Resource recommendation method and device, electronic equipment, storage medium and program product

Abstract

The embodiment of the invention provides a resource recommendation method and device, electronic equipment, a storage medium and a program product, and relates to the technical field of artificial intelligence and cloud. The method comprises the steps of obtaining at least one first resource to be recommended to a target object, and determining a first resource fingerprint of each first resource in the at least one first resource; then obtaining a resource fingerprint pre-estimation group corresponding to at least one second resource set related to the target object, wherein the resource fingerprint pre-estimation group comprises a target resource fingerprint and a target pre-estimation value; then, for each first resource, matching the first resource fingerprint with the target resource fingerprint of each resource fingerprint pre-estimation group; and under the condition that the first resource fingerprint of the first resource is successfully matched with the target resource fingerprint of one resource fingerprint estimation group, recommending a target resource in the at least one first resource to the target object. According to the embodiment of the invention, the computing power resource consumption for calculating the estimated value can be reduced, so that the resource recommendation efficiency is improved.

Description

Technical Field

[0001] The present application relates to the field of data processing technology. Specifically, the present application relates to a resource recommendation method, device, electronic device, storage medium and program product. Background Art

[0002] Resource recommendation refers to recommending resources that the subject may like to the subject, and displaying them on the subject's mobile terminal for the subject to choose, such as recommending commodities, multimedia, links and other resources. Before recommending resources to the subject, it is generally necessary to filter out the resources to be recommended from the resources to be recommended.

[0003] In related technologies, a specific algorithm is required to calculate the estimated value of each resource in order to determine which resources need to be recommended to the object. For example, it is generally necessary to go through the screening stages of recall, rough sorting, and fine sorting to determine the estimated value of the resource.

[0004] However, using a specific algorithm to calculate the estimated value of each resource will result in a large consumption of computing power resources for calculating the estimated value, which in turn leads to a lower efficiency in resource recommendation. Summary of the invention

[0005] The embodiments of the present application provide a resource recommendation method, device, electronic device, storage medium and program product, which are used to solve the technical problem that calculating the estimated value of each resource through a specific algorithm will result in a large consumption of computing power resources for calculating the estimated value, and thus lead to a low efficiency of resource recommendation, thereby reducing the consumption of computing power resources for calculating the estimated value and improving the efficiency of resource recommendation.

[0006] On the one hand, an embodiment of the present application provides a resource recommendation method, including:

[0007] Acquire at least one first resource to be recommended to the target object, and determine a first resource fingerprint of each first resource in the at least one first resource, where the first resource fingerprint is determined based on multimedia material information of the first resource;

[0008] Acquire a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, the resource fingerprint estimation group set includes a resource fingerprint estimation group corresponding to each second resource set, the resource fingerprint estimation group includes a target resource fingerprint and a target estimation value, the target resource fingerprint corresponding to the second resource set is determined based on multimedia material information of a second resource in the second resource set, and the target estimation value corresponding to the second resource set represents the possibility that the second resource in the second resource set is selected by the target object;

[0009] For each first resource, matching a first resource fingerprint of the first resource with a target resource fingerprint of each resource fingerprint estimation group;

[0010] In the case where the first resource fingerprint of the first resource successfully matches the target resource fingerprint of one of the resource fingerprint estimation groups, determining an estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group, the estimated value of the first resource representing the possibility of the first resource being selected by the target object;

[0011] Based on the estimated values ​​of the first resources, a target resource in at least one first resource is recommended to the target object.

[0012] On the other hand, the embodiment of the present application further provides a resource recommendation device, including:

[0013] A first acquisition module, configured to acquire at least one first resource to be recommended to a target object, and determine a first resource fingerprint of each first resource in the at least one first resource, wherein the first resource fingerprint is determined based on multimedia material information of the first resource;

[0014] A second acquisition module is used to acquire a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, the resource fingerprint estimation group set includes a resource fingerprint estimation group corresponding to each second resource set, the resource fingerprint estimation group includes a target resource fingerprint and a target estimation value, the target resource fingerprint corresponding to the second resource set is determined based on the multimedia material information of the second resource in the second resource set, and the target estimation value corresponding to the second resource set represents the possibility that the second resource in the second resource set is selected by the target object;

[0015] A matching module, configured to match, for each first resource, a first resource fingerprint of the first resource with a target resource fingerprint of each resource fingerprint estimation group;

[0016] An estimated value determination module is used to determine an estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group when the first resource fingerprint of the first resource successfully matches the target resource fingerprint of one of the resource fingerprint estimation groups, wherein the estimated value of the first resource represents the possibility that the first resource is selected by the target object;

[0017] The resource recommendation module is used to recommend a target resource from at least one first resource to a target object based on the estimated value of each first resource.

[0018] Optionally, the estimated value determination module is further used for:

[0019] In the case that the first resource fingerprint of the first resource fails to match the target resource fingerprint of any resource fingerprint estimation group, the estimated value of the first resource is determined based on a preset estimated value determination algorithm.

[0020] Optionally, a second resource set includes one or more second resources, and the resource fingerprint estimation group corresponding to the second resource set is determined by the second acquisition module in the following manner:

[0021] For a second resource set including a second resource, determining a second resource fingerprint of the second resource, and determining the second resource fingerprint of the second resource as a target resource fingerprint corresponding to the second resource set, and determining a second estimated value of the second resource, and determining the second estimated value of the second resource as a target estimated value corresponding to the second resource set, the second estimated value representing the possibility of the second resource being selected by the target object;

[0022] For a second resource set including multiple second resources, determine the second resource fingerprint of each second resource in the multiple second resources, and fuse the second resource fingerprints of each second resource in the multiple second resources to obtain a fused second resource fingerprint, and determine the fused second resource fingerprint as a target resource fingerprint corresponding to the second resource set, and determine the second estimated value of each second resource in the multiple second resources, and fuse the second estimated value of each second resource in the multiple second resources to obtain a fused second estimated value, and determine the fused second estimated value as a target estimated value corresponding to the second resource set, the resource fingerprint similarity between any two second resources in the multiple second resources is greater than a resource fingerprint similarity threshold, and the second resource fingerprint of the second resource is determined based on the multimedia material information of the second resource.

[0023] Optionally, at least one of the first resource fingerprint of the first resource or the second resource fingerprint of the second resource is determined by the first acquisition module in the following manner:

[0024] Extracting at least one item of resource information of a fingerprint resource to be determined, where the fingerprint resource to be determined is at least one of the first resource or the second resource, and the at least one item of resource information includes multimedia material information of the fingerprint resource to be determined;

[0025] Determine the creative fingerprint corresponding to the multimedia material information;

[0026] Based on the creative fingerprint corresponding to the multimedia material information, a resource fingerprint of the fingerprint resource to be determined is obtained.

[0027] Optionally, at least one item of information resource information also has to determine landing page information corresponding to the fingerprint resource, the landing page information indicates the landing page to be redirected when the fingerprint resource to be determined is selected, and the first acquisition module is further used to:

[0028] Determine the landing page fingerprint corresponding to the landing page information;

[0029] When the first acquisition module obtains the resource fingerprint of the fingerprint resource to be determined based on the creative fingerprint corresponding to the multimedia material information, it can be used to:

[0030] Determine respectively a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information;

[0031] A weighted calculation is performed based on the creative fingerprint corresponding to the multimedia material information, the first fusion weight, the landing page fingerprint corresponding to the landing page information, and the second fusion weight to obtain a resource fingerprint of the fingerprint resource to be determined, where the resource fingerprint includes a hash value.

[0032] Optionally, when the first acquisition module determines the first fusion weight corresponding to the multimedia material information and the second fusion weight corresponding to the landing page information respectively, it can be used to:

[0033] Determining the material richness of the multimedia material information, the material type of the multimedia material information includes at least one of text or image, and the material richness is related to at least one of the number of material types of the multimedia material information, the length of the text, the number of images, or the number of image elements of the images;

[0034] Determine, based on the material richness, a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information;

[0035] Among them, the first fusion weight is positively correlated with the richness of the material, and the second fusion weight is negatively correlated with the richness of the material.

[0036] Optionally, when matching the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group, the matching module may be used to:

[0037] Performing similarity matching processing on the first resource fingerprint of the first resource and the target resource fingerprint of each resource fingerprint estimation group to obtain the resource fingerprint similarity corresponding to each resource fingerprint estimation group;

[0038] For each resource fingerprint estimation group, when the resource fingerprint similarity corresponding to the resource fingerprint estimation group is greater than or equal to the resource fingerprint similarity threshold, it is determined that the resource fingerprint estimation group is successfully matched with the first resource; and when the resource fingerprint similarity corresponding to the resource fingerprint estimation group is less than the resource fingerprint similarity threshold, it is determined that the resource fingerprint estimation group fails to match the first resource.

[0039] Optionally, each resource fingerprint estimation group corresponds to a random number probability threshold, and the random number probability threshold corresponding to each resource fingerprint estimation group is determined by the matching module in the following manner:

[0040] For each resource fingerprint estimation group, based on the second estimated value of the second resource in the second resource set corresponding to the resource fingerprint estimation group, determine the estimated value mean and estimated value standard deviation of the second resource set corresponding to the resource fingerprint estimation group;

[0041] For each resource fingerprint estimation group, based on the estimated value mean and estimated value standard deviation corresponding to the resource fingerprint estimation group, determine the coefficient of variation corresponding to the resource fingerprint estimation group;

[0042] Based on the coefficient of variation corresponding to each resource fingerprint estimation group, determine the random number probability threshold corresponding to each resource fingerprint estimation group;

[0043] When determining the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group, the estimated value determination module can be used to:

[0044] Determine the random number probability based on the target estimated value of the successfully matched resource fingerprint estimation group;

[0045] When the random number probability is greater than the random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimated value of the first resource is determined based on the target estimated value of the successfully matched resource fingerprint estimation group.

[0046] Optionally, the estimated value determination module is further used for:

[0047] When the random number probability is less than or equal to the random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimation value of the first resource is determined based on a preset estimation value determination algorithm.

[0048] Optionally, when determining the random number probability threshold corresponding to each resource fingerprint estimation group based on the coefficient of variation corresponding to each resource fingerprint estimation group, the matching module may be used to:

[0049] Based on the coefficient of variation corresponding to each resource fingerprint estimation group, determine the maximum coefficient of variation and the minimum coefficient of variation;

[0050] Based on the maximum coefficient of variation and the minimum coefficient of variation, at least two coefficient of variation intervals are divided;

[0051] Determine the random number probability threshold corresponding to each coefficient of variation interval. For each coefficient of variation interval, the random number probability threshold corresponding to the coefficient of variation interval is negatively correlated with the interval maximum value corresponding to the coefficient of variation interval.

[0052] For each resource fingerprint estimation group, determine the variation coefficient interval to which the resource fingerprint estimation group belongs, and determine the random number probability threshold corresponding to the variation coefficient interval to which the resource fingerprint estimation group belongs as the random number probability threshold corresponding to the resource fingerprint estimation group.

