A multimedia sequence recommendation method, an operation prediction model training method, device, equipment and storage medium

Abstract

The present disclosure relates to a multimedia sequence recommendation method, an operation prediction model training method, an apparatus, a device and a storage medium. The target object and the target corresponding relationship information of the first object are used as the input of the operation prediction model. The target operation time information corresponding to the target object is used as the input of the operation prediction model. The operation prediction model is used to predict the target corresponding relationship information of the target object and at least one first object, and the target operation time information corresponding to the target object. The first object corresponding target prediction information is obtained. The target prediction information corresponding to at least one first object is combined to determine the target multimedia sequence matched with the target object. The correlation between the target multimedia sequence and the virtual resource interaction operation can be improved. The target multimedia sequence with higher correlation with the virtual resource interaction operation is pushed to the target object. The overall performance of the target multimedia sequence in the actual recommendation scene can be optimized, and the overall benefit of the recommended multimedia sequence can be improved.

Description

Technical Field

[0001] This disclosure relates to the field of computer technology, and in particular to a multimedia sequence recommendation method, an operation prediction model training method, an apparatus, a device, and a storage medium. Background Technology

[0002] With the development of internet technology, numerous online platforms are constantly upgrading. Besides allowing users to share short videos of their daily lives, they also provide platforms for broadcasters to promote their desired content. In related technologies, within a single business operation scenario, the data for that operation is often quite dense. Predictive models can be trained based on the types of items in these dense historical business operations to estimate the probability of a user performing a particular business operation on a certain type of item. This method typically requires a large amount of historical business operation records at the item level. However, pushing multimedia resources to users based on these item-level historical business operation records is challenging due to the diversity of user business operations in short video and live streaming formats. It may be impossible to recommend favorite broadcasters or live streams to users, resulting in low user conversion rates and overall low revenue in short video and live streaming services. Summary of the Invention

[0003] This disclosure provides a multimedia sequence recommendation method, an operation prediction model training method, an apparatus, a device, and a storage medium to at least address the problem in related technologies where, in business scenarios involving multiple services, the correlation between multimedia sequence recommendation and virtual resource interaction operations within these operations cannot be improved. The technical solution of this disclosure is as follows:

[0004] According to a first aspect of the present disclosure, a multimedia sequence recommendation method is provided, comprising:

[0005] Obtain target correspondence information and target operation time information of the target object; the target correspondence information indicates that the target object has performed a virtual resource interaction operation on the virtual resources of at least one first object; the target operation time information is the time information of the target object performing the virtual resource interaction operation on the virtual resources of each first object, and the time from the current time is less than a time threshold.

[0006] The target correspondence information and the target operation time information are input into the operation prediction model to perform prediction processing of the virtual resource interaction operation, thereby obtaining target prediction information corresponding to each of the at least one first object; the target prediction information represents the probability that the predicted target object will perform the virtual resource interaction operation on the at least one first object at a target time; the target time is a preset duration after the current time;

[0007] Based on the target prediction information corresponding to each of the at least one first object, a target multimedia sequence matching the target object is determined; the target multimedia sequence is at least one multimedia resource of the at least one first object; the at least one multimedia resource is associated with a virtual resource;

[0008] The target multimedia sequence is pushed to the target object.

[0009] In one possible implementation, the target operation prediction model includes a feature extraction network, a distributed processing network, and a prediction information processing module. The step of inputting the target correspondence information and the target operation time information into the operation prediction model to perform prediction processing for the virtual resource interaction operation, and obtaining the target prediction information corresponding to each of the at least one first object, includes:

[0010] The target correspondence information is input into the feature extraction network for feature extraction processing to obtain the target correspondence feature information;

[0011] The target correspondence feature information is input into the distributed processing network for distributed processing to obtain the time distribution of virtual resource interaction operations; the time distribution of virtual resource interaction operations refers to the time distribution of the target object performing the virtual resource interaction operation on the at least one first object.

[0012] The prediction information processing module is used to perform prediction processing on the target operation time information and the time distribution to obtain the target prediction information corresponding to each of the at least one first object.

[0013] In one possible implementation, determining the target multimedia sequence matched by the target object based on the target prediction information corresponding to each of the at least one first object includes:

[0014] Obtain the conversion rate information corresponding to the target object; the conversion rate information represents the proportion of multimedia resources in the total multimedia resources pushed to the target object in which the target object performs virtual resource interaction operations; each multimedia resource in the total multimedia resources is associated with the virtual resource;

[0015] The target multimedia sequence is determined based on the conversion rate information and the target prediction information corresponding to each of the at least one first object.

[0016] In one possible implementation, determining the target multimedia sequence matched by the target object based on the target prediction information corresponding to each of the at least one first object includes:

[0017] Based on the target prediction information corresponding to each of the at least one first object, the priority information of the at least one first object is determined;

[0018] Acquire the target multimedia resources of the at least one first object;

[0019] The target multimedia resources are sorted based on the priority information to obtain the target multimedia sequence.

