportion.

[0105] In some embodiments, when the PID method adjusts the ranking, there may be a uniform directional deviation. For example, the actual proportion should be increased to approach the expected proportion, but the PID requires the actual proportion to be lowered. When a directional deviation occurs, if the integral term accumulates infinitely, it may be impossible to quickly exit from the saturation zone to return to the normal adjustment direction, thereby causing the continuous recommendation of undesirable resources, which will affect the accuracy of the recommendation ranking. In view of this, in the embodiment of the present disclosure, when the value of the integral term in the PID method is greater than the integral upper limit, the value of the integral term needs to be corrected to be less than or equal to the integral upper limit.

[0106] Among them, the points cap can be dynamically adjusted according to specific business scenarios.

[0107] In the disclosed embodiment, since the integral item needs to be determined based on the expected proportion, the integral upper limit of the integral item is set on the basis of setting the expected upper limit of the expected proportion. This is beneficial to avoid the infinite accumulation of integral items in the PID method, which causes the resources recommended to the target object to be too single. It is also beneficial to timely adjust the situation where the accumulated value of the integral item is too large to reduce the value of the integral item.

[0108] In some embodiments, the target category may be determined based on the following steps, including:

[0109] E1, obtain the recommendation records for the target object within a preset time period.

[0110] E2, filter out the categories whose resource quantity is lower than the preset threshold from the recommended records as the target categories.

[0111] For example, if the preset time period is 15 minutes and the preset threshold of the number of information resources is 6, then the recommended records for the target object within 15 minutes are obtained, and the recommended records show that the number of information resources is 4, which is lower than the number of entertainment resources (i.e., the preset threshold), then the information category is used as the target category. In other words, the category with a smaller number of display can be selected from the recommended records at one time as the target category. Of course, the preset threshold can also be determined based on an empirical value, and the embodiments of the present disclosure are not limited to this.

[0112] In the disclosed embodiment, the target categories are screened out based on whether the number of resources is lower than a preset threshold, which is conducive to enriching the recommended resources, meeting the user's needs for different resources, and guiding the user to pay attention to more resources.

[0113] In some embodiments, the target category can be a sub-category. However, the sub-category may result in a single recommended content. Therefore, the embodiments of the present disclosure may consider adjusting the recommended resources from a macro perspective. Thus, the target category is a category in a category set, and the category set includes at least two categories, and each category includes multiple sub-categories. For example, a target category that is a primary category, that is, a category that includes multiple subordinate sub-categories, can be used as a target category.

[0114] For example, the category set includes two categories: pan-entertainment and pan-information. Among them, the pan-entertainment category can include subcategories such as travel, film and television, and games; the pan-information category includes subcategories such as domestic news and foreign news. Taking the pan-entertainment category as the target category, different entertainment resources can be provided to users, which can avoid the recommended entertainment content being too single. Taking the pan-information category as the target category, different information content can be provided to users, such as finance, law, news, etc., which can also avoid the recommended information content being too single.

[0115] In the disclosed embodiment, the target category is a large category containing rich subcategories, which is conducive to ensuring that the resources to be recommended are more diversified and is conducive to improving the diversity of the resources to be recommended.

[0116] In some embodiments, N resources are selected from the m resources of the candidate resource set. It can be understood that when a resource subset (i.e., N resources) is selected from the candidate resource set and recommended to the target object, since the resources in the resource subset are selected one by one, the sequence of resources to be recommended includes the existing resources in the resource subset and the resources to be adjusted, and the existing resources are sorted in the order of being added to the resource subset, and the resources to be adjusted are sorted at the end. In other embodiments, the sequence of resources to be recommended includes the resource subset, and the difference set between the candidate resource set and the resource subset, and the order of the resources in the resource subset in the sequence of resources to be recommended is before the difference set, and the resources in the difference set are sorted according to the candidate resource set. For example, the first resource is selected from the m resources of the candidate resource set, and the first resource is set in the first place, and the remaining m-1 resources are sorted according to the order in the candidate resource set. Of course, when the resource subset is empty, that is, when the first resource is not selected, the m resources are sorted completely according to the candidate resource set.