[0053] Optionally, when determining the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group, the estimated value determination module may be used to:

[0054] Determine the similarity difference between the resource fingerprint similarity corresponding to the successfully matched resource fingerprint estimation group and the resource fingerprint similarity threshold;

[0055] When the similarity difference is less than the similarity difference threshold, the target estimated value of the successfully matched resource fingerprint estimation group is determined as the estimated value of the first resource;

[0056] When the similarity difference is greater than or equal to the similarity difference threshold, the target estimated value of the successfully matched resource fingerprint estimation group is adjusted, and the adjusted target estimated value is determined as the estimated value of the first resource.

[0057] Optionally, when adjusting the target estimated value of the successfully matched resource fingerprint estimation group, the estimated value determination module may be used to:

[0058] Acquire a target object feature of the target object, and acquire a differential resource feature in the first resource, where the differential resource feature is different from a second resource feature of any second resource in the second resource set corresponding to the successfully matched resource fingerprint estimation group;

[0059] Obtain a trained deep neural network model, and use the target object features and the difference resource features as inputs of the deep neural network model to obtain a first output estimate value of the deep neural network model output, wherein the deep neural network model is trained by using the training object features and the training resource features in the training data as inputs and the training estimate value in the training data as outputs of the deep neural network model;

[0060] When the first output estimated value is greater than the estimated value threshold, increasing the target estimated value of the successfully matched resource fingerprint estimated group;

[0061] When the first output estimated value is less than or equal to the estimated value threshold, the target estimated value of the successfully matched resource fingerprint estimation group is reduced.

[0062] Optionally, the estimated value determination module is further used for:

[0063] Obtaining a determined number of a second estimated value determined based on the estimated value determination algorithm, and determining a quantity difference between a target number and the determined number, wherein the target number is less than or equal to a maximum determined number, and the maximum determined number is: when the second estimated value of the second resource is determined based on the estimated value determination algorithm, the maximum number of the second estimated value determined that is supported by the computing power resources of the electronic device;

[0064] Obtaining a third resource with a quantity difference, where the third resource is used to explore the preferences of the target object;

[0065] Determine an estimated value of each third resource in the number difference third resources based on a preset estimated value determination algorithm, where the estimated value of the third resource represents the possibility that the third resource is selected by the target object;

[0066] The resource recommendation module is also used to recommend a target resource among the quantity difference third resources to the target object based on the estimated value of each third resource among the quantity difference third resources.

[0067] The second estimated value of the second resource is determined by the estimated value determination module based on a preset estimated value determination algorithm. When executing the estimated value determination algorithm, the estimated value determination module may be used to:

[0068] Acquire a target object feature of a target object, and acquire a resource feature of a resource with an estimated value to be determined, where the resource with an estimated value to be determined is at least one of a first resource, a second resource, or a third resource;

[0069] Obtain a trained deep neural network model, and use the target object characteristics and the resource characteristics of the estimated value resource to be determined as inputs of the deep neural network model to obtain a second output estimated value output by the deep neural network model, wherein the deep neural network model is trained by using the training object characteristics and the training resource characteristics in the training data as inputs and using the training estimated value in the training data as outputs of the deep neural network model;

[0070] The second output estimated value is determined as the estimated value of the estimated value resource to be determined.

[0071] On the other hand, an embodiment of the present application further provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method of the embodiment of the present application.

[0072] On the other hand, an embodiment of the present application further provides a computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the steps of the method of the embodiment of the present application are implemented.

[0073] On the other hand, an embodiment of the present application further provides a computer program product, including a computer program, which implements the steps of the method of the embodiment of the present application when the computer program is executed by a processor.

[0074] The technical solution of this embodiment is to obtain at least one first resource to be recommended to the target object, and determine the first resource fingerprint of each first resource in the at least one first resource; then obtain a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, and the resource fingerprint estimation group set includes the resource fingerprint estimation group corresponding to each second resource set, and the resource fingerprint estimation group includes a target resource fingerprint and a target estimated value; then, for each first resource, match the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group; and when the first resource fingerprint of the first resource successfully matches the target resource fingerprint of one of the resource fingerprint estimation groups, determine the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group; based on the estimated value of each first resource, recommend the target resource in at least one first resource to the target object, because the estimated value of the first resource is determined in a matching manner, then for the first resource, it is not necessary to use a complex algorithm to determine the corresponding estimated value, which can reduce the computing power resource consumption of calculating the estimated value, thereby improving the efficiency of resource recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0075] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in describing the embodiments of the present application are briefly introduced below.

[0076] Figure 1 A schematic diagram of an application environment provided for an embodiment of the present application;

[0077] Figure 2 A schematic diagram of an implementation environment of a resource recommendation method provided in an embodiment of the present application;

[0078] Figure 3 A schematic diagram of a resource recommendation method provided in an embodiment of the present application;

[0079] Figure 4 A flowchart of another resource recommendation method provided in an embodiment of the present application;

[0080] Figure 5 A schematic diagram of a difference resource feature provided in an embodiment of the present application;

[0081] Figure 6 A schematic diagram of calculating an estimated value of a resource based on a preset estimated value determination algorithm provided in an embodiment of the present application;

[0082] Figure 7 A schematic diagram of a flow chart of an estimated value of a computing resource provided in an embodiment of the present application;

[0083] Figure 8 A flowchart of another resource recommendation method provided in an embodiment of the present application;

[0084] Fig. 9 A schematic diagram of the structure of a resource recommendation device provided in an embodiment of the present application;

[0085] Fig.10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0086] The embodiments of the present application are described below in conjunction with the drawings in the present application. It should be understood that the implementation methods described below in conjunction with the drawings are exemplary descriptions for explaining the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions of the embodiments of the present application.

[0087] It will be understood by those skilled in the art that, unless specifically stated, the singular forms "one", "an", and "the" used herein may also include plural forms. It should be further understood that the terms "including" and "comprising" used in the embodiments of the present application refer to that the corresponding features can be implemented as the presented features, information, data, steps, operations, elements and / or components, but do not exclude the implementation as other features, information, data, steps, operations, elements, components and / or combinations thereof supported by the technical field. It should be understood that when we call an element "connected" or "coupled" to another element, the one element may be directly connected or coupled to the other element, or it may refer to that the one element and the other element establish a connection relationship through an intermediate element. In addition, the "connection" or "coupling" used here may include wireless connection or wireless coupling. The term "and / or" used here represents at least one of the items defined by the term, for example, "A and / or B" represents the implementation as "A", or as "A", or as "A and B". "Multiple" refers to at least two.

[0088] In order to make the objectives, technical solutions and advantages of the present application clearer, the implementation methods of the present application will be further described in detail below with reference to the accompanying drawings.

[0089] See also Figure 1 , Figure 1 A schematic diagram of an application environment provided by an embodiment of the present application. Figure 1As shown, when the object uses a mobile terminal to browse some applications, some resources that the object may be interested in will be recommended to the mobile terminal, so that the recommended resources are exposed in the application for the object to choose. For example, it can be recommended goods, multimedia and links, etc. Among them, multimedia can refer to video as a combination of multiple media, generally including multiple media forms such as text, sound and image. Specifically, the resources can be resources related to the function of the application. Taking the application as a news application as an example, the resources recommended to the mobile terminal can be news resources, such as the latest current affairs, etc. For another example, taking the application as a video application as an example, the resources recommended to the mobile terminal can be video resources, such as movie resources. In addition, the resources can also be resources unrelated to the function of the application. Taking the application as a news application as an example, the resources recommended to the mobile terminal can be an advertisement, such as a jump link to other applications, a purchase link for goods, etc., which is not limited here. For another example, taking the application as a video application as an example, the resources recommended to the mobile terminal can be news resources, etc., which is not limited here.

[0090] See also Figure 2 , Figure 2 Schematic diagram of an implementation environment of a resource recommendation method provided in an embodiment of the present application. The resource recommendation method provided in an embodiment of the present application can be applied to Figure 2 In the implementation environment shown, the terminal 202 communicates with the server 204 via a network. The data storage system can store data that the server 204 needs to process. The data storage system can be integrated on the server 204, or placed on the cloud or other network servers.

[0091] In this embodiment, the server 204 can obtain at least one first resource to be recommended to the target object, and determine the first resource fingerprint of each first resource in the at least one first resource; obtain a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, the resource fingerprint estimation group set includes a resource fingerprint estimation group corresponding to each second resource set, and the resource fingerprint estimation group includes a target resource fingerprint and a target estimation value; for each first resource, match the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group; when the first resource fingerprint of the first resource successfully matches the target resource fingerprint of one of the resource fingerprint estimation groups, determine the estimation value of the first resource based on the target estimation value of the successfully matched resource fingerprint estimation group, and the estimation value of the first resource represents the possibility of the first resource being selected by the target object; based on the estimation value of each first resource, recommend the target resource in the at least one first resource to the target object. Then the display device, such as the screen, of the mobile terminal 202 used by the target object can display the target resource for selection by the target object.

[0092] In another embodiment, the target resource may be determined in the background by the mobile terminal 202 , so that the target resource is displayed through the display device of the mobile terminal 202 .

[0093] The mobile terminal 202 may be, but is not limited to, various personal computers, laptops, smart phones, tablet computers, IoT devices, and portable wearable devices. The IoT devices may be smart speakers, smart TVs, smart air conditioners, smart car-mounted devices, etc. The portable wearable devices may be smart watches, smart bracelets, head-mounted devices, etc. The server 204 may be implemented as an independent server or a server cluster consisting of multiple servers.

[0094] In related technologies, specific algorithms are mainly used to calculate the estimated value of resources. Taking advertising as an example, the existing solutions are mainly based on the traffic value model, that is, from the perspective of the posterior value of the ad slot, the traffic value is differentiated and the fitting granularity is set at the granularity of the ad slot. For high-value ad slots, more sufficient and higher-quality computing resources are provided to ensure that more candidate ads can be calculated, while for low-value ad slots, elastic resources are used to ensure that they can meet the lower limit of the number of candidate ads.

[0095] However, research has found that from the perspective of the object, the same or similar materials have the same sensory perception for the object, so it can be considered that the behavior of the object is consistent on the same or similar materials. Generally, a portion of ads will be extracted from the ad library as recommended ads and stored in the candidate ad queue. Among the ads in the candidate ad queue, there will be more or less ads with the same or similar materials. Therefore, the algorithm's repeated calculations on the same or similar materials are to some extent a waste or unnecessary. However, the optimization of computing power for repeated calculations of ads with the same or highly similar materials is not considered by the current solution based on the traffic value model.

[0096] Therefore, in response to at least one of the above-mentioned technical problems or areas that need improvement in the relevant technology, the present application proposes a resource recommendation method, device, electronic device, storage medium and program product. The solution obtains at least one first resource to be recommended to the target object, and determines the first resource fingerprint of each first resource in the at least one first resource; then obtains a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, and the resource fingerprint estimation group set includes a resource fingerprint estimation group corresponding to each second resource set, the resource fingerprint estimation group includes a target resource fingerprint and a target estimation value, and the target estimation value corresponding to the second resource set represents the second resource set. the possibility of a second resource being selected by a target object; then, for each first resource, matching the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group; and in the case where the first resource fingerprint of the first resource is successfully matched with the target resource fingerprint of one of the resource fingerprint estimation groups, determining an estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group, the estimated value of the first resource representing the possibility of the first resource being selected by the target object; based on the estimated values ​​of each first resource, recommending at least one target resource in the first resources to the target object, thereby reducing the consumption of computing power resources for calculating the estimated value, and thereby improving the efficiency of resource recommendation.