[0020] In one possible implementation, the method further includes:

[0021] Obtain the identification information of a second object that has performed the virtual resource interaction operation; the second object includes the target object;

[0022] Determine the identification information of at least one third object corresponding to the second object; the third object is an object that the second object has performed the virtual resource interaction operation on, and the at least one third object includes the at least one first object;

[0023] Based on the identification information of the second object and the identification information of the third object, construct the correspondence information between the second object and the third object;

[0024] The step of obtaining the target correspondence information between the target object and at least one first object and the target operation time information of the target object includes:

[0025] Obtain the identification information of the target object and the target operation time information;

[0026] The target correspondence information is obtained from the correspondence information between the second object and the third object based on the identification information of the target object.

[0027] In one possible implementation, the distributed processing network includes a parameter fitting network and a parameter processing module; the step of inputting the target correspondence feature information into the distributed determination network for distribution determination processing to obtain the time distribution of virtual resource interaction operations includes:

[0028] The target correspondence feature information is input into the parameter fitting network for parameter fitting processing to obtain the target prediction mean and target prediction variance of the time distribution.

[0029] The parameter processing module is used to perform distribution determination processing on the estimated mean and variance of the target to obtain the time distribution of the virtual resource interaction operation.

[0030] According to a second aspect of the present disclosure, a method for training an operational prediction model is provided, the method comprising:

[0031] Obtain sample correspondence and sample operation time information for each of the sample correspondences; the sample correspondence represents that the target sample object has performed virtual resource interaction operations on the virtual resources of at least one first sample object, and the sample operation time information is the time sequence information of the at least one sample object performing the virtual resource interaction operations on each first sample object;

[0032] The sample correspondence and the sample operation time information are input into the initial operation prediction model for operation prediction processing to obtain sample prediction information; the sample prediction information represents the probability that the target sample object will perform the virtual resource interaction operation during the target sample time period; the target sample time period is a time period obtained based on the time series information;

[0033] Based on the sample operation prediction information and the operation statistics of the target sample object performing the virtual resource interaction operation on the at least one first sample object within the target sample time period, the loss information is determined.

[0034] The initial operation prediction model is trained based on the loss information to obtain the target operation prediction model.

[0035] In one possible implementation, the initial operation prediction model includes: a feature extraction network, a preset distribution processing network, and a preset prediction information processing module; the step of inputting the sample correspondence and the sample operation time information into the initial operation prediction model for operation prediction processing to obtain sample prediction information includes:

[0036] The sample correspondence is input into the feature extraction network for feature extraction processing to obtain sample correspondence feature information;

[0037] The sample correspondence feature information is input into the preset distribution processing network for distribution processing to obtain the sample time distribution of virtual resource interaction operations; the sample time distribution refers to the time distribution of the sample object performing the virtual resource interaction operation on the at least one first sample object.

[0038] The sample time distribution and the sample time information are input into the preset prediction information processing module to obtain the sample operation prediction information.

[0039] In one possible implementation, the preset distribution processing network includes a preset parameter fitting network and a parameter processing module; the step of inputting the sample correspondence feature information into the preset distribution processing network for distribution processing to obtain the sample time distribution of virtual resource interaction operations includes:

[0040] The sample correspondence feature information is input into the preset parameter fitting network for parameter fitting processing to obtain the sample predicted mean and sample predicted variance.

[0041] The parameter processing module is used to perform distribution determination processing on the estimated mean and variance of the samples to obtain the sample time distribution of the virtual resource interaction operation.

[0042] In one possible implementation, the preset parameter fitting network includes a preset number of fully connected layers and a preset number of sigmoid function processing modules; the step of inputting the sample correspondence feature information into the preset parameter fitting network for parameter fitting processing to obtain the sample predicted mean and sample predicted variance includes:

[0043] A preset number of fully connected layers are used to perform parameter fitting on the sample correspondence feature information to obtain a first intermediate value and a second intermediate value.

[0044] The first intermediate value and the second intermediate value are input into the preset number of S-shaped function processing modules for compression processing to obtain the sample estimated mean and the sample estimated variance.

[0045] According to a third aspect of the present disclosure, a multimedia sequence recommendation apparatus is provided, comprising:

[0046] The target information acquisition module is configured to acquire target correspondence information and target operation time information of the target object; the target correspondence information indicates that the target object has performed a virtual resource interaction operation on the virtual resources of at least one first object; the target operation time information is the time information of the target object performing the virtual resource interaction operation on the virtual resources of each first object, and the time since the current time is less than a time threshold.

[0047] The target information prediction module is configured to input the target correspondence information and the target operation time information into the operation prediction model to perform prediction processing of the virtual resource interaction operation, and obtain target prediction information corresponding to each of the at least one first object; the target prediction information represents the probability that the predicted target object will perform the virtual resource interaction operation on the at least one first object at a target time; the target time is a preset duration after the current time;

[0048] A multimedia sequence determination module is configured to determine a target multimedia sequence matching the target object based on target prediction information corresponding to each of the at least one first object; the target multimedia sequence is at least one multimedia resource of the at least one first object; the at least one multimedia resource is associated with a virtual resource.

[0049] The push module is configured to push the target multimedia sequence to the target object.

[0050] In one possible implementation, the target operation prediction model includes a feature extraction network, a distributed processing network, and a prediction information processing module, wherein the information prediction module includes:

[0051] The first feature extraction unit is configured to input the target correspondence information into the feature extraction network for feature extraction processing to obtain target correspondence feature information;

[0052] The first parameter processing unit is configured to input the target correspondence feature information into the distributed processing network for distributed processing to obtain the time distribution of virtual resource interaction operations; the time distribution of virtual resource interaction operations refers to the time distribution of the target object performing the virtual resource interaction operation on the at least one first object.