[0117] Accordingly, based on the corrected scores of the resources to be adjusted, recommending resources to the target object can be implemented as follows:

[0118] When the existing resources in the resource subset do not reach the specified number and the corrected score of the resource to be adjusted is the maximum value of the corrected scores of all resources in the remaining resource set, the resource to be adjusted is added to the resource subset, where the remaining resource set is the resources in the sequence of resources to be recommended that have not been added to the resource subset.

[0119] The disclosed embodiment can construct a sequence of resources to be recommended based on a candidate resource set or a resource subset, and filter out resources that need to be recommended to the user based on the recommended resource sequence. By filtering, resources of the target category can be effectively filtered out from multiple candidate resources, so as to enrich the resources recommended to the target object. Based on this, processing resources can be effectively utilized for recommendation to improve the user retention rate.

[0120] For example, there are m resources in the candidate resource set, and the number of resource subsets to be screened is N. In the order of recommendation scores, A1 represents the first resource in the candidate resource set, A2 represents the second resource in the candidate resource set, ..., A m Indicates the mth resource in the candidate resource set. Specifically, when constructing a sequence of resources to be recommended based on the candidate resource set, the screening process is as follows.

[0121] 1) The process of selecting the first resource top1 is as follows:

[0122] like Figure 3aAs shown, the sequence of resources to be recommended includes m resources sorted in sequence according to the candidate resource set. According to the sorting in the candidate resource set, for the first resource A1 in the m resources, where K=1, K-1=0, since e(0)=0 indicates that there is no resource, then: e(K-1)=e(0)=X-0=X;

[0123] When A1 is a resource of the target category, e(K)=e(1)=X-1; when A1 is not a resource of the target category, e(K)=e(1)=X-0=X;

[0124] The integral term requires the integration of A1, thus Substitute the obtained related parameters into formula (1) to obtain the adjustment factor for the first resource A1. Modify the recommended score of A1 to obtain the modified score of A1.

[0125] According to the ranking in the candidate resource set, for the second resource A2 in the m resources, K=2, including resource A1 and resource A2. K-1=1, including resource A1. Then: e(K-1)=e(1); when resource A1 is a target category resource, e(1)=1; when resource A1 is not a target category resource, e(1)=0.

[0126] Based on resource A1 and resource A2 as a resource set of resource A2, determine e(K)=e(2);

[0127] The integral term needs to perform an integral operation on the resource set [A1] corresponding to K-1 and the resource set [A1, A2] corresponding to K, thus Substitute the obtained relevant parameters into formula (1) to obtain the adjustment factor for the first resource A2, thereby obtaining the corrected score of A2.

[0128] In this way, the recommended scores of the m resources are revised respectively to obtain the revised scores of the m resources.

[0129] Afterwards, the resource corresponding to the maximum value is selected from the corrected scores of the m resources, such as Figure 3a The assumption shown is A2, which is the top 1 filtered out.

[0130] 2) The process of screening the i-th resource topi is as follows:

[0131] Taking top2 as an example, since A2 is selected as top1, Figure 3a As shown, the sequence of resources to be recommended is constructed in the order of candidate resource sets in front and difference sets in the back, namely A2, A1, A3, A4...Am.

[0132] According to the sequence of resources to be recommended, the recommendation scores of each resource in the remaining resource set [A1, A3, A4…Am] are corrected to obtain the corrected scores of each resource in the remaining resource set. The resource with the largest corrected score is selected from the remaining resource set as the top2.

[0133] Taking A1 in the remaining resource set as an example, K=2, corresponding to the resource set [A2, A1], and the resource set [A2] corresponding to K-1, the revised score of A1 is determined by these two resource sets. Taking A3 in the remaining resource set as an example, K=3, corresponding to the resource set [A2, A1, A3], and the resource set [A2, A1] corresponding to K-1, the revised score of A3 is determined by these two resource sets.

[0134] Similarly, the N resources can be filtered out one by one through iterative execution.