[0097] The following describes several exemplary embodiments to illustrate the technical solutions of the embodiments of the present application and the technical effects produced by the technical solutions of the present application. It should be noted that the following embodiments can refer to, draw on or combine with each other, and the same terms, similar features and similar implementation steps in different embodiments will not be described repeatedly.

[0098] Optionally, the embodiments of the present application may also relate to the field of cloud technology. For example, the solution of the embodiments may be deployed and implemented on the cloud.

[0099] Among them, cloud technology is a general term for network technology, information technology, integration technology, management platform technology, application technology, etc. based on the cloud computing business model. It can form a resource pool, which is used on demand and is flexible and convenient. Cloud computing technology will become an important support. Cloud storage is a new concept extended and developed from the concept of cloud computing. Distributed cloud storage system (hereinafter referred to as storage system) refers to a storage system that uses cluster applications, grid technology, and distributed storage file systems to bring together a large number of different types of storage devices (storage devices are also called storage nodes) in the network through application software or application interfaces to work together and provide data storage and target business access functions to the outside world.

[0100] Optionally, embodiments of the present application may also involve the field of artificial intelligence (AI), for example, when an estimated value is calculated based on a preset estimated value determination algorithm, artificial intelligence may be used to implement it.

[0101] Artificial intelligence is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results. In other words, artificial intelligence is a comprehensive technology in computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a similar way to human intelligence. Artificial intelligence is to study the design principles and implementation methods of various intelligent machines so that machines have the functions of perception, reasoning and decision-making.

[0102] Artificial intelligence technology is a comprehensive discipline that covers a wide range of fields, including both hardware-level and software-level technologies. The basic technologies of artificial intelligence generally include sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technology, pre-trained model technology, operation / interaction systems, mechatronics, etc. Among them, pre-trained models are also called large models and basic models. After fine-tuning, they can be widely used in downstream tasks in various major directions of artificial intelligence. Artificial intelligence software technology mainly includes computer vision technology, speech processing technology, natural language processing technology, and machine learning / deep learning. Pre-trained models, an important technology for model training in the field of artificial intelligence, are developed from large language models in the field of natural language processing (NLP). Large language models are trained on large amounts of text data and can perform a wide range of tasks, including text summarization, translation, sentiment analysis, and so on. LLMs are characterized by their large scale and billions of parameters, which help them learn complex patterns in language data. These models are usually based on deep learning architectures such as transformers, which helps them achieve impressive performance on various NLP tasks.

[0103] See also Figure 3 , Figure 3 A schematic diagram of a resource recommendation method provided in an embodiment of the present application. Figure 3 The method shown in the figure may be executed by an electronic device, which may include at least one of a server or a mobile terminal. Figure 3 The methods shown may include:

[0104] S310: Acquire at least one first resource to be recommended to a target object, and determine a first resource fingerprint of each first resource in the at least one first resource.

[0105] Among them, the first resource may refer to the resource to be recommended. The first resource fingerprint is determined based on the multimedia material information of the first resource. Generally, the resource fingerprint is unique, that is, the resource fingerprints corresponding to different resources are generally different. The multimedia material information may include at least one of text material information or picture material information. Optionally, the resource fingerprint may represent the design creativity of the resource, or it may represent the attribute of the resource. The attribute of the resource may include but is not limited to at least one of the category or price of the resource, etc., which is not limited here. Specifically, if the resource fingerprint represents the attribute of the resource, the attribute of the resource can be extracted from the multimedia material information.

[0106] Optionally, the first resource may include but is not limited to at least one of a commodity, a recommended multimedia or a link, etc., which is not limited here. Generally, in order to attract the target, some special creativity is designed to attract the target's attention, thereby prompting the target to select the resource. Therefore, the resource fingerprint may include a creative fingerprint, which can be used to characterize the design creativity of the resource, and the creative fingerprint can be determined based on the multimedia material information of the resource.

[0107] S320, obtaining a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, wherein the resource fingerprint estimation group set includes a resource fingerprint estimation group corresponding to each second resource set, and the resource fingerprint estimation group includes a target resource fingerprint and a target estimation value.

[0108] Among them, for each second resource set, the target resource fingerprint corresponding to the second resource set is determined based on the multimedia material information of the second resource in the second resource set, and the target estimated value corresponding to the second resource set represents the possibility that the second resource in the second resource set is selected by the target object. Optionally, the larger the target estimated value, the higher the possibility.

[0109] The second resource set may include one or more second resources, and the second resources may be recommended resources or non-recommended resources.

[0110] Optionally, for each second resource set, its corresponding target resource fingerprint and target estimated value may be preset or calculated in a specific manner, which is not limited here.

[0111] S330: For each first resource, match the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group.

[0112] In this embodiment, the purpose of using the first resource fingerprint of the first resource as the target resource fingerprint of each resource fingerprint estimation group is to find a target resource fingerprint that is the same or similar to the first resource fingerprint from the target resource fingerprints of each resource fingerprint estimation group. If a target resource fingerprint is the same or similar to the first resource fingerprint, it means that the design creativity between the second resource in the second resource set corresponding to the target resource fingerprint and the first resource is the same or similar.

[0113] Optionally, in this embodiment, the target resource fingerprints of each resource fingerprint estimation group may be traversed to match the first resource fingerprint with the target resource fingerprint of each resource fingerprint estimation group.

[0114] S340: When the first resource fingerprint of the first resource successfully matches the target resource fingerprint of one of the resource fingerprint estimation groups, determine the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group.

[0115] The estimated value of the first resource represents the possibility of the first resource being selected by the target object. In this embodiment, if the first resource fingerprint successfully matches the target estimated value of one of the resource fingerprint estimated values, it means that the target resource fingerprint of the successfully matched resource fingerprint estimated group is the same as or similar to the first resource fingerprint. Optionally, the larger the estimated value of the first resource, the greater the possibility of it being selected by the target object.

[0116] S350: Recommend a target resource from at least one first resource to a target object based on the estimated value of each first resource.

[0117] Among them, the target resource may refer to a resource determined to be recommended. In this embodiment, based on the estimated value of each first resource, part or all of at least one first resource may be used as a target resource in order from large to small, so that the target resource is recommended to the target object. Optionally, the target resource may be stored in a candidate advertising queue, and when recommending resources, the resources in the candidate advertising queue are recommended to the target object. Optionally, the second resources in each second resource set may also be stored in the candidate advertising queue as resources recommended to the target object.

[0118] The technical solution of this embodiment is to obtain at least one first resource to be recommended to the target object, and determine the first resource fingerprint of each first resource in the at least one first resource; then obtain a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, and the resource fingerprint estimation group set includes the resource fingerprint estimation group corresponding to each second resource set, and the resource fingerprint estimation group includes a target resource fingerprint and a target estimated value; then, for each first resource, match the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group; and when the first resource fingerprint of the first resource successfully matches the target resource fingerprint of one of the resource fingerprint estimation groups, determine the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group; based on the estimated value of each first resource, recommend the target resource in at least one first resource to the target object, because the estimated value of the first resource is determined in a matching manner, then for the first resource, it is not necessary to use a complex algorithm to determine the corresponding estimated value, which can reduce the computing power resource consumption of calculating the estimated value, thereby improving the efficiency of resource recommendation.

[0119] See also Figure 4 , Figure 4 A flowchart of another resource recommendation method provided in an embodiment of the present application. Figure 4 The methods shown may include:

[0120] S310: Acquire at least one first resource to be recommended to a target object, and determine a first resource fingerprint of each first resource in the at least one first resource.

[0121] S320, obtaining a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, wherein the resource fingerprint estimation group set includes a resource fingerprint estimation group corresponding to each second resource set, and the resource fingerprint estimation group includes a target resource fingerprint and a target estimation value.

[0122] S330: For each first resource, match the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group.

[0123] S340: When the first resource fingerprint of the first resource successfully matches the target resource fingerprint of one of the resource fingerprint estimation groups, determine the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group.

[0124] S350: Recommend a target resource from at least one first resource to a target object based on the estimated value of each first resource.

[0125] S360: When the first resource fingerprint of the first resource fails to match the target resource fingerprint of any resource fingerprint estimation group, determine an estimated value of the first resource based on a preset estimated value determination algorithm.

[0126] In this embodiment, if the first resource fingerprint fails to match the target resource fingerprint of any resource fingerprint estimation group, it means that among the second resources whose estimated values ​​have been determined, there may not be a second resource with the same or similar design creativity as the first resource. At this time, it is necessary to determine the estimated value of the first resource based on a preset estimated value determination algorithm, so as to ensure the success rate of determining the estimated value of the resource.

[0127] The technical solution of this embodiment can ensure the success rate of determining the estimated value of the resource by determining the estimated value of the first resource based on a preset estimated value determination algorithm when the first resource fingerprint of the first resource fails to match the target resource fingerprint of any resource fingerprint estimation group.

[0128] Optionally, after determining the estimated value of the first resource based on a preset estimated value determination algorithm, the method further includes:

[0129] A new resource fingerprint estimation group is added to the resource fingerprint estimation group set based on the first resource fingerprint of the first resource and the estimation value of the first resource.

[0130] Specifically, the target resource fingerprint included in the new resource fingerprint estimation group may be the first resource fingerprint of the first resource, and the target estimated value included in the new resource fingerprint estimation group may be the estimated value of the first resource.

[0131] It can be understood that, by determining the estimated value of the first resource based on a preset estimated value determination algorithm, adding a new resource fingerprint estimation group to the resource fingerprint estimation group set based on the first resource fingerprint of the first resource and the estimated value of the first resource, the probability of successful matching can be increased, thereby further reducing computing power resources.

[0132] It should be noted that the second resource set may include one second resource or multiple second resources. Therefore, in the following embodiments, the case where the second resource set may include one second resource or multiple second resources is described respectively.

[0133] In a possible implementation, for a second resource set including a second resource, a resource fingerprint estimation group corresponding to the second resource set is determined in the following manner:

[0134] For a second resource set including a second resource, determine a second resource fingerprint of the second resource, and determine the second resource fingerprint of the second resource as a target resource fingerprint corresponding to the second resource set, and determine a second estimated value of the second resource, and determine the second estimated value of the second resource as a target estimated value corresponding to the second resource set.

[0135] The second estimated value represents the possibility that the second resource is selected by the target object. Optionally, the larger the second estimated value of the second resource is, the greater the possibility that the second resource is selected by the target object.

[0136] In this embodiment, exemplarily, assuming that the second resource set includes a second resource A, the second resource fingerprint of the second resource A is determined, and the second estimated value of the second resource A is determined, and the second resource fingerprint of the second resource A is used as the target resource fingerprint corresponding to the second resource set, and the second estimated value of the second resource A is used as the target estimated value corresponding to the second resource set.