[0053] The information prediction unit is configured to use the prediction information processing module to perform prediction processing on the target operation time information and the time distribution to obtain target prediction information corresponding to each of the at least one first object.

[0054] In one possible implementation, the multimedia sequence determination module includes:

[0055] A conversion rate acquisition unit is configured to acquire conversion rate information corresponding to the target object; the conversion rate information represents the proportion of multimedia resources in the total multimedia resources pushed to the target object in which the target object performs virtual resource interaction operations; each multimedia resource in the total multimedia resources is associated with the virtual resource;

[0056] The multimedia sequence determination unit is configured to determine the target multimedia sequence based on the conversion rate information and the target prediction information corresponding to each of the at least one first object.

[0057] In one possible implementation, the multimedia sequence determination module includes:

[0058] The priority information determination unit is configured to determine the priority information of the at least one first object based on the target prediction information corresponding to each of the at least one first object;

[0059] A multimedia resource acquisition unit is configured to acquire the target multimedia resources of the at least one first object;

[0060] The sorting unit is configured to sort the target multimedia resources based on the priority information to obtain the target multimedia sequence.

[0061] In one possible implementation, the device further includes:

[0062] The identification information acquisition module is configured to acquire identification information of a second object that has performed the virtual resource interaction operation; the second object includes the target object;

[0063] The identification information determination module is configured to determine the identification information of at least one third object corresponding to the second object; the third object is an object that the second object has performed the virtual resource interaction operation on, and the at least one third object includes the at least one first object;

[0064] The correspondence information construction module is configured to construct correspondence information between the second object and the third object based on the identification information of the second object and the identification information of the third object.

[0065] The target information acquisition module includes:

[0066] The first information acquisition unit is configured to acquire the identification information of the target object and the target operation time information;

[0067] The target correspondence information acquisition unit is configured to acquire the target correspondence information from the correspondence information between the second object and the third object based on the identification information of the target object.

[0068] In one possible implementation, the target information prediction module includes:

[0069] The parameter fitting unit is configured to input the target correspondence feature information into the parameter fitting network for parameter fitting processing to obtain the target prediction mean and target prediction variance of the time distribution.

[0070] The time distribution determination unit is configured to use the parameter processing module to perform distribution determination processing on the target estimated mean and the target estimated variance to obtain the time distribution of the virtual resource interaction operation.

[0071] According to a fourth aspect of the present disclosure, an operational prediction model training apparatus is provided, the apparatus comprising:

[0072] The sample information acquisition module is configured to acquire sample correspondence and sample operation time information of each of the sample correspondences; the sample correspondence represents that the target sample object has performed virtual resource interaction operation on the virtual resource of at least one first sample object, and the sample operation time information is the time sequence information of the at least one sample object performing the virtual resource interaction operation on each first sample object;

[0073] The sample information prediction module is configured to input the sample correspondence and the sample operation time information into an initial operation prediction model for operation prediction processing to obtain sample prediction information; the sample prediction information represents the probability of the target sample object performing the virtual resource interaction operation during the target sample time period; the target sample time period is a time period obtained based on the time series information;

[0074] The loss determination module is configured to determine loss information based on the sample operation prediction information and the operation statistics of the target sample object performing the virtual resource interaction operation on the at least one first sample object within the target sample time period.

[0075] The training module is configured to train the initial operation prediction model based on the loss information to obtain the target operation prediction model.

[0076] In one possible implementation, the initial operation prediction model includes: a feature extraction network, a preset distribution processing network, and a preset prediction information processing module; the sample information prediction module includes:

[0077] The second feature extraction unit is configured to input the sample correspondence into the feature extraction network for feature extraction processing to obtain sample correspondence feature information.

[0078] The first sample time distribution determination unit is configured to input the sample correspondence feature information into the preset distribution processing network for distribution processing to obtain the sample time distribution of virtual resource interaction operations; the sample time distribution refers to the time distribution of the sample object performing the virtual resource interaction operation on the at least one first sample object.

[0079] The second information prediction unit is configured to input the sample time distribution and the sample time information into the preset prediction information processing module to obtain the sample operation prediction information.

[0080] In one possible implementation, the preset distribution processing network includes a preset parameter fitting network and a parameter processing module; the first sample time distribution determination unit includes:

[0081] The sample parameter fitting unit is configured to input the sample correspondence feature information into the preset parameter fitting network for parameter fitting processing to obtain the sample estimated mean and sample estimated variance.

[0082] The second sample time distribution determination unit is configured to use the parameter processing module to perform distribution determination processing on the estimated mean and the estimated variance of the samples to obtain the sample time distribution of the virtual resource interaction operation.

[0083] In one possible implementation, the preset parameter fitting network includes a preset number of fully connected layers and the preset number of sigmoid function processing modules; the second parameter processing unit includes:

[0084] The intermediate value determination unit is configured to perform parameter fitting processing on the sample correspondence feature information using a preset number of fully connected layers to obtain a first intermediate value and a second intermediate value.