[0135] In another embodiment, the sequence to be recommended is obtained by sorting the existing resources and the resources to be adjusted in the resource subset. Assume that there are m resources in the candidate resource set and the number of resource subsets to be screened is N. In which, according to the recommendation score sorting, A1 represents the first resource in the candidate resource set, A2 represents the second resource in the candidate resource set, ..., A m Indicates the mth resource in the candidate resource set. Specifically:

[0136] 1) The process of selecting the first resource top1 is as follows:

[0137] like Figure 3b As shown, the sequence of resources to be recommended is empty.

[0138] According to the ranking in the candidate resource set, for the first resource A1 in the m resources, where K=1, K-1=0, since e(0)=0 indicates no resource, then: e(K-1)=e(0)=X-0=X;

[0139] When A1 is a resource of the target category, e(K)=e(1)=X-1; when A1 is not a resource of the target category, e(K)=e(1)=X-0=X;

[0140] The integral term requires the integration of A1, thus Substitute the obtained relevant parameters into formula (1) to obtain the adjustment factor for the first resource A1, thereby obtaining the corrected score of A1.

[0141] According to the ranking in the candidate resource set, for the second resource A2 in the m resources, where K=1, K-1=0, since e(0)=0 indicates no resource, then: e(K-1)=e(0)=X-0=X;

[0142] When A2 is a resource of the target category, e(K)=e(1)=X-1; when A2 is not a resource of the target category, e(K)=e(1)=X-0=X;

[0143] The integral term requires the integration of A2, so Substitute the obtained relevant parameters into formula (1) to obtain the adjustment factor for the first resource A2, thereby obtaining the corrected score of A2.

[0144] In summary, each resource in the candidate resource set is processed in the same way, and the recommended scores of the m resources are respectively corrected to obtain the corrected scores of the m resources.

[0145] Afterwards, the resource corresponding to the maximum value is selected from the corrected scores of the m resources, such as Figure 3b The assumption shown is A2, which is the top 1 filtered out.

[0146] 2) The process of screening the i-th resource topi is as follows:

[0147] Taking top2 as an example, since A2 is selected as top1, the resource subset is [A2], and the difference between the resource subset and the candidate resource set is [A1, A3, A4…Am]. The resource sequence to be recommended is constructed from the resource subset [A2], such as Figure 3b Shown is [A2].

[0148] According to the sequence of resources to be recommended, the recommendation scores of each resource in the remaining resource set [A1, A3, A4…Am] are corrected to obtain the corrected scores of each resource in the remaining resource set. The resource with the largest corrected score is selected from the remaining resource set as the top2.

[0149] Taking A1 in the remaining resource sets as an example, K=2, corresponding to the resource set [A2, A1], and K-1 corresponding to the resource set [A2]. The corrected score of A1 is determined by these two resource sets.

[0150] Taking A3 in the remaining resource sets as an example, K=2, corresponding to the resource sets [A2, A3], and K-1 corresponding to the resource set [A2]. The corrected score of A3 is determined by these two resource sets.

[0151] Similarly, the N resources can be filtered out one by one through iterative execution.

[0152] It should be noted that Figure 3a and Figure 3b In the example, the resource sorting displayed on the front-end page is only an example. During implementation, the display sorting and display method can be determined according to the processing strategy after the resources are screened out, and the present disclosure does not limit this.

[0153] In summary, in the embodiment of the present disclosure, the overall process of the resource recommendation method is as follows: Figure 3c As shown, including:

[0154] S301 , constructing a sequence of resources to be recommended when the i-th (i ranges from 1 to N) resource is selected from a candidate resource set.

[0155] Taking the resource sequence to be recommended including m resources as an example, the resource sequence to be recommended is constructed in the order of resources in the resource subset first and resources in the difference set later. The difference set is the difference between the candidate resource set and the resource subset, and the resources in the difference set in the resource sequence to be recommended are sorted according to the candidate resource set.