[0137] In a possible implementation, for a second resource set including multiple second resources, a resource fingerprint estimation group corresponding to the second resource set is determined in the following manner:

[0138] Determine the second resource fingerprint of each second resource in the multiple second resources, and fuse the second resource fingerprints of each second resource in the multiple second resources to obtain the fused second resource fingerprint, and determine the fused second resource fingerprint as the target resource fingerprint corresponding to the second resource set, and determine the second estimated value of each second resource in the multiple second resources, and fuse the second estimated value of each second resource in the multiple second resources to obtain the fused second estimated value, and determine the fused second estimated value as the target estimated value corresponding to the second resource set, the resource fingerprint similarity between any two second resources in the multiple second resources is greater than the resource fingerprint similarity threshold, and the second resource fingerprint of the second resource is determined based on the multimedia material information of the second resource.

[0139] In this embodiment, illustratively, assuming that the second resource set includes a second resource A and a second resource B, the second resource fingerprint of the second resource A and the second resource fingerprint of the second resource B are determined, and the second estimated value of the second resource A and the second estimated value of the second resource B are determined. Then, the second resource fingerprint of the second resource A and the second resource fingerprint of the second resource B are merged to obtain a target estimated value corresponding to the second resource set, and the second estimated value of the second resource A and the second estimated value of the second resource B are merged to obtain a target estimated value corresponding to the second resource set.

[0140] In this embodiment, optionally, the second estimated values ​​of each second resource in the plurality of second resources are integrated by calculating the average value of the second resource fingerprint or selecting the median of the second resource fingerprint, which is not limited here. In addition, the second estimated values ​​of each second resource in the plurality of second resources are integrated by calculating the average value of the second estimated values ​​or selecting the median of the second estimated values, which is not limited here.

[0141] It should be noted that if the resource fingerprint similarity between any two second resources among the multiple second resources is greater than the resource fingerprint similarity threshold, it means that any two second resources are identical or similar, that is, multiple identical or similar second resources form a second resource set.

[0142] The following embodiments illustrate how to determine the resource fingerprint of a resource based on any of the above embodiments.

[0143] In a possible implementation, at least one of the first resource fingerprint of the first resource or the second resource fingerprint of the second resource is determined in the following manner:

[0144] Extracting at least one item of resource information of a fingerprint resource to be determined, where the fingerprint resource to be determined is at least one of the first resource or the second resource, and the at least one item of resource information includes multimedia material information of the fingerprint resource to be determined;

[0145] Determine the creative fingerprint corresponding to the multimedia material information;

[0146] Based on the creative fingerprint corresponding to the multimedia material information, a resource fingerprint of the fingerprint resource to be determined is obtained.

[0147] The multimedia material information may include at least one of text material, image material or video material (multiple image materials). The resource fingerprint may be used to indicate uniqueness. The creative fingerprint corresponding to the multimedia material information may be used to indicate the design creativity of the multimedia material information. Generally, the creative fingerprint is unique.

[0148] Optionally, if the multimedia material information includes text material and image material, the text fingerprint corresponding to the text material and the image fingerprint corresponding to the image material can be determined, and then the text fingerprint and the image fingerprint are added to obtain the creative fingerprint corresponding to the multimedia material information. If there are multiple image materials, the image sub-fingerprint corresponding to each of the multiple image materials can be determined respectively, and then the average value or weighted calculation is calculated based on the image sub-fingerprint corresponding to each image material to obtain the image fingerprint corresponding to the image material.

[0149] Among them, the fingerprint of this embodiment may include a hash value. For example, the resource fingerprint may include a hash value, and the creative fingerprint corresponding to the multimedia material information also includes a hash value. The hash value is calculated based on a hash function. A hash function (or hash algorithm, also known as a hash function, English: Hash Function) is a method of creating a small digital "fingerprint" from any kind of data. The hash function compresses a message or data into a summary, making the amount of data smaller and fixing the format of the data. The function shuffles and mixes the data to recreate a fingerprint called a hash value (hash values, hash codes, hashsums, or hashes). Hash values ​​are usually represented by a short string of random letters and numbers. Optionally, the hash value can be an MD5 value.

[0150] In this embodiment, the creative fingerprint corresponding to the multimedia material information may be used as the resource fingerprint of the fingerprint resource to be determined.

[0151] In another possible implementation, at least one item of information resource also includes landing page information corresponding to the fingerprint resource to be determined, and the landing page information indicates a landing page to be redirected when the fingerprint resource to be determined is selected. The method further includes:

[0152] Determine the landing page fingerprint corresponding to the landing page information;

[0153] Based on the creative fingerprint corresponding to the multimedia material information, a resource fingerprint of the fingerprint resource to be determined is obtained, including:

[0154] Determine respectively a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information;

[0155] A weighted calculation is performed based on the creative fingerprint corresponding to the multimedia material information, the first fusion weight, the landing page fingerprint corresponding to the landing page information, and the second fusion weight to obtain the resource fingerprint of the fingerprint resource to be determined.

[0156] Among them, the landing page is also called a guide page or a landing page. In this embodiment, the landing page can be a page of another application program to which the object jumps after clicking on the recommended resource, or a page of the current application program to which the object jumps, which is not limited here. Generally, the landing page fingerprint is unique. The landing page fingerprint of this embodiment can include a hash value.

[0157] Specifically, the product of the creative fingerprint corresponding to the multimedia material information and the first fusion weight can be calculated to obtain a first product result, and then the product of the landing page fingerprint corresponding to the landing page information and the second fusion weight can be calculated to obtain a second product result, and then the first product result and the second product result can be added together to obtain the resource fingerprint of the resource.

[0158] The technical solution of this embodiment also utilizes the landing page information of the resource to calculate the resource fingerprint of the resource, that is, it integrates other information to calculate the resource fingerprint of the resource, thereby improving the richness of the resource fingerprint calculation, and further improving the richness of resource fingerprint matching.

[0159] It should be noted that the first fusion weight and the second fusion weight may be predetermined.

[0160] In a possible implementation, respectively determining a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information includes:

[0161] Determining the material richness of the multimedia material information, the material type of the multimedia material information includes at least one of text or image, and the material richness is related to at least one of the number of material types of the multimedia material information, the length of the text, the number of images, or the number of image elements of the images;

[0162] Determine, based on the material richness, a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information;

[0163] Among them, the first fusion weight is positively correlated with the richness of the material, and the second fusion weight is negatively correlated with the richness of the material.

[0164] In this embodiment, specifically, if the multimedia material is relatively rich, the reason why the object selects the resource may be because it is attracted by the multimedia material. If the multimedia material is relatively simple, the reason why the object selects the resource may be the landing page. Therefore, the first fusion weight and the second fusion weight can be determined based on the richness of the material.

[0165] Optionally, the more material types there are, the higher the material richness is. The longer the text is, the higher the material richness is. The more images there are, the higher the material richness is. The more image elements there are in an image, the higher the material richness is.

[0166] The technical solution of this embodiment determines the material richness of the multimedia material information, the material type of the multimedia material information includes at least one of text or image, and the material richness is related to at least one of the number of material types of the multimedia material information, the length of the text, the number of images, or the number of image elements of the image; based on the material richness, the first fusion weight corresponding to the multimedia material information and the second fusion weight corresponding to the landing page information are determined respectively, which can improve the accuracy of determining the fusion weight, and then improve the accuracy of calculating the resource fingerprint, thereby improving the accuracy of resource fingerprint matching.

[0167] It can be understood that the first resource fingerprint of the first resource and the second resource fingerprint of the second resource can both be determined by the method of the above embodiment.

[0168] The following embodiments describe how to match a first resource fingerprint of a first resource with a target resource fingerprint of each resource fingerprint estimation group based on any of the above embodiments.

[0169] In a possible implementation, matching the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group includes:

[0170] Performing similarity matching processing on the first resource fingerprint of the first resource and the target resource fingerprint of each resource fingerprint estimation group to obtain the resource fingerprint similarity corresponding to each resource fingerprint estimation group;

[0171] For each resource fingerprint estimation group, when the resource fingerprint similarity corresponding to the resource fingerprint estimation group is greater than or equal to the resource fingerprint similarity threshold, it is determined that the resource fingerprint estimation group is successfully matched with the first resource; and when the resource fingerprint similarity corresponding to the resource fingerprint estimation group is less than the resource fingerprint similarity threshold, it is determined that the resource fingerprint estimation group fails to match the first resource.

[0172] Among them, the resource fingerprint similarity represents the similarity between resource fingerprints. Optionally, the greater the resource fingerprint similarity, the more similar the resource fingerprints are. In this embodiment, optionally, taking the resource fingerprint as a hash value as an example, the resource fingerprint similarity can be the reciprocal of the absolute value of the difference between the hash values, and the resource fingerprint similarity threshold can be the threshold corresponding to the reciprocal of the absolute value of the difference between the hash values. The larger the absolute value of the difference between the hash values, that is, the smaller the reciprocal of the absolute value of the difference, the less similar the resource fingerprints are; the smaller the absolute value of the difference between the hash values, that is, the larger the reciprocal of the absolute value of the difference, the more similar the resource fingerprints are.

[0173] Specifically, if the similarity between the first resource fingerprint of the first resource and the target resource fingerprint corresponding to one of the resource fingerprint estimation groups is greater than or equal to the resource fingerprint similarity threshold, it means that the first resource is relatively similar to the target resource fingerprint corresponding to one of the resource fingerprint estimation groups, that is, the first resource is relatively similar to the second recommended resource set in one of the second resource sets, and at this time, it can be determined that the resource fingerprint estimation group and the first resource are successfully matched. If the similarity between the first resource fingerprint of the first resource and the target resource fingerprint corresponding to any resource fingerprint estimation group is less than the resource fingerprint similarity threshold, it means that the first resource and the target resource fingerprint corresponding to any resource fingerprint estimation group are greatly different, that is, the first resource and the second recommended resource set in any second resource set are greatly different, and at this time, it can be determined that any resource fingerprint estimation group and the first resource have failed to match.

[0174] In a possible implementation, each resource fingerprint estimation group corresponds to a random number probability threshold, and the random number probability threshold corresponding to each resource fingerprint estimation group is determined by:

[0175] For each resource fingerprint estimation group, based on the second estimated value of the second resource in the second resource set corresponding to the resource fingerprint estimation group, determine the estimated value mean and estimated value standard deviation of the second resource set corresponding to the resource fingerprint estimation group;

[0176] For each resource fingerprint estimation group, based on the estimated value mean and estimated value standard deviation corresponding to the resource fingerprint estimation group, determine the coefficient of variation corresponding to the resource fingerprint estimation group;

[0177] Based on the coefficient of variation corresponding to each resource fingerprint estimation group, a random number probability threshold corresponding to each resource fingerprint estimation group is determined.