[0085] The sample prediction value determination unit is configured to input the first intermediate value and the second intermediate value into the preset number of S-shaped function processing modules for compression processing to obtain the sample prediction mean and the sample prediction variance.

[0086] According to a fifth aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the method as described in any one of the first aspects above.

[0087] According to a sixth aspect of the present disclosure, a computer-readable storage medium is provided such that, when instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to perform any of the methods described in the first aspect of the present disclosure.

[0088] According to a seventh aspect of the present disclosure, a computer program product is provided, including computer instructions that, when executed by a processor, cause a computer to perform the method described in any one of the first aspects of the present disclosure. The technical solutions provided by the embodiments of the present disclosure offer at least the following beneficial effects:

[0089] By using the target object's target correspondence information and target operation time information as inputs to the operation prediction model, the correlation between the target object and the object that has performed virtual resource interaction operations can be improved. Since the target correspondence information represents the relationship between the target object and the object that has performed virtual resource interaction operations, and the target operation time information represents the operation time corresponding to the target object's virtual resource interaction operations, the correlation between the predicted probability of virtual resource interaction operations and the first object and historical virtual resource interaction operation times can be improved. The operation prediction model is then used to predict the target correspondence information and target operation time information to obtain target prediction information corresponding to at least one first object. Combining the target prediction information corresponding to at least one first object, the target multimedia sequence matching the target object is determined. Since the target prediction information can characterize the predicted probability that the target object will perform virtual resource interaction operations on the first object at the target time, the correlation between the target multimedia sequence and the virtual resource interaction operation can be improved. Pushing the target multimedia sequence with a higher correlation to the virtual resource interaction operation to the target object can optimize the overall performance of the target multimedia sequence in the actual recommendation scenario, improve the virtual resource conversion rate during multimedia sequence recommendation, and thus improve the overall benefit of multimedia sequence recommendation. Attached Figure Description

[0090] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure, and are not intended to unduly limit this disclosure.

[0091] Figure 1 This is a schematic diagram illustrating an implementation environment according to an exemplary embodiment.

[0092] Figure 2 This is a flowchart illustrating a multimedia sequence recommendation method according to an exemplary embodiment.

[0093] Figure 3 This is a flowchart illustrating, according to an exemplary embodiment, a method for determining a target multimedia sequence that matches a target object based on target prediction information corresponding to at least one first object.

[0094] Figure 4 This is a flowchart illustrating a method for obtaining target prediction information using an operational prediction model, according to an exemplary embodiment.

[0095] Figure 5 This is a flowchart illustrating a method for determining a target multimedia sequence that matches a target object based on target prediction information corresponding to at least one first object, according to an exemplary embodiment.

[0096] Figure 6 This is a flowchart illustrating a method for constructing correspondence information between a second object and a third object, according to an exemplary embodiment.

[0097] Figure 7 This is a flowchart illustrating a method for training an operational prediction model according to an exemplary embodiment.

[0098] Figure 8 This is a flowchart illustrating a method for inputting sample correspondence and sample operation time information into an initial operation prediction model to perform operation prediction processing and obtain sample prediction information, according to this embodiment.

[0099] Figure 9 This is a block diagram illustrating a multimedia sequence recommendation device according to an exemplary embodiment.

[0100] Figure 10 This is a block diagram of an operational prediction model training apparatus according to an exemplary embodiment.

[0101] Figure 11 This is a block diagram illustrating an electronic device for use in a multimedia sequence recommendation method and / or an operational prediction model training method, according to an exemplary embodiment. Detailed Implementation

[0102] To enable those skilled in the art to better understand the technical solutions of this disclosure, the technical solutions in the embodiments of this disclosure will be clearly and completely described below with reference to the accompanying drawings.

[0103] It should be noted that the terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this disclosure are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure as detailed in the appended claims.

[0104] Please see Figure 1 It illustrates a schematic diagram of an implementation environment provided by an embodiment of this disclosure, which may include:

[0105] At least one terminal 01 and at least one server 02. The at least one terminal 01 and the at least one server 02 can communicate data via a network.

[0106] In an optional embodiment, terminal 01 can be a terminal that sends a target multimedia sequence recommendation request or a terminal that pushes multimedia sequences to server 02. Terminal 01 can be, but is not limited to, electronic devices such as smartphones, desktop computers, tablets, laptops, smart speakers, digital assistants, augmented reality (AR) / virtual reality (VR) devices, and smart wearable devices. The operating system running on terminal 01 can be, but is not limited to, Android, iOS, Linux, Windows, and Unix.

[0107] In an optional embodiment, server 02 may be a server that pushes the target multimedia sequence to terminal 01. Optionally, server 02 may be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms.

[0108] It should be noted that the following diagram illustrates one possible sequence of steps, and it is not strictly required to follow this order. Some steps can be performed in parallel without interdependence. The user information (including but not limited to user device information, user personal information, user behavior information, etc.) and data (including but not limited to data used for display, training data, etc.) involved in this disclosure are all information and data authorized by the user or fully authorized by all parties.

[0109] Figure 2 This is a flowchart illustrating a multimedia sequence recommendation method according to an exemplary embodiment. This multimedia sequence recommendation method can be applied to server O2, such as... Figure 2 As shown, the multimedia sequence recommendation method may include the following steps:

[0110] In step S21, the target correspondence information and target operation time information of the target object are obtained.