[0156] Taking the example that the sequence of resources to be recommended needs to be constructed based on resource subsets, when the resource subset is empty, the sequence of resources to be recommended includes one resource to be adjusted. When the resource subset includes at least one existing resource, the existing resources are sorted in the order in which they are added to the resource subset and the resource to be adjusted is sorted at the end to obtain the sequence of resources to be recommended. Of course, the order may not be followed, as long as the resource to be recommended includes more existing resources in all resource subsets and is sorted before the resource to be adjusted.

[0157] During implementation, the selected resources are removed from the candidate resource set to obtain the remaining resource set, and each resource in the remaining resource set is used as a resource to be adjusted.

[0158] S302: For each resource to be adjusted, perform the following operations respectively:

[0159] S3021, determine the resource set corresponding to the resource to be adjusted, that is, including the resources before the recommended order of the resource to be adjusted in the sequence of recommended resources and the resource to be adjusted. And determine the resource set of the previous resource of the resource to be adjusted in the sequence of recommended resources, that is, including the resources in the sequence of recommended resources up to the previous resource.

[0160] S3022: Determine a first proportion of resources of the target category in a resource set corresponding to the resource to be adjusted, and a second proportion of resources of the target category in a resource set corresponding to the previous resource.

[0161] S3023: Based on the first proportion, the second proportion, and the expected proportion, an impact factor of the resource to be adjusted is determined by a PID method.

[0162] S3024, based on the sigmoid activation function that satisfies the mapping rule, the influencing factor is mapped to the target value range to obtain the adjustment factor.

[0163] S3025: Modify the recommended score of the resource to be adjusted based on the adjustment factor to obtain a modified score of the resource to be adjusted.

[0164] S303, selecting the resource with the largest corrected score from the remaining resource set as the topi-th resource, and adding it to the resource subset.

[0165] S304: When the number of resources in the resource subset is less than N, the process returns to S301; when the number of resources in the resource subset is equal to N, the screening process ends.

[0166] Of course, in the case where the resource subset needs to be further filtered later, if at least one resource is transitioned out of the resource subset, a new resource is added to the resource subset based on the method of FIG. 3 .

[0167] In other embodiments, for example, Figure 4 As shown, it is assumed that m resources are screened out from a large number of resources based on icf or ucf, and the recommended scores of each resource are sorted to obtain a sequence of resources to be recommended. Each time the target object is refreshed, n resources need to be recommended to it. After obtaining the corrected score for each resource to be adjusted, the sorting of the resource to be adjusted in the candidate resource set can be determined to obtain a new sequence of resources to be recommended. Recommending resources to the target object can be implemented as selecting the first n resources from the new sequence of resources to be recommended, and determining the resource proportion of the target category in the first n resources. If the difference between the resource proportion of the target category and the expected proportion is less than the preset difference, the n resources are recommended to the target object. Among them, m is greater than n, and m and n are both positive integers. In another case, if the resource proportion of the target category in the first n resources does not meet the above requirements, the new sequence of resources to be recommended is re-sorted according to the PID method until the first n resources meet the above requirements.

[0168] Based on the same technical concept, the embodiment of the present disclosure also provides a resource recommendation device 500, such as Figure 5 As shown, the device comprises:

[0169] The first acquisition module 501 is used to acquire a resource to be adjusted and a resource set corresponding to the resource to be adjusted; the resource to be adjusted is any resource in the sequence of resources to be recommended, and the resource set includes the resource to be adjusted and the resource before the resource to be adjusted in the sequence of resources to be recommended;

[0170] A first proportion determination module 502, configured to determine a first proportion of resources of a target category in a resource set;

[0171] An adjustment factor determination module 503 is used to determine the adjustment factor of the resource to be adjusted by a PID method based on the first proportion and the expected proportion, so that the first proportion approaches the expected proportion;

[0172] A first correction module 504, configured to correct the recommended score of the resource to be adjusted based on the adjustment factor to obtain a corrected score of the resource to be adjusted;

[0173] The recommendation module 505 is used to recommend resources to the target object based on the corrected scores of the resources to be adjusted.