[0178] Among them, the coefficient of variation corresponding to the resource fingerprint estimation group is the ratio of the standard deviation of the estimation value corresponding to the resource fingerprint estimation group to the corresponding estimation value mean. Specifically, taking the resource as advertising resources as an example, the coefficient of variation is used instead of other indicators to measure the difference, such as standard deviation, variance, etc., mainly considering that there are natural category dimensions between advertisements. For example, the average click-through rate of advertisements in industry A is 10%, and its estimated value is basically about 10%. The average click-through rate of advertisements in industry B is 1%, and its estimated value is basically about 1%. If the standard deviation and variance are used, the indicators of these two industries will naturally have a 10-fold dimensional difference, and the use of the coefficient of variation can avoid the influence of the dimension. Optionally, in this embodiment, for each resource fingerprint estimation group, the smaller the corresponding coefficient of variation, the larger the corresponding random number probability threshold.

[0179] Accordingly, determining the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group includes:

[0180] Determine the random number probability based on the target estimated value of the successfully matched resource fingerprint estimation group;

[0181] When the random number probability is greater than the random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimated value of the first resource is determined based on the target estimated value of the successfully matched resource fingerprint estimation group.

[0182] In this embodiment, specifically, the smaller the coefficient of variation of the bucket is, the more likely it is that the preset estimated value recommendation algorithm has similar estimated values ​​for the same or similar resources, and the higher the random number probability threshold can be set.

[0183] Optionally, when the random number probability is less than or equal to a random number probability threshold corresponding to a successfully matched resource fingerprint estimation group, the estimated value of the first resource is determined based on a preset estimated value determination algorithm.

[0184] Specifically, if the random number probability is less than or equal to the random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimated value of the first resource determined based on the target estimated value of the successfully matched resource fingerprint estimation group may be inaccurate. Therefore, it is necessary to determine the estimated value of the first resource based on a preset estimated value determination algorithm to ensure the accuracy of the determination of the estimated value of the first resource.

[0185] The technical solution of this embodiment determines the random number probability based on the target estimated value of the successfully matched resource fingerprint estimation group; when the random number probability is greater than the random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimated value of the first resource is determined based on the target estimated value of the successfully matched resource fingerprint estimation group; when the random number probability is less than or equal to the random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimated value of the first resource is determined based on a preset estimated value determination algorithm, which can ensure the calculation accuracy of the estimated value of the first resource and thereby improve the accuracy of resource recommendation.

[0186] In a possible implementation, based on the coefficient of variation corresponding to each resource fingerprint estimation group, determining the random number probability threshold corresponding to each resource fingerprint estimation group includes:

[0187] Based on the coefficient of variation corresponding to each resource fingerprint estimation group, determine the maximum coefficient of variation and the minimum coefficient of variation;

[0188] Based on the maximum coefficient of variation and the minimum coefficient of variation, at least two coefficient of variation intervals are divided;

[0189] Determine the random number probability threshold corresponding to each coefficient of variation interval. For each coefficient of variation interval, the random number probability threshold corresponding to the coefficient of variation interval is negatively correlated with the interval maximum value corresponding to the coefficient of variation interval.

[0190] For each resource fingerprint estimation group, determine the variation coefficient interval to which the resource fingerprint estimation group belongs, and determine the random number probability threshold corresponding to the variation coefficient interval to which the resource fingerprint estimation group belongs as the random number probability threshold corresponding to the resource fingerprint estimation group.

[0191] In this embodiment, the random number probability threshold may be determined once every certain period of time. Optionally, the interval time may be related to the number of target resources that need to be recommended each time. Specifically, if the number of target resources that need to be recommended each time is larger, the computing power required to determine the target resources is larger, and the interval time for updating the random number probability threshold at this time may be larger, so as to avoid the problem of insufficient computing power caused by the random number probability threshold being updated too quickly, which in turn leads to the problem of slow recommendation efficiency of target resources. The smaller the number of target resources that need to be recommended each time, the smaller the computing power required to determine the target resources, and the smaller the interval time for updating the random number probability threshold at this time may be, so as to ensure the accuracy of the recommendation.

[0192] For example, if the probability threshold is updated once every hour, the calculation can be selected for the last 15 minutes (min). Assuming that there are N second resource sets in these 15 minutes, the N second resource sets correspond to N resource fingerprint estimation groups. In these N resource fingerprint estimation groups, the corresponding coefficients of variation are arranged from small to large, namely CV_1, CV_2, ..., CV_N. According to the business situation, it is divided into 4 buckets. At this time, the N / 4th, 2*(N / 4), and 3*(N / 4) are the bucket thresholds (less than CV_N / 4, greater than or equal to CV_N / 4 and less than CV_2N / 4, greater than or equal to CV_2N / 4 and less than CV_3N / 4, greater than or equal to CV_3N / 4), and each bucket corresponds to a random number probability threshold, which are r1, r2, r3, and r4 respectively. If the coefficient of variation corresponding to the resource fingerprint estimation group is less than CV_N / 4, the random number probability threshold corresponding to the resource fingerprint estimation group is r1; if the coefficient of variation corresponding to the resource fingerprint estimation group is greater than or equal to CV_N / 4 and less than CV_2N / 4, the random number probability threshold corresponding to the resource fingerprint estimation group is r2; if the coefficient of variation corresponding to the resource fingerprint estimation group is greater than or equal to CV_2N / 4 and less than CV_3N / 4, the random number probability threshold corresponding to the resource fingerprint estimation group is r3; if the coefficient of variation corresponding to the resource fingerprint estimation group is greater than or equal to CV_3N / 4, the random number probability threshold corresponding to the resource fingerprint estimation group is r4.

[0193] The following embodiments further illustrate how to determine the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group based on any of the above embodiments.

[0194] In a possible implementation, determining the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group includes:

[0195] Determine the similarity difference between the resource fingerprint similarity corresponding to the successfully matched resource fingerprint estimation group and the resource fingerprint similarity threshold;

[0196] When the similarity difference is less than the similarity difference threshold, the target estimated value of the successfully matched resource fingerprint estimation group is determined as the estimated value of the first resource;

[0197] When the similarity difference is greater than or equal to the similarity difference threshold, the target estimated value of the successfully matched resource fingerprint estimation group is adjusted, and the adjusted target estimated value is determined as the estimated value of the first resource.

[0198] In this embodiment, specifically, if the similarity difference is less than the similarity difference threshold, it means that the first resource is particularly close to the second resource in the second resource set, and at this time, the target estimated value of the successfully matched resource fingerprint estimation group can be directly determined as the estimated value of the first resource. If the similarity difference is greater than or equal to the similarity difference threshold, it means that the first resource is not particularly close to the second resource in the second resource set, and at this time, the target estimated value of the successfully matched resource fingerprint estimation group can be adjusted, and the adjusted target estimated value can be determined as the estimated value of the first resource.

[0199] Optionally, adjusting the target estimated value may be to increase the target estimated value or to decrease the target estimated value, which is determined based on actual conditions and is not limited here.

[0200] The technical solution of this embodiment is to determine the target estimated value of the successfully matched resource fingerprint estimation group as the estimated value of the first resource when the similarity difference is less than the similarity difference threshold; and adjust the target estimated value of the successfully matched resource fingerprint estimation group when the similarity difference is greater than or equal to the similarity difference threshold, and determine the adjusted target estimated value as the estimated value of the first resource. This can ensure the calculation accuracy of the estimated value of the first resource as much as possible, thereby ensuring the accuracy of resource recommendations.

[0201] In a possible implementation, adjusting the target estimated value of the successfully matched resource fingerprint estimation group includes:

[0202] Acquire a target object feature of the target object, and acquire a differential resource feature in the first resource, where the differential resource feature is different from a second resource feature of any second resource in the second resource set corresponding to the successfully matched resource fingerprint estimation group;

[0203] Obtain a trained deep neural network model, and use the target object features and the difference resource features as inputs of the deep neural network model to obtain a first output estimate value of the deep neural network model output, wherein the deep neural network model is trained by using the training object features and the training resource features in the training data as inputs and the training estimate value in the training data as outputs of the deep neural network model;

[0204] When the first output estimated value is greater than the estimated value threshold, increasing the target estimated value of the successfully matched resource fingerprint estimated group;

[0205] When the first output estimated value is less than or equal to the estimated value threshold, the target estimated value of the successfully matched resource fingerprint estimation group is reduced.

[0206] Among them, the deep neural network (DNN) is a multi-layer unsupervised neural network, and the output features of the previous layer are used as the input of the next layer for feature learning. After layer-by-layer feature mapping, the features of the existing space samples are mapped to another feature space. Among them, the resource feature can represent the attribute of the resource, and the attribute of the resource can include at least one of the category or price of the resource.

[0207] In this embodiment, the difference resource feature may be at least one of the difference in multimedia material layout, the difference in text material, or the difference in image material. The difference in text material may be at least one of the content of the text, the font of the text, or the font color of the text. The difference in image material may be at least one of the content of the image or the color of the image.

[0208] See also Figure 5 , Figure 5 A schematic diagram of a difference resource feature provided in an embodiment of the present application. Figure 5 As shown, the difference in image material between the first recommended resource and the second recommended resource can be determined as the difference resource feature.

[0209] Specifically, Figure 5 As shown, the image actions between the first recommended resource and the second recommended resource are the same, so the image action of the first recommended resource can be used as the input of the DNN.

[0210] The estimated value threshold may be a threshold used to determine the target object's preference. Specifically, if the first output estimated value is greater than the estimated value threshold, it means that the target object is relatively fond of the part of the differential resource features, so the target estimated value of the successfully matched resource fingerprint estimated group may be increased. If the first output estimated value is less than or equal to the estimated value threshold, it means that the target object is not particularly fond of the part of the differential resource features, so the target estimated value of the successfully matched resource fingerprint estimated group may be reduced.

[0211] The technical solution of this embodiment obtains the target object features of the target object and the differential resource features in the first resource, obtains a trained deep neural network model, and uses the target object features and the differential resource features as inputs of the deep neural network model to obtain a first output estimate value output by the deep neural network model, wherein the deep neural network model increases the target estimate value of the successfully matched resource fingerprint estimate group in the training data when the first output estimate value is greater than the estimate value threshold; and reduces the target estimate value of the successfully matched resource fingerprint estimate group when the first output estimate value is less than or equal to the estimate value threshold. Since the differential resource features are generally partial resource features in the first resource, the computing power required for model prediction can be reduced.

[0212] In a possible implementation, the method further includes:

[0213] Obtaining a determined number of a second estimated value determined based on the estimated value determination algorithm, and determining a quantity difference between a target number and the determined number, wherein the target number is less than or equal to a maximum determined number, and the maximum determined number is: when the second estimated value of the second resource is determined based on the estimated value determination algorithm, the maximum number of the second estimated value determined that is supported by the computing power resources of the electronic device;

[0214] Obtaining a third resource with a quantity difference, where the third resource is used to explore the preferences of the target object;

[0215] Determine an estimated value of each third resource in the number difference third resources based on a preset estimated value determination algorithm, where the estimated value of the third resource represents the possibility that the third resource is selected by the target object;

[0216] Based on the estimated value of each third resource in the quantity difference third resources, a target resource in the quantity difference third resources is recommended to the target object.

[0217] The computing power resource may include but is not limited to at least one of the processing frequency of a central processing unit and the number of processing cores of the central processing unit.