[0111] In the embodiments of this specification, the target object can be an object that has performed at least one virtual resource interaction operation. The target correspondence information can characterize that the target object has performed a virtual resource interaction operation on the virtual resources of at least one first object. The at least one first object can be an object on which the target object has performed a virtual resource interaction operation. The virtual resource interaction operation can be used for the target object to interact with at least one first object on virtual resources. In practical applications, the target object can perform virtual resource interaction operations based on the multimedia resources of at least one first object. Specifically, the virtual resource interaction operation can be a virtual resource interaction operation triggered by the target object in a virtual room provided by an Internet platform, or it can be a virtual resource interaction operation triggered by the target object in multimedia resources published on an Internet platform. By performing the virtual resource interaction operation, the target object can obtain the corresponding virtual resource, which can be a virtual resource associated with at least one first object.

[0112] It should be emphasized that each first object has a corresponding relationship with a virtual resource. For example, multiple first objects can correspond to one virtual resource, or each first object can correspond to a different virtual resource.

[0113] In practical applications, the virtual space or multimedia resource corresponding to the first object can be configured with target functions corresponding to virtual resource interaction operations. The target object can achieve these target functions within the virtual space or multimedia resource corresponding to the first object by executing the virtual resource interaction operation. The virtual space corresponding to the first object can be a virtual space created based on the first object's request, such as a chat room or live streaming room; the multimedia resource corresponding to the first object can refer to multimedia content captured and / or produced by the first object. It should be noted that the virtual resource interaction operation is one preset operation among multiple business operations in the business scenarios of this specification's embodiments. In the business scenarios of this specification's embodiments, in addition to the virtual resource interaction operation, the target object can perform various other preset operations in the virtual space or multimedia resource corresponding to the first object to achieve other corresponding functions. For example, these various other preset operations can be browsing, liking, commenting, sharing, etc.

[0114] In the embodiments of this specification, the target operation time information can be operation time information that is less than a time threshold from the current time among at least one operation time information. This at least one operation time information is the operation time information corresponding to the virtual resource interaction operation performed by the target object. In practical applications, different time thresholds can be set for different target objects. For example, if the time span between two consecutive virtual resource interaction operations performed by the target object is long, the time threshold can be set to a larger threshold; if the time span between two consecutive virtual resource interaction operations performed by the target object is short, the time threshold can be set to a smaller threshold. Specifically, at least one operation time information can be a time sequence of virtual resource interaction operations performed by the target object. For example, according to the order in which the target object performs virtual resource interaction operations, the time sequence is arranged from the time furthest from the current time to the time closest as t1, t2, t3, t4, where t4 is less than the current time threshold, and t4 can be used as the target operation time.

[0115] In step S22, the target correspondence information and target operation time information are input into the operation prediction model to perform prediction processing of virtual resource interaction operations, so as to obtain target prediction information corresponding to at least one first object.

[0116] In the embodiments of this specification, target prediction information can characterize the probability that a target object will perform a virtual resource interaction operation on at least one first object at a target time. The target time can be a preset duration after the current time. The target prediction information corresponding to each of the at least one first object characterizes the predicted probability that the target object will perform a virtual resource interaction operation on each of the at least one first object at the target time. Taking a short video platform scenario as an example, the target correspondence information is the correspondence information between user 1 (the target object) and users 2, 3, and 4. The target prediction information can be the prediction information that user 1 will perform a virtual resource interaction operation in the virtual space or multimedia resources corresponding to users 2, 3, and 4, respectively. In practical applications, the target prediction information can represent the probability of performing a virtual resource interaction operation through numerical values. For example, 0.6 in the target prediction information represents a higher probability of performing a virtual resource interaction operation than 0.3. It should be noted that the numerical values ​​in the target prediction information can be values ​​within the range of 0 to 1, 0 to 100, etc., and this disclosure does not limit the numerical range of the target prediction information. In the embodiments described in this specification, the operation prediction model can be a model obtained by training a preset neural network. The specific training steps will be introduced in the following content.

[0117] In step S23, the target multimedia sequence matching the target object is determined based on the target prediction information corresponding to at least one first object.

[0118] In the embodiments of this specification, the target multimedia sequence is at least one multimedia resource of at least one first object. For example, the target multimedia sequence may refer to the virtual space sequence of each of the at least one first object, or it may refer to the multimedia resource sequence corresponding to each of the at least one first object. The at least one multimedia resource is associated with a virtual resource. In actual use, multimedia resources can be directly pushed based on the target multimedia sequence in the embodiments of this specification, or the target multimedia sequence in the embodiments of this specification can be combined with other target multimedia sequence recommendation methods to recommend multimedia sequences. For example, the target object may correspond to at least one first object and an object that has not undergone virtual resource interaction. The object that has not undergone virtual resource interaction can be an object determined based on the target object's points of interest. In order to improve the association between multimedia sequence recommendation and virtual resource interaction in multi-service operations, the priority of the target multimedia sequence can be set higher than the multimedia corresponding to the object that has not undergone virtual resource interaction. Optionally, such as Figure 3 As shown, determining the target multimedia sequence matching the target object based on the target prediction information corresponding to at least one first object may include the following steps:

[0119] In step S231, priority information of at least one first object is determined based on the target prediction information corresponding to each of the at least one first object.