[0174] In some embodiments, the PID method includes a proportional term, an integral term and a differential term, wherein the differential term is used to represent the change between a first gap and a second gap, the first gap being the gap between a first proportion and an expected proportion, and the second gap being the gap between a second proportion of resources of the target category in a resource set corresponding to reference resources and the expected proportion when adjacent resources of the resources to be adjusted in the recommended resource sequence are used as the resources to be adjusted.

[0175] In some embodiments, the expression of the PID method is:

[0176]

[0177] Among them, U(K) represents the impact factor of the resource to be adjusted obtained based on the PID method, and the impact factor is used to determine the adjustment factor; K represents the resource to be adjusted, and represents the recommended ranking of the resource to be adjusted in the sequence of resources to be recommended; (K-1) represents the reference resource with a recommended ranking of (K-1) in the sequence of resources to be recommended; e(K) represents the difference between the first proportion and the expected proportion; e(K-1) represents the difference between the second proportion and the expected proportion, and the second proportion is the proportion of the target category resources in the resource set corresponding to the reference resource; e(n) represents the difference between the third proportion and the expected proportion, and the third proportion is the proportion of the target category resources in the resource set corresponding to the resource with a recommended ranking of n; K p represents the weight of the proportional term, K i represents the weight of the integral term, K d Represents the weight of the derivative term.

[0178] In some embodiments, a weight determination module is further included, which is used to determine the weight of any one of the proportional term, the integral term and the differential term in the PID method based on the following method:

[0179] Determine the difference between the first share and the expected share;

[0180] Used to determine the weight of any item based on the mapping relationship between gaps and weights.

[0181] In some embodiments, the adjustment factor determination module includes:

[0182] An impact factor determination unit, used for determining the impact factor through an expression of a PID method;

[0183] A mapping unit, used for mapping the influencing factor to a target value range based on a mapping rule to obtain an adjustment factor;

[0184] The mapping rules meet the following conditions:

[0185] When the impact factor is a negative number, the impact factor is mapped to the first sub-interval of the target value interval, and the first sub-interval is used to reduce the recommended ranking of the resources to be adjusted;

[0186] When the impact factor is a positive number, the impact factor is mapped to the second sub-interval of the target value interval, and the second sub-interval is used to improve the recommended ranking of the resources to be adjusted.

[0187] In some embodiments, the mapping unit is specifically configured to:

[0188] Based on the sigmoid activation function that satisfies the mapping rules, the influencing factors are mapped to the target value range.

[0189] In some embodiments, the resource recommendation device further includes:

[0190] An interest determination module is used to determine the interest of the target object in the target category of resources within a specified time period;

[0191] The expected proportion determination module is used to adjust the interest level based on a preset bias to obtain the expected proportion.

[0192] In some embodiments, the interest determination module includes:

[0193] A first acquisition unit is used to acquire a set of recommended resources for a target object within a specified time period;

[0194] A second acquisition unit is used to acquire the operation record of the target object on the resources of the target category in the recommended resource set;

[0195] A screening unit, configured to screen resources having a target operation from resources of a target category in a recommended resource set based on the operation record;

[0196] The proportion determination unit is used to determine the proportion of resources with target operations in the recommended resources to obtain the interest degree.

[0197] In some embodiments, the resource recommendation device further includes:

[0198] The second correction module is used to correct the expected proportion to be less than or equal to the expected upper limit when the expected proportion is greater than the expected upper limit.

[0199] In some embodiments, the resource recommendation device further includes:

[0200] The third correction module is used to correct the value of the integral term to be less than or equal to the integral upper limit when the value of the integral term in the PID method is greater than the integral upper limit.

[0201] In some embodiments, the resource recommendation device further includes a target category determination module, which is used to:

[0202] Obtain recommendation records for target objects within a preset time period;

[0203] The categories whose resource quantity is lower than the preset threshold are selected from the recommended records as the target categories.

[0204] In some embodiments, the target category is a category in a category set, the category set includes at least two categories, and each category includes multiple subcategories.