[0218] In this embodiment, specifically, since the estimated value of the first resource can be determined based on a matching method, the determined number of the second estimated value determined based on the estimated value determination algorithm can be less than the target number supported by the computing power resources of the electronic device, and the saved computing power resources can be used to calculate the estimated value of the exploration resource (third resource), and then based on the estimated value of each third resource in the number difference third resources, the target resource in the number difference third resources is recommended to the target object, and the saved computing power resources can be used for calculation of newly created advertisements or other advertisements that need to be explored, thereby improving the exploration success rate.

[0219] The second estimated value of the second resource is determined based on a preset estimated value determination algorithm, the estimated value determination algorithm including:

[0220] Acquire a target object feature of a target object, and acquire a resource feature of a resource with an estimated value to be determined, where the resource with an estimated value to be determined is at least one of a first resource, a second resource, or a third resource;

[0221] Obtain a trained deep neural network model, and use the target object characteristics and the resource characteristics of the estimated value resource to be determined as inputs of the deep neural network model to obtain a second output estimated value output by the deep neural network model, wherein the deep neural network model is trained by using the training object characteristics and the training resource characteristics in the training data as inputs and using the training estimated value in the training data as outputs of the deep neural network model;

[0222] The second output estimated value is determined as the estimated value of the estimated value resource to be determined.

[0223] See also Figure 6 , Figure 6 A schematic diagram of calculating an estimated value of a resource based on a preset estimated value determination algorithm provided in an embodiment of the present application.

[0224] like Figure 6 As shown, the target object identifier of the target object and the resource identifier of the resource can be obtained first, and then the target object feature corresponding to the target object identifier is searched based on the target object identifier, and the corresponding resource feature is searched based on the resource identifier of the resource, and then the DNN network is used to determine the estimated value.

[0225] It should be noted that each time the estimated value is calculated based on the preset estimated value determination algorithm, generally including the following: Figure 6 There are several processes, namely querying object features (one query, multiple uses), querying ad features (each ad must be queried), and then taking object features and ad features as input, passing through the DNN network, to get the final estimated value (score).

[0226] It is understandable that if it is necessary to calculate the estimated value of the resource based on a preset estimated value determination algorithm, it can be achieved through the method of this embodiment.

[0227] For ease of understanding, the following embodiments are based on any of the above embodiments and are described by taking the resource being an advertisement as an example.

[0228] After research, it was found that, in fact, for each recommendation request, the ads in the candidate ad queue recalled from the ad library will have more or less duplicate or highly similar materials. The main reasons include that different advertisers may create the same material; when recalling, it is necessary to ensure diversity in multiple dimensions, including the advertiser dimension. If the candidate ads cannot have duplicate materials, then the diversity of advertisers may be affected, thereby affecting the bidding ecology, so the existence of ads with duplicate materials in the candidate ad queue is an objective fact in the current ad recommendation system. The behavior of the object for ads with the same material is basically the same, so this solution is designed to optimize the computational amount of ads with the same or similar materials from the perspective of algorithm stability, so as to achieve the purpose of saving advertising computing power.

[0229] Therefore, based on the algorithm stability, we optimize the computing power of ads with the same or similar materials, so as to save the computing power of ads. Figure 7 and Figure 8 . Figure 7 A schematic diagram of a flow chart of estimating a computing resource provided in an embodiment of the present application. Figure 8 A flowchart of another resource recommendation method provided in an embodiment of the present application.

[0230] The specific steps are as follows:

[0231] Step 1: Extract resource fingerprints from the ads that are filtered from the first candidate ad queue to the second candidate ad queue. The resource fingerprints are designed based on the business scenario and the system process position of the algorithm. For example, the resource fingerprint can be the md5 value of the material image (video), or it can be the fusion md5 of the material image (video) md5 and the landing page md5.

[0232] Step 2: For the candidate advertisement sets with the same or similar fingerprints, calculate the estimated value deviation based on the estimated value of the algorithm. The coefficient of variation (CV) is used to measure the estimated deviation. σ is the standard deviation of the estimated value of the same or similar fingerprint advertisement set, and μ is the mean of the estimated value of the same or similar fingerprint advertisement set. The coefficient of variation is used instead of other indicators to measure the difference, such as standard deviation, variance, etc., mainly because there are natural category dimensions in the advertisement recommendation system. For example, the average click-through rate of advertisements in industry A is 10%, and its estimated value is basically around 10%. The average click-through rate of advertisements in industry B is 1%, and its estimated value is basically around 1%. If standard deviation and variance are used, the indicators of these two industries will naturally have a 10-fold dimension difference. Using the coefficient of variation can avoid the influence of dimension, which provides theoretical support for the aggregation in step three.

[0233] Step 3: According to the update frequency of the advertising recommendation algorithm, calculate the stability index of the algorithm in the recent period, and sort and bucket it. Each bucket corresponds to a computing power optimization probability (random number probability threshold). For example, if the algorithm is updated once an hour, you can choose to calculate the last 15 minutes. Assume that there are N fingerprint sets in this 15 minutes, and the coefficient of variation is arranged from small to large as CV_1, CV_2, ..., CV_N. According to the business situation, it is divided into 4 buckets. At this time, the N / 4th, 2*(N / 4), and 3*(N / 4) are the bucket thresholds (less than CV_N / 4, greater than or equal to CV_N / 4 and less than CV_2N / 4), greater than or equal to CV_2N / 4 and less than CV_3N / 4), greater than or equal to CV_3N / 4). Each bucket corresponds to a computing power optimization probability value r1, r2, r3, and r4. The smaller the coefficient of variation of the bucket, the more similar the estimated value of the advertising recommendation algorithm for the same material advertisement is, and the higher the computing power optimization probability can be set.

[0234] Step 4: When an object ad display request is triggered, a certain number of candidate ads will enter the ad recommendation system. The ad recommendation system needs to calculate an estimated value for each candidate ad for the object. At this time, according to the definition of the same or similar materials in step 1, such as the same or similar resource fingerprints, a Set set (resource fingerprint estimation group set) with resource fingerprints as the key is constructed to store the estimated value of the ad algorithm corresponding to the fingerprint for the current request. Generally, the keys in the Set set will not be repeated. The ad algorithm traverses the ads in the second candidate ad queue. Before calculating each candidate ad, it first queries whether there is a value in the Set set through its resource fingerprint. If there is a value, it determines whether to use the value in the set according to the optimization computing power probability. If so, there is no need to calculate, and the corresponding estimated value in the Set set is directly assigned. If the value in the set is not used according to the optimization computing power probability, it is calculated in the normal estimation method, and the Set set is updated based on the calculated estimated value; if there is no value, it is also calculated in the normal estimation method, and the calculated result and the corresponding fingerprint key are stored in the Set set.

[0235] Among them, key can represent resource fingerprint, and score can represent estimated value.

[0236] Step 5: Pass the number of uncalculated ads to the discovery ad library, and add a number of discovery ads to the second candidate ad queue, not exceeding the number of uncalculated ads, to form a final candidate ad queue with the calculated ads. The resources in the final candidate ad queue will eventually be recommended to the mobile terminal used by the target object.

[0237] Since the calculation of the algorithm is relatively stable, the calculated estimated values ​​of ads with the same material in the candidate ad queue are almost the same, that is, the calculated ad ranking positions are also very close, and the objects behave the same way for ads with the same material, so the calculation of ads with the same material seems redundant. This solution introduces algorithm stability as a reference standard. Algorithms with high stability are more likely to use the same fingerprint to replace the existing estimated values ​​for computing power optimization, thereby not losing the recommendation effect and reducing the amount of calculation for the same ad. In addition, the computing power resources saved by ads with the same fingerprint can be used to calculate newly created ads or other ads that need to be explored, thereby improving the exploration success rate.

[0238] Specifically, the stability dimension may not be limited to creative fingerprints. For industry recommendation algorithms, aggregation can be performed based on industry categories. For e-commerce advertising recommendation algorithms, SPU (the smallest unit of aggregation) or products can be used for aggregation to achieve similar effects.

[0239] See also Fig. 9 , Fig. 9 This is a schematic diagram of the structure of a resource recommendation device provided in an embodiment of the present application. Fig. 9 The resource recommendation device 900 shown includes a first acquisition module 910, a second acquisition module 920, a matching module 930, an estimated value determination module 940 and a resource recommendation module 950, wherein:

[0240] A first acquisition module 910 is used to acquire at least one first resource to be recommended to a target object, and determine a first resource fingerprint of each first resource in the at least one first resource, where the first resource fingerprint is determined based on multimedia material information of the first resource;

[0241] A second acquisition module 920 is used to acquire a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, the resource fingerprint estimation group set includes a resource fingerprint estimation group corresponding to each second resource set, the resource fingerprint estimation group includes a target resource fingerprint and a target estimation value, the target resource fingerprint corresponding to the second resource set is determined based on the multimedia material information of the second resource in the second resource set, and the target estimation value corresponding to the second resource set represents the possibility that the second resource in the second resource set is selected by the target object;

[0242] A matching module 930, configured to match, for each first resource, a first resource fingerprint of the first resource with a target resource fingerprint of each resource fingerprint estimation group;

[0243] An estimated value determination module 940 is used to determine an estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group when a first resource fingerprint of the first resource successfully matches a target resource fingerprint of one of the resource fingerprint estimation groups, wherein the estimated value of the first resource represents the possibility that the first resource is selected by the target object;

[0244] The resource recommendation module 950 is configured to recommend a target resource from at least one first resource to a target object based on the estimated value of each first resource.

[0245] Optionally, the estimated value determination module 940 is further configured to:

[0246] In the case that the first resource fingerprint of the first resource fails to match the target resource fingerprint of any resource fingerprint estimation group, the estimated value of the first resource is determined based on a preset estimated value determination algorithm.

[0247] Optionally, a second resource set includes one or more second resources, and the resource fingerprint estimation group corresponding to the second resource set is determined by the second acquisition module 920 in the following manner:

[0248] For a second resource set including a second resource, determining a second resource fingerprint of the second resource, and determining the second resource fingerprint of the second resource as a target resource fingerprint corresponding to the second resource set, and determining a second estimated value of the second resource, and determining the second estimated value of the second resource as a target estimated value corresponding to the second resource set, the second estimated value representing the possibility of the second resource being selected by the target object;

[0249] For a second resource set including multiple second resources, determine the second resource fingerprint of each second resource in the multiple second resources, and fuse the second resource fingerprints of each second resource in the multiple second resources to obtain a fused second resource fingerprint, and determine the fused second resource fingerprint as a target resource fingerprint corresponding to the second resource set, and determine the second estimated value of each second resource in the multiple second resources, and fuse the second estimated value of each second resource in the multiple second resources to obtain a fused second estimated value, and determine the fused second estimated value as a target estimated value corresponding to the second resource set, the resource fingerprint similarity between any two second resources in the multiple second resources is greater than a resource fingerprint similarity threshold, and the second resource fingerprint of the second resource is determined based on the multimedia material information of the second resource.