[0120] In the embodiments of this specification, the target prediction information corresponding to at least one first object can represent the predicted probability that the target object will perform a virtual resource interaction operation on each first object at a target time. For example, the probability that the target object will perform a virtual resource interaction operation on each first object at time T+Δt can be predicted at time T. The target prediction information corresponding to each first object can be sorted, and the priority information of the at least one first object can be determined according to the sorting of the target prediction information. Specifically, the higher the probability that the predicted target object will perform a virtual resource interaction operation at the current time, the higher the priority ranking of the corresponding first object.

[0121] For example, taking the target prediction information corresponding to at least one first object as M1, M2, and M3 corresponding to first object A1, first object A2, and first object A3 respectively, where M1, M2, and M3 are 0.3, 0.6, and 0.1 respectively, the priority information of the at least one first object can be shown in Table 1 below:

[0122] Table 1

[0123]

[0124] In step S232, at least one target multimedia resource of the first object is acquired.

[0125] In the embodiments of this specification, the target multimedia resource of at least one first object can be a multimedia resource published by the first object that meets preset conditions. Optionally, the acquisition conditions for multimedia resources can be set according to at least one of the following information: time information of the multimedia resources published by the first object, quantity information of the multimedia resources published by the first object, whether it is delivered to the target object, quantity of the first object, priority information of the first object, etc., and the target multimedia resource of the at least one first object can be acquired based on the acquisition conditions. For example, taking the acquisition conditions set according to the quantity of the first object and the quantity information of the multimedia resources published by the first object as an example, it can be set that when the quantity of the first object is less than 3, at least 10 of the most recently published multimedia resources of each first object are acquired and the most recently published multimedia resources of each first object are used as target multimedia resources; when the quantity of the first object is not less than 3, at least 5 of the most recently published multimedia resources of each first object are acquired and the most recently published multimedia resources of each first object are used as target multimedia resources.

[0126] In step S233, the target multimedia resources are sorted based on priority information to obtain the target multimedia sequence.

[0127] In the embodiments described in this specification, the order of the target multimedia resources can be positively correlated with priority information.

[0128] In one specific embodiment, the number of the at least one first object can be one, and the target prediction information can be the prediction information of the first object. Multiple multimedia resources corresponding to the first object can be sorted based on the multimedia resource release time of the first object, whether the target object has browsed the multimedia resources of the first object, etc., to obtain a target multimedia sequence, and the target multimedia sequence can be pushed to the target object.

[0129] In one specific embodiment, the number of the at least one first object can be multiple, and the target prediction information can be the prediction information of the multiple first objects. The target multimedia sequence can be obtained based on the prediction information of the multiple first objects, or the prediction information of the multiple first objects can be combined with the release time of the multimedia resources of the multiple first objects, whether the target object has browsed the multimedia resources of the first objects, etc., to sort the multiple multimedia resources of the multiple first objects, obtain the target multimedia sequence, and push the target multimedia sequence to the target object.

[0130] In a specific embodiment, a target multimedia sequence matching the target object can be determined based on the target prediction information corresponding to each of the at least one first object. Then, the target multimedia sequence is combined with the multimedia resource sequence of an object that has not undergone virtual resource interaction operation to obtain a multimedia resource sequence to be pushed to the target object. The combined multimedia resource sequence is then pushed to the target object.

[0131] By determining the priority information of at least one first object based on the target prediction information corresponding to each first object, the priority information can effectively express the target prediction information. By acquiring the target multimedia resources of the at least one first object and sorting the target multimedia resources based on the priority information to obtain the target multimedia sequence, the correlation between the target multimedia sequence and the priority information can be improved, thereby making the target multimedia sequence determined by the priority information more accurate and improving the overall benefit of the recommended multimedia sequence.

[0132] In step S24, the target multimedia sequence is pushed to the target object.

[0133] In the above embodiments, using the target correspondence information between the target object and at least one first object, as well as the target operation time information corresponding to the target object, as input to the operation prediction model can improve the correlation between the predicted probability of virtual resource interaction operations and the interaction time of the first object and historical virtual resource operations. The operation prediction model is used to predict the target correspondence information between the target object and at least one first object, as well as the target operation time information corresponding to the target object, to obtain target prediction information corresponding to each of the at least one first object. Combining the target prediction information corresponding to each of the at least one first object, the target multimedia sequence matching the target object is determined, which can improve the correlation between the target multimedia sequence and the virtual resource interaction operation. Pushing the target multimedia sequence with a higher correlation to the virtual resource interaction operation to the target object can optimize the overall performance of the target multimedia sequence in the actual recommendation scenario, improve the virtual resource conversion rate during multimedia sequence recommendation, and thus improve the overall benefit of multimedia sequence recommendation. Taking the virtual resource interaction operation of purchasing goods in multimedia resources published by users as an example, the target user has purchased goods in the multimedia resources of user 1 and user 2. The target correspondence information records the correspondence between the target user and user 1 and user 2. The time of the target user's most recent purchase in user 1's multimedia resources is U1t, and the time of the most recent purchase in user 2's multimedia resources is U2t. When recommending multimedia resources to the target user, the correspondence between the target user and user 1 and user 2, as well as U1t and U2t, can be input into the operation prediction model to predict the probabilities P1 and P2 of the target user making purchases to user 1 and user 2 at the target time, respectively. Based on the probabilities P1 and P2, the target multimedia sequence to be pushed to the target user can be determined, which can make the recommended target multimedia sequence more in line with the target user's preferences or needs, thereby improving the conversion rate of multimedia resources in the target multimedia sequence.