[0205] In some embodiments, when a resource subset is screened out from a candidate resource set and recommended to a target object, the sequence of resources to be recommended includes existing resources in the resource subset and resources to be adjusted, and the existing resources are sorted in the order of being added to the resource subset, and the resources to be adjusted are sorted at the end; or, the sequence of resources to be recommended includes the resource subset, and a difference set between the candidate resource set and the resource subset, and in the sequence of resources to be recommended, the resources in the resource subset are sorted before the difference set, and the resources in the difference set are sorted according to the candidate resource set;

[0206] Recommended modules for:

[0207] When the existing resources in the resource subset do not reach the specified number and the corrected score of the resource to be adjusted is the maximum value of the corrected scores of all resources in the remaining resource set, the resource to be adjusted is added to the resource subset; wherein the remaining resource set is the resources in the sequence of resources to be recommended that have not been added to the resource subset.

[0208] For the description of specific functions and examples of each module and unit of the device in the embodiment of the present disclosure, reference can be made to the relevant description of the corresponding steps in the above method embodiment, which will not be repeated here.

[0209] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium and a computer program product.

[0210] Figure 6A schematic block diagram of an example electronic device 600 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0211] like Figure 6 As shown, the device 600 includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. In the RAM 603, various programs and data required for the operation of the device 600 can also be stored. The computing unit 601, the ROM 602, and the RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0212] A number of components in the device 600 are connected to the I / O interface 605, including: an input unit 606, such as a keyboard, a mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a disk, an optical disk, etc.; and a communication unit 609, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 609 allows the device 600 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0213] The computing unit 601 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 601 performs the various methods and processes described above, such as the resource recommendation method. For example, in some embodiments, the resource recommendation method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 608. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the resource recommendation method described above may be performed. Alternatively, in other embodiments, the computing unit 601 may be configured to perform the resource recommendation method in any other appropriate manner (e.g., by means of firmware).

[0214] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0215] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0216] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0217] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0218] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0219] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0220] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0221] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the principles of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. A resource recommendation method, comprising: Acquire the resource to be adjusted and the resource set corresponding to the resource to be adjusted; Any resource in the sequence of resources to be recommended is used as the resource to be adjusted, wherein the resource set includes the resource to be adjusted and resources before the resource to be adjusted in the sequence of resources to be recommended; The sequence of resources to be recommended is screened based on a resource scoring method; Determine a first proportion of resources of the target category in the resource set; Based on the first proportion and the expected proportion, determining an adjustment factor of the resource to be adjusted by a PID method, so that the first proportion approaches the expected proportion; Modifying the recommended score of the resource to be adjusted based on the adjustment factor to obtain a modified score of the resource to be adjusted; Recommending resources to a target object based on the corrected scores of the resources to be adjusted; Wherein, the expression of the PID method is: Among them, U(K) represents the impact factor of the resource to be adjusted obtained based on the PID method, and the impact factor is used to determine the adjustment factor; K represents the resource to be adjusted, and represents the recommended ranking of the resource to be adjusted in the sequence of resources to be recommended; (K-1) represents the reference resource with a recommended ranking of (K-1) in the sequence of resources to be recommended; e(K) represents the difference between the first proportion and the expected proportion; e(K-1) represents the difference between the second proportion and the expected proportion, and the second proportion is the proportion of the resources of the target category in the resource set corresponding to the reference resource; e(n) represents the difference between the third proportion and the expected proportion, and the third proportion is the proportion of the resources of the target category in the resource set corresponding to the resources with a recommended ranking of n; K p represents the weight of the proportional term, K i represents the weight of the integral term, k d Represents the weight of the derivative term.

2. The method according to claim 1, wherein: Determine the weight of any of the proportional, integral, and derivative terms in the PID method, including: Determining a difference between the first proportion and the expected proportion; Based on the mapping relationship between the gap and the weight, the weight of any one of the items is determined.