[0250] Optionally, at least one of the first resource fingerprint of the first resource or the second resource fingerprint of the second resource is determined by the first acquisition module 910 in the following manner:

[0251] Extracting at least one item of resource information of a fingerprint resource to be determined, where the fingerprint resource to be determined is at least one of the first resource or the second resource, and the at least one item of resource information includes multimedia material information of the fingerprint resource to be determined;

[0252] Determine the creative fingerprint corresponding to the multimedia material information;

[0253] Based on the creative fingerprint corresponding to the multimedia material information, a resource fingerprint of the fingerprint resource to be determined is obtained.

[0254] Optionally, at least one item of information resource information also has to determine landing page information corresponding to the fingerprint resource, and the landing page information indicates the landing page to be jumped when the fingerprint resource to be determined is selected. The first acquisition module 910 is further used to:

[0255] Determine the landing page fingerprint corresponding to the landing page information;

[0256] When the first acquisition module 910 obtains the resource fingerprint of the fingerprint resource to be determined based on the creative fingerprint corresponding to the multimedia material information, it can be used to:

[0257] Determine respectively a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information;

[0258] A weighted calculation is performed based on the creative fingerprint corresponding to the multimedia material information, the first fusion weight, the landing page fingerprint corresponding to the landing page information, and the second fusion weight to obtain a resource fingerprint of the fingerprint resource to be determined, where the resource fingerprint includes a hash value.

[0259] Optionally, when respectively determining the first fusion weight corresponding to the multimedia material information and the second fusion weight corresponding to the landing page information, the first acquisition module 910 may be used to:

[0260] Determining the material richness of the multimedia material information, the material type of the multimedia material information includes at least one of text or image, and the material richness is related to at least one of the number of material types of the multimedia material information, the length of the text, the number of images, or the number of image elements of the images;

[0261] Determine, based on the material richness, a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information;

[0262] Among them, the first fusion weight is positively correlated with the richness of the material, and the second fusion weight is negatively correlated with the richness of the material.

[0263] Optionally, when matching the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group, the matching module 930 may be used to:

[0264] Performing similarity matching processing on the first resource fingerprint of the first resource and the target resource fingerprint of each resource fingerprint estimation group to obtain the resource fingerprint similarity corresponding to each resource fingerprint estimation group;

[0265] For each resource fingerprint estimation group, when the resource fingerprint similarity corresponding to the resource fingerprint estimation group is greater than or equal to the resource fingerprint similarity threshold, it is determined that the resource fingerprint estimation group is successfully matched with the first resource; and when the resource fingerprint similarity corresponding to the resource fingerprint estimation group is less than the resource fingerprint similarity threshold, it is determined that the resource fingerprint estimation group fails to match the first resource.

[0266] Optionally, each resource fingerprint estimation group corresponds to a random number probability threshold, and the random number probability threshold corresponding to each resource fingerprint estimation group is determined by the matching module 930 in the following manner:

[0267] For each resource fingerprint estimation group, based on the second estimated value of the second resource in the second resource set corresponding to the resource fingerprint estimation group, determine the estimated value mean and estimated value standard deviation of the second resource set corresponding to the resource fingerprint estimation group;

[0268] For each resource fingerprint estimation group, based on the estimated value mean and estimated value standard deviation corresponding to the resource fingerprint estimation group, determine the coefficient of variation corresponding to the resource fingerprint estimation group;

[0269] Based on the coefficient of variation corresponding to each resource fingerprint estimation group, determine the random number probability threshold corresponding to each resource fingerprint estimation group;

[0270] When determining the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group, the estimated value determination module 940 may be used to:

[0271] Determine the random number probability based on the target estimated value of the successfully matched resource fingerprint estimation group;

[0272] When the random number probability is greater than the random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimated value of the first resource is determined based on the target estimated value of the successfully matched resource fingerprint estimation group.

[0273] Optionally, the estimated value determination module 940 is further configured to:

[0274] When the random number probability is less than or equal to the random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimation value of the first resource is determined based on a preset estimation value determination algorithm.

[0275] Optionally, when determining the random number probability threshold corresponding to each resource fingerprint estimation group based on the coefficient of variation corresponding to each resource fingerprint estimation group, the matching module 930 may be used to:

[0276] Based on the coefficient of variation corresponding to each resource fingerprint estimation group, determine the maximum coefficient of variation and the minimum coefficient of variation;

[0277] Based on the maximum coefficient of variation and the minimum coefficient of variation, at least two coefficient of variation intervals are divided;

[0278] Determine the random number probability threshold corresponding to each coefficient of variation interval. For each coefficient of variation interval, the random number probability threshold corresponding to the coefficient of variation interval is negatively correlated with the interval maximum value corresponding to the coefficient of variation interval.

[0279] For each resource fingerprint estimation group, determine the variation coefficient interval to which the resource fingerprint estimation group belongs, and determine the random number probability threshold corresponding to the variation coefficient interval to which the resource fingerprint estimation group belongs as the random number probability threshold corresponding to the resource fingerprint estimation group.

[0280] Optionally, when determining the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group, the estimated value determination module 940 may be used to:

[0281] Determine the similarity difference between the resource fingerprint similarity corresponding to the successfully matched resource fingerprint estimation group and the resource fingerprint similarity threshold;

[0282] When the similarity difference is less than the similarity difference threshold, the target estimated value of the successfully matched resource fingerprint estimation group is determined as the estimated value of the first resource;

[0283] When the similarity difference is greater than or equal to the similarity difference threshold, the target estimated value of the successfully matched resource fingerprint estimation group is adjusted, and the adjusted target estimated value is determined as the estimated value of the first resource.

[0284] Optionally, when adjusting the target estimated value of the successfully matched resource fingerprint estimation group, the estimated value determination module 940 may be used to:

[0285] Acquire a target object feature of the target object, and acquire a differential resource feature in the first resource, where the differential resource feature is different from a second resource feature of any second resource in the second resource set corresponding to the successfully matched resource fingerprint estimation group;

[0286] Obtain a trained deep neural network model, and use the target object features and the difference resource features as inputs of the deep neural network model to obtain a first output estimate value of the deep neural network model output, wherein the deep neural network model is trained by using the training object features and the training resource features in the training data as inputs and the training estimate value in the training data as outputs of the deep neural network model;

[0287] When the first output estimated value is greater than the estimated value threshold, increasing the target estimated value of the successfully matched resource fingerprint estimated group;

[0288] When the first output estimated value is less than or equal to the estimated value threshold, the target estimated value of the successfully matched resource fingerprint estimation group is reduced.

[0289] Optionally, the estimated value determination module 940 is further configured to:

[0290] Obtaining a determined number of a second estimated value determined based on the estimated value determination algorithm, and determining a quantity difference between a target number and the determined number, wherein the target number is less than or equal to a maximum determined number, and the maximum determined number is: when the second estimated value of the second resource is determined based on the estimated value determination algorithm, the maximum number of the second estimated value determined that is supported by the computing power resources of the electronic device;

[0291] Obtaining a third resource with a quantity difference, where the third resource is used to explore the preferences of the target object;

[0292] Determine an estimated value of each third resource in the number difference third resources based on a preset estimated value determination algorithm, where the estimated value of the third resource represents the possibility that the third resource is selected by the target object;

[0293] The resource recommendation module 950 is further configured to recommend a target resource among the quantity difference third resources to the target object based on the estimated value of each third resource among the quantity difference third resources.

[0294] The second estimated value of the second resource is determined by the estimated value determination module 940 based on a preset estimated value determination algorithm. When executing the estimated value determination algorithm, the estimated value determination module 940 may be used to:

[0295] Acquire a target object feature of a target object, and acquire a resource feature of a resource with an estimated value to be determined, where the resource with an estimated value to be determined is at least one of a first resource, a second resource, or a third resource;

[0296] Obtain a trained deep neural network model, and use the target object characteristics and the resource characteristics of the estimated value resource to be determined as inputs of the deep neural network model to obtain a second output estimated value output by the deep neural network model, wherein the deep neural network model is trained by using the training object characteristics and the training resource characteristics in the training data as inputs and using the training estimated value in the training data as outputs of the deep neural network model;

[0297] The second output estimated value is determined as the estimated value of the estimated value resource to be determined.

[0298] The device of the embodiments of the present application can execute the method provided by the embodiments of the present application, and the implementation principles are similar. The actions performed by each module in the device of each embodiment of the present application correspond to the steps in the method of each embodiment of the present application. For the detailed functional description of each module of the device, please refer to the description in the corresponding method shown in the previous text, which will not be repeated here.

[0299] An embodiment of the present application provides an electronic device, including a memory, a processor, and a computer program stored in the memory, and the processor executes the above computer program to implement the steps of the method of any embodiment of the present application.

[0300] In an alternative embodiment, an electronic device is provided, such as Fig.10 As shown, Fig.10 The electronic device 1000 shown includes: a processor 1001 and a memory 1003. The processor 1001 and the memory 1003 are connected, such as through a bus 1002. Optionally, the electronic device 1000 may also include a transceiver 1004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 1004 is not limited to one, and the structure of the electronic device 1000 does not constitute a limitation on the embodiments of the present application.

[0301] Processor 1001 may be a CPU (Central Processing Unit), a general purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 1001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0302] The bus 1002 may include a path to transmit information between the above components. The bus 1002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 1002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Fig.10 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0303] The memory 1003 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), magnetic disk storage media, other magnetic storage devices, or any other medium that can be used to carry or store computer programs and can be read by a computer, without limitation herein.

[0304] The memory 1003 is used to store the computer program for executing the embodiment of the present application, and the execution is controlled by the processor 1001. The processor 1001 is used to execute the computer program stored in the memory 1003 to implement the steps shown in the above method embodiment.

[0305] An embodiment of the present application provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps and corresponding contents of the aforementioned method embodiment can be implemented.

[0306] The embodiment of the present application also provides a computer program product, including a computer program, which can implement the steps and corresponding contents of the aforementioned method embodiment when executed by a processor.

[0307] It should be understood that, although each operation step is represented by arrows in the flowchart of the embodiment of the present application, the implementation order of these steps is not limited to the order represented by the arrows. Unless there is a clear description in this article, in some implementation scenarios of the embodiment of the present application, the implementation steps in each flowchart can be performed in other orders according to demand. In addition, some or all of the steps in each flowchart can include multiple sub-steps or multiple stages based on the actual implementation scenario. Some or all of these sub-steps or stages can be executed at the same time, and each sub-step or stage in these sub-steps or stages can also be executed at different times respectively. In the scenario where the execution time is different, the execution order of these sub-steps or stages can be flexibly configured according to demand, and the embodiment of the present application does not limit this.

[0308] The above is only an optional implementation method for some implementation scenarios of the present application. It should be pointed out that for ordinary technicians in this technical field, without departing from the technical concept of the solution of the present application, other similar implementation methods based on the technical ideas of the present application are also within the protection scope of the embodiments of the present application.