[0134] Figure 4 This is a flowchart illustrating a method for obtaining target prediction information using an operation prediction model according to an exemplary embodiment. In this embodiment, the target operation prediction model may include a feature extraction network, a distributed processing network, and a prediction information processing module. Based on this, inputting target correspondence information and target operation time information into the operation prediction model for prediction processing of virtual resource interaction operations to obtain target prediction information corresponding to at least one first object may include:

[0135] In step S41, the target correspondence information is input into the feature extraction network for feature extraction processing to obtain the target correspondence feature information.

[0136] In the embodiments of this specification, the feature extraction network can be at least one DenseNet network module. For example, the feature extraction network can be obtained by concatenating two DenseNet (a type of neural network) network modules. The two DenseNet network modules used to extract correspondence feature information can be pre-trained. The pre-training of the DenseNet network modules can be performed before the parameter fitting network is trained, or it can be trained together with the parameter fitting network; this disclosure does not limit this.

[0137] In step S42, the target correspondence feature information is input into the distributed processing network for distributed processing to obtain the time distribution of virtual resource interaction operations.

[0138] In the embodiments of this specification, the time distribution of virtual resource interaction operations can refer to the time distribution of a target object performing virtual resource interaction operations on at least one first object. Specifically, the time distribution of virtual resource interaction operations can be...

[0139] In the embodiments of this specification, the time distribution of virtual resource interaction operations can be a log-normal distribution, that is, the correspondence between the target object and at least one first object, and the time span of the target object performing virtual resource interaction operations, follow a log-normal distribution. Specifically, the horizontal axis of this log-normal distribution can represent the time span of the virtual resource interaction operation being executed, and the vertical axis can represent the intensity of the target object performing the virtual resource interaction operation. The intensity of the target object performing the virtual resource interaction operation can be statistical information about the target object's performance of the virtual resource interaction operation. On the vertical axis, the greater the intensity of the target object performing the virtual resource interaction operation, the more times the target object performs the virtual resource interaction operation.

[0140] In a specific embodiment, the distribution processing network may include a parameter fitting network and a parameter processing network. Accordingly, inputting the target correspondence feature information into the distribution determination network for distribution determination processing to obtain the time distribution of virtual resource interaction operations may include: inputting the target correspondence feature information into the parameter fitting network for parameter fitting processing to obtain the target estimated mean and target estimated variance of the time distribution; and using the parameter processing module to perform distribution determination processing on the target estimated mean and target estimated variance to obtain the time distribution of virtual resource interaction operations.

[0141] In a specific embodiment, any point (x, f(x)) in the log-normal distribution can represent the number of times the target object performs virtual resource interaction operations on each first object within the time span x.

[0142] In step S43, the prediction information processing module is used to perform prediction processing on the time distribution and target prediction variance to obtain target prediction information corresponding to at least one first object.

[0143] In this embodiment of the specification, after obtaining the target estimated mean and target estimated variance in step S42, the target operation time information, the target estimated mean and the target estimated variance can be input into the prediction information processing module, and the output results can be used as the target prediction information corresponding to at least one object.

[0144] In the embodiments of this specification, the time distribution of virtual resource interaction operations can be determined based on the target estimated value and the target estimated variance. The prediction information processing module can be used to predict the target operation time information, the target estimated mean, and the target estimated variance, which can be achieved by combining a log-normal distribution function. This log-normal distribution function can be any function corresponding to the aforementioned log-normal distribution.

[0145] In a specific embodiment, the log-normal distribution function can be expressed as follows:

[0146]

[0147] Where f(μ, σ; x) represents the intensity of the virtual resource interaction operation performed by the target object on at least one first object, x represents the time span of the virtual resource interaction operation performed by the target object, μ represents the mean of the log-normal distribution, σ represents the variance of the log-normal distribution, exp is an exponential function with the natural constant e as the base, and ln(x) is a logarithmic function with x as the independent variable.

[0148] Specifically, if the target operation time is T1 and the current time is T2, the time span x in the above formula can be obtained by calculating the difference between the target operation time T1 and the current time T2; and the target prediction mean and target prediction variance in step S32 are respectively used as μ and σ in the log-normal distribution function. The time span x, the target prediction mean μ and the target prediction variance σ are substituted into the above log-normal distribution function to obtain the target prediction information of each first object corresponding to the target object.

[0149] In the above embodiments, the target correspondence information is input into the feature extraction network for feature extraction processing. By extracting features from the target correspondence, relatively dense target correspondence feature information is obtained. The relatively dense target correspondence feature information is then input into the distribution processing network for distribution processing to obtain the time distribution of virtual resource interaction operations, thereby obtaining the time distribution of target object matching. The prediction information processing module is used to predict the target operation time information and time distribution to obtain target prediction information corresponding to at least one first object. Since the time and object of each target object performing virtual resource interaction operations on the first object are different, this disclosure uses the time distribution corresponding to the target object and the target operation time information to perform probability prediction processing of virtual resource interaction operations. This allows for targeted prediction of virtual resource interaction operations by target objects, thereby improving the accuracy of target prediction information.