3. The method according to claim 1, wherein: The determining the adjustment factor of the resource to be adjusted by the PID method includes: Determine the influencing factor through the expression of the PID method; Based on a mapping rule, the influencing factor is mapped to a target value range to obtain the adjustment factor; The mapping rule satisfies the following conditions: In the case where the impact factor is a negative number, the impact factor is mapped to a first sub-interval of the target value interval, and the first sub-interval is used to reduce the recommended ranking of the resource to be adjusted; In the case where the impact factor is a positive number, the impact factor is mapped to a second sub-interval of the target value interval, and the second sub-interval is used to improve the recommended ranking of the to-be-adjusted resources.

4. The method according to claim 3, wherein: Mapping the impact factor to a target value range based on a mapping rule includes: Based on a sigmoid activation function that satisfies the mapping rule, the influencing factor is mapped to the target value range.

5. The method according to any one of claims 1 to 4, further comprising determining the expected proportion based on the following method: Determining the interest of the target object in resources of the target category within a specified period of time; The interest level is adjusted based on a preset bias to obtain the expected proportion.

6. The method according to claim 5, wherein: The determining the interest of the target object in the resources of the target category within a specified period of time includes: Obtain a set of recommended resources for the target object within the specified time period; Acquire operation records of the target object on resources of the target category in the recommended resource set; Based on the operation record, filtering out resources having a target operation from resources of the target category in the recommended resource set; The proportion of resources having the target operation in the recommended resources is determined to obtain the interest level.

7. The method according to claim 5, further comprising: When the expected proportion is greater than the expected upper limit, the expected proportion is corrected to be less than or equal to the expected upper limit.

8. The method according to claim 1, further comprising: When the value of the integral term in the PID method is greater than the integral upper limit, the value of the integral term is corrected to be less than or equal to the integral upper limit.

9. The method of claim 1 , further comprising determining the target category based on: Obtaining recommendation records for the target object within a preset time period; The categories whose resource quantities are lower than a preset threshold are screened out from the recommendation records as the target categories.

10. The method according to claim 1, wherein: The target category is a category in a category set, the category set includes at least two categories, and each category includes multiple subcategories.

11. The method according to claim 1, wherein: In the case where a resource subset is screened out from the candidate resource set and recommended to the target object, the sequence of resources to be recommended includes existing resources in the resource subset and the resources to be adjusted, and the existing resources are sorted in the order of being added to the resource subset, and the resources to be adjusted are sorted last; or, the sequence of resources to be recommended includes the resource subset and a difference set between the candidate resource set and the resource subset, and in the sequence of resources to be recommended, the resources in the resource subset are sorted before the difference set, and the resources in the difference set are sorted according to the candidate resource set; The recommending resources to the target object based on the corrected scores of the resources to be adjusted includes: When the existing resources in the resource subset do not reach the specified number and the corrected score of the resource to be adjusted is the maximum value of the corrected scores of all resources in the remaining resource set, the resource to be adjusted is added to the resource subset; wherein the remaining resource set is the resources in the sequence of resources to be recommended that have not been added to the resource subset.

12. A resource recommendation device, comprising: A first acquisition module, used to acquire resources to be adjusted and resource sets corresponding to the resources to be adjusted; Any resource in the sequence of resources to be recommended is used as the resource to be adjusted, wherein the resource set includes the resource to be adjusted and resources before the resource to be adjusted in the sequence of resources to be recommended; The sequence of resources to be recommended is screened based on a resource scoring method; A first proportion determination module, used to determine a first proportion of resources of a target category in the resource set; an adjustment factor determination module, configured to determine, based on the first proportion and the expected proportion, an adjustment factor of the resource to be adjusted by a PID method, so that the first proportion approaches the expected proportion; A first correction module, configured to correct the recommended score of the resource to be adjusted based on the adjustment factor to obtain a corrected score of the resource to be adjusted; A recommendation module, configured to recommend resources to a target object based on the corrected scores of the resources to be adjusted; Wherein, the expression of the PID method is: Among them, U(K) represents the impact factor of the resource to be adjusted obtained based on the PID method, and the impact factor is used to determine the adjustment factor; K represents the resource to be adjusted, and represents the recommended ranking of the resource to be adjusted in the sequence of resources to be recommended; (K-1) represents the reference resource with a recommended ranking of (K-1) in the sequence of resources to be recommended; e(K) represents the difference between the first proportion and the expected proportion; e(K-1) represents the difference between the second proportion and the expected proportion, and the second proportion is the proportion of the resources of the target category in the resource set corresponding to the reference resource; e(n) represents the difference between the third proportion and the expected proportion, and the third proportion is the proportion of the resources of the target category in the resource set corresponding to the resources with a recommended ranking of n; K p represents the weight of the proportional term, K i represents the weight of the integral term, K d Represents the weight of the derivative term.