Claims

1. A resource recommendation method, characterized in that: include: Acquire at least one first resource to be recommended to the target object, and determine a first resource fingerprint of each first resource in the at least one first resource, where the first resource fingerprint is determined based on multimedia material information of the first resource; Acquire a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, the resource fingerprint estimation group set includes a resource fingerprint estimation group corresponding to each second resource set, the resource fingerprint estimation group includes a target resource fingerprint and a target estimation value, the target resource fingerprint corresponding to the second resource set is determined based on multimedia material information of a second resource in the second resource set, and the target estimation value corresponding to the second resource set represents the possibility that the second resource in the second resource set is selected by the target object; For each first resource, matching a first resource fingerprint of the first resource with a target resource fingerprint of each resource fingerprint estimation group; In the case where the first resource fingerprint of the first resource successfully matches the target resource fingerprint of one of the resource fingerprint estimation groups, determining an estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group, the estimated value of the first resource representing the possibility of the first resource being selected by the target object; Based on the estimated value of each first resource, a target resource from among the at least one first resource is recommended to the target object.

2. The method according to claim 1, characterized in that The method further comprises: In a case where the first resource fingerprint of the first resource fails to match the target resource fingerprint of any resource fingerprint estimation group, an estimated value of the first resource is determined based on a preset estimated value determination algorithm.

3. The method according to claim 1, characterized in that A second resource set includes one or more second resources, and a resource fingerprint estimation group corresponding to the second resource set is determined in the following manner: For a second resource set including a second resource, determining a second resource fingerprint of the second resource, and determining the second resource fingerprint of the second resource as a target resource fingerprint corresponding to the second resource set, and determining a second estimated value of the second resource, and determining the second estimated value of the second resource as a target estimated value corresponding to the second resource set, wherein the second estimated value represents a possibility that the second resource is selected by the target object; For a second resource set including multiple second resources, determine the second resource fingerprint of each second resource in the multiple second resources, merge the second resource fingerprints of each second resource in the multiple second resources to obtain a merged second resource fingerprint, and determine the merged second resource fingerprint as a target resource fingerprint corresponding to the second resource set, determine the second estimated value of each second resource in the multiple second resources, merge the second estimated value of each second resource in the multiple second resources to obtain a merged second estimated value, and determine the merged second estimated value as a target estimated value corresponding to the second resource set, the resource fingerprint similarity between any two second resources in the multiple second resources is greater than a resource fingerprint similarity threshold, and the second resource fingerprint of the second resource is determined based on the multimedia material information of the second resource.

4. The method according to claim 1, characterized in that: At least one of the first resource fingerprint of the first resource or the second resource fingerprint of the second resource is determined by: Extracting at least one item of resource information of a fingerprint resource to be determined, where the fingerprint resource to be determined is at least one of the first resource or the second resource, and the at least one item of resource information includes multimedia material information of the fingerprint resource to be determined; Determining a creative fingerprint corresponding to the multimedia material information; Based on the creative fingerprint corresponding to the multimedia material information, a resource fingerprint of the fingerprint resource to be determined is obtained.

5. The method according to claim 4, characterized in that The at least one information resource also includes landing page information corresponding to the fingerprint resource to be determined, the landing page information indicating a landing page to be redirected when the fingerprint resource to be determined is selected, and the method further includes: Determine a landing page fingerprint corresponding to the landing page information; The obtaining the resource fingerprint of the fingerprint resource to be determined based on the creative fingerprint corresponding to the multimedia material information includes: Respectively determining a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information; A weighted calculation is performed based on the creative fingerprint corresponding to the multimedia material information, the first fusion weight, the landing page fingerprint corresponding to the landing page information, and the second fusion weight to obtain a resource fingerprint of the fingerprint resource to be determined, wherein the resource fingerprint includes a hash value.

6. The method according to claim 5, characterized in that The respectively determining a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information includes: Determining material richness of the multimedia material information, the material type of the multimedia material information comprising at least one of text or image, the material richness being related to at least one of the number of material types of the multimedia material information, the length of the text, the number of the images, or the number of image elements of the images; Determine, based on the material richness, a first fusion weight corresponding to the multimedia material information and a second fusion weight corresponding to the landing page information; Among them, the first fusion weight is positively correlated with the material richness, and the second fusion weight is negatively correlated with the material richness.

7. The method according to claim 1, characterized in that The matching of the first resource fingerprint of the first resource with the target resource fingerprint of each resource fingerprint estimation group includes: Performing similarity matching processing on the first resource fingerprint of the first resource and the target resource fingerprint of each resource fingerprint estimation group to obtain the resource fingerprint similarity corresponding to each resource fingerprint estimation group; For each resource fingerprint estimation group, when the resource fingerprint similarity corresponding to the resource fingerprint estimation group is greater than or equal to the resource fingerprint similarity threshold, it is determined that the resource fingerprint estimation group is successfully matched with the first resource; and when the resource fingerprint similarity corresponding to the resource fingerprint estimation group is less than the resource fingerprint similarity threshold, it is determined that the resource fingerprint estimation group fails to match with the first resource.

8. The method according to claim 7, characterized in that Each resource fingerprint estimation group corresponds to a random number probability threshold, which is determined by the following method: For each resource fingerprint estimation group, based on the second estimated value of the second resource in the second resource set corresponding to the resource fingerprint estimation group, determine the estimated value mean and estimated value standard deviation of the second resource set corresponding to the resource fingerprint estimation group; For each resource fingerprint estimation group, determining a coefficient of variation corresponding to the resource fingerprint estimation group based on an estimation value mean and an estimation value standard deviation corresponding to the resource fingerprint estimation group; Based on the coefficient of variation corresponding to each resource fingerprint estimation group, determine the random number probability threshold corresponding to each resource fingerprint estimation group; The determining the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group includes: Determine the random number probability based on the target estimated value of the successfully matched resource fingerprint estimation group; When the random number probability is greater than a random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimated value of the first resource is determined based on the target estimated value of the successfully matched resource fingerprint estimation group.

9. The method according to claim 8, characterized in that The method further comprises: When the random number probability is less than or equal to the random number probability threshold corresponding to the successfully matched resource fingerprint estimation group, the estimation value of the first resource is determined based on a preset estimation value determination algorithm.

10. The method according to claim 8, characterized in that The determining of the random number probability threshold corresponding to each resource fingerprint estimation group based on the coefficient of variation corresponding to each resource fingerprint estimation group includes: Based on the coefficient of variation corresponding to each resource fingerprint estimation group, determine the maximum coefficient of variation and the minimum coefficient of variation; Based on the maximum coefficient of variation and the minimum coefficient of variation, dividing at least two coefficient of variation intervals; Determine a random number probability threshold corresponding to each coefficient of variation interval, wherein for each coefficient of variation interval, the random number probability threshold corresponding to the coefficient of variation interval is negatively correlated with the interval maximum value corresponding to the coefficient of variation interval; For each resource fingerprint estimation group, determine the variation coefficient interval to which the resource fingerprint estimation group belongs, and determine the random number probability threshold corresponding to the variation coefficient interval to which the resource fingerprint estimation group belongs as the random number probability threshold corresponding to the resource fingerprint estimation group.

11. The method according to claim 7, characterized in that The determining the estimated value of the first resource based on the target estimated value of the successfully matched resource fingerprint estimation group includes: Determine the similarity difference between the resource fingerprint similarity corresponding to the successfully matched resource fingerprint estimation group and the resource fingerprint similarity threshold; In a case where the similarity difference is less than a similarity difference threshold, determining the target estimated value of the successfully matched resource fingerprint estimation group as the estimated value of the first resource; When the similarity difference is greater than or equal to the similarity difference threshold, the target estimated value of the successfully matched resource fingerprint estimation group is adjusted, and the adjusted target estimated value is determined as the estimated value of the first resource.

12. The method according to claim 11, characterized in that The target estimated value of the resource fingerprint estimation group that is adjusted to match successfully includes: Acquire a target object feature of the target object, and acquire a differential resource feature in the first resource, wherein the differential resource feature is different from a second resource feature of any second resource in a second resource set corresponding to the successfully matched resource fingerprint estimation group; Acquire a trained deep neural network model, and use the target object feature and the difference resource feature as inputs of the deep neural network model to obtain a first output estimate value output by the deep neural network model, wherein the deep neural network model is trained by using the training object feature and the training resource feature in the training data as inputs and the training estimate value in the training data as outputs of the deep neural network model; When the first output estimated value is greater than the estimated value threshold, increasing the target estimated value of the successfully matched resource fingerprint estimated group; When the first output estimated value is less than or equal to the estimated value threshold, the target estimated value of the successfully matched resource fingerprint estimation group is reduced.

13. The method according to claim 1, characterized in that Applied to electronic equipment, the method further includes: Obtaining a determined number of a second estimated value determined based on the estimated value determination algorithm, and determining a quantity difference between a target number and the determined number, wherein the target number is less than or equal to a maximum determined number, and the maximum determined number is: when the second estimated value of the second resource is determined based on the estimated value determination algorithm, the maximum number of the second estimated value determined that is supported by the computing power resources of the electronic device; Acquire a third resource of the quantity difference, where the third resource is used to explore the preferences of the target object; Determine an estimated value of each third resource in the number-difference third resources based on a preset estimated value determination algorithm, wherein the estimated value of the third resource represents a possibility that the third resource is selected by the target object; Based on the estimated value of each third resource in the number of third resources, a target resource in the number of third resources is recommended to the target object.

14. The method according to any one of claims 1 to 13, characterized in that: The second estimated value of the second resource is determined based on a preset estimated value determination algorithm, and the estimated value determination algorithm includes: Acquire a target object feature of the target object, and acquire a resource feature of a resource with an estimated value to be determined, wherein the resource with an estimated value to be determined is at least one of the first resource, the second resource, or the third resource; Acquire a trained deep neural network model, and use the target object features and the resource features of the estimated value resource to be determined as inputs of the deep neural network model to obtain a second output estimated value output by the deep neural network model, wherein the deep neural network model is trained by using the training object features and training resource features in the training data as inputs and the training estimated value in the training data as outputs of the deep neural network model; The second output estimated value is determined as the estimated value of the estimated value resource to be determined.

15. A resource recommendation device, characterized in that: include: A first acquisition module, configured to acquire at least one first resource to be recommended to a target object, and determine a first resource fingerprint of each first resource in the at least one first resource, wherein the first resource fingerprint is determined based on multimedia material information of the first resource; A second acquisition module is used to acquire a resource fingerprint estimation group set corresponding to at least one second resource set related to the target object, the resource fingerprint estimation group set includes a resource fingerprint estimation group corresponding to each second resource set, the resource fingerprint estimation group includes a target resource fingerprint and a target estimation value, the target resource fingerprint corresponding to the second resource set is determined based on the multimedia material information of the second resource in the second resource set, and the target estimation value corresponding to the second resource set represents the possibility that the second resource in the second resource set is selected by the target object; a matching module, configured to match, for each first resource, a first resource fingerprint of the first resource with a target resource fingerprint of each resource fingerprint estimation group; an estimated value determination module, configured to determine an estimated value of the first resource based on a target estimated value of a successfully matched resource fingerprint estimation group when a first resource fingerprint of the first resource successfully matches a target resource fingerprint of one of the resource fingerprint estimation groups, wherein the estimated value of the first resource represents a possibility that the first resource is selected by the target object; The resource recommendation module is used to recommend a target resource from the at least one first resource to the target object based on the estimated value of each first resource.

16. An electronic device comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 14.

17. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 14 are implemented.

18. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 14 are implemented.

Images