13. The apparatus according to claim 12, further comprising a weight determination module, configured to determine the weight of any one of a proportional term, an integral term, and a differential term in the PID method based on the following method: Determining a difference between the first proportion and the expected proportion; Used to determine the weight of any one of the items based on the mapping relationship between the gap and the weight.

14. The device according to any one of claims 12, wherein: The adjustment factor determination module includes: An influence factor determination unit, used to determine the influence factor through the expression of the PID method; A mapping unit, configured to map the influencing factor to a target value range based on a mapping rule to obtain the adjustment factor; The mapping rule satisfies the following conditions: In the case where the impact factor is a negative number, the impact factor is mapped to a first sub-interval of the target value interval, and the first sub-interval is used to reduce the recommended ranking of the resource to be adjusted; In the case where the impact factor is a positive number, the impact factor is mapped to a second sub-interval of the target value interval, and the second sub-interval is used to improve the recommended ranking of the to-be-adjusted resources.

15. The device according to claim 14, wherein: The mapping unit is specifically used for: Based on a sigmoid activation function that satisfies the mapping rule, the influencing factor is mapped to the target value range.

16. The apparatus according to any one of claims 12 to 15, further comprising: An interest determination module, used to determine the interest of the target object in the resources of the target category within a specified time period; The expected proportion determination module is used to adjust the interest level based on a preset bias to obtain the expected proportion.

17. The device according to claim 16, wherein: The interest level determination module comprises: A first acquisition unit, configured to acquire a set of recommended resources for the target object within the specified time period; A second acquisition unit, configured to acquire an operation record of the target object on the resources of the target category in the recommended resource set; A screening unit, configured to screen out resources having a target operation from resources of the target category in the recommended resource set based on the operation record; The proportion determining unit is used to determine the proportion of resources having the target operation in the recommended resources to obtain the interest level.

18. The apparatus according to claim 16, further comprising: The second correction module is used to correct the expected proportion to be less than or equal to the expected upper limit when the expected proportion is greater than the expected upper limit.

19. The apparatus according to claim 12, further comprising: The third correction module is used to correct the value of the integral term in the PID method to be less than or equal to the integral upper limit when the value of the integral term is greater than the integral upper limit.

20. The apparatus according to claim 12, further comprising a target category determination module, configured to: Obtaining recommendation records for the target object within a preset time period; The categories whose resource quantities are lower than a preset threshold are screened out from the recommendation records as the target categories.

21. The device according to claim 12, wherein: The target category is a category in a category set, the category set includes at least two categories, and each category includes multiple subcategories.

22. The device according to claim 12, wherein: In the case where a resource subset is screened out from the candidate resource set and recommended to the target object, the sequence of resources to be recommended includes existing resources in the resource subset and the resources to be adjusted, and the existing resources are sorted in the order of being added to the resource subset, and the resources to be adjusted are sorted last; or, the sequence of resources to be recommended includes the resource subset and a difference set between the candidate resource set and the resource subset, and in the sequence of resources to be recommended, the resources in the resource subset are sorted before the difference set, and the resources in the difference set are sorted according to the candidate resource set; The recommendation module is specifically used for: When the existing resources in the resource subset do not reach the specified number and the corrected score of the resource to be adjusted is the maximum value of the corrected scores of all resources in the remaining resource set, the resource to be adjusted is added to the resource subset; wherein the remaining resource set is the resources in the sequence of resources to be recommended that have not been added to the resource subset.

23. An electronic device, comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 11.

24. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1-11.

25. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 11.

Images