[0150] Furthermore, determining the time distribution of virtual resource interaction operations based on the target estimate and target estimate variance can improve the matching degree between the time distribution and the target object.

[0151] Figure 5 This is a flowchart illustrating a method for determining a target multimedia sequence that matches a target object based on target prediction information corresponding to at least one first object, according to an exemplary embodiment.

[0152] In step S51, the conversion rate information corresponding to the target object is obtained.

[0153] In the embodiments of this specification, the conversion rate information can characterize the proportion of multimedia resources in the total multimedia resources pushed to the target object, in which the target object performs virtual resource interaction operations; wherein, each multimedia resource in the total multimedia resources is associated with a virtual resource. The conversion rate information can be in numerical form, such as a value in the range of 0 to 1, and this disclosure does not limit it in this way.

[0154] In practical applications, each multimedia resource in the full set of multimedia resources can be a multimedia resource for which the target object performs a preset operation. In the embodiments of this specification, the preset operation can be an associated operation of the virtual resource interaction operation. In some examples, the preset operation can be a necessary operation for the virtual resource interaction operation, that is, the virtual resource interaction operation needs to be executed after the preset operation. For example, before performing the virtual resource interaction operation (clicking the link information in the multimedia resource), the preset operation (browsing and playing the multimedia resource after it is exposed) needs to be performed first; or the preset operation can be another preset operation besides the virtual resource interaction operation. This other preset operation can be an operation that, together with the preset operation, reflects the target object's point of interest. For example, during the playback of a multimedia resource, the target object performs a preset operation (like) and also performs a virtual resource interaction operation (clicking the link information in the multimedia resource).

[0155] Taking 100,000 virtual spaces or multimedia resources that have performed preset operations as an example, the number of virtual spaces or multimedia resources that the target object has also performed virtual resource interaction operations can be counted from these 100,000 historical records, denoted as N; based on the 100,000 virtual spaces or multimedia resources that have performed preset operations, and the counted number N of virtual spaces or multimedia resources that the target object has performed preset operations, the statistical information of the target object performing preset operations can be determined as N / 100,000.

[0156] In step S52, the target multimedia sequence is determined based on the conversion rate information and the target prediction information corresponding to at least one first object.

[0157] In the embodiments of this specification, based on conversion rate information and target prediction information corresponding to at least one first object, the conversion rate information and target prediction information can be fused. Optionally, the conversion rate information of the first object to the target object can be estimated by a preset conversion rate model, and the conversion rate information is multiplied with the target prediction information to obtain the ranking criteria of the target multimedia resources. Based on the ranking criteria of the target multimedia resources, the target multimedia sequence is determined.

[0158] The above embodiments, by acquiring conversion rate information and determining the target multimedia sequence based on the conversion rate and the target prediction information corresponding to at least one first object, can improve the correlation between the target multimedia sequence and the preset operation, further optimize the overall performance of the target multimedia sequence in the actual recommendation scenario, and thus improve the overall benefit of the recommended multimedia sequence.

[0159] Figure 6 This is a flowchart illustrating a method for constructing correspondence information between a second object and a third object, according to an exemplary embodiment. The method may include the following steps:

[0160] In step S61, the identification information of the second object that has performed virtual resource interaction operation is obtained.

[0161] In the embodiments of this specification, the second object can refer to all objects that have performed virtual resource interaction operations. The second object may include the target object. The identification information of the second object can characterize its identity information, such as an identity serial number.

[0162] In step S62, the identification information of at least one third object corresponding to the second object is determined.

[0163] In the embodiments of this specification, at least one third object corresponding to the second object can refer to an object on which the second object has performed a virtual resource interaction operation, and the at least one third object may include at least one first object. In practical applications, an object can act as either the second object when performing a virtual resource interaction operation or the third object when being subjected to a virtual resource interaction operation, depending on the judgment of the executor or the executed party of the virtual resource interaction operation.

[0164] In the embodiments of this specification, the identification information of at least one third object corresponding to the second object can characterize the identity information of the third object, such as an identity identification serial number.

[0165] In step S63, based on the identification information of the second object and the identification information of the third object, the correspondence information between the second object and the third object is constructed.

[0166] In the embodiments of this specification, a correspondence between the second object and the third object can be constructed based on the identification information of the second object and the identification information of at least one third object corresponding to the second object. Specifically, the correspondence between the second object and the third object can include multiple correspondence information entries, where each correspondence information entry records the identification information of each second object and the third object corresponding to that second object. In practical use, this correspondence information between the second object and the third object can be stored on a server.

[0167] Optionally, each correspondence information may record the identification information of a second correspondence and a third object corresponding to different times. For example, if the second object (user 1) has performed virtual resource interaction operations on the third objects (user 2, user 2, user 3, user 3) respectively, the correspondence information may be recorded as: user 1 corresponds to user 2, user 2, user 3, user 3.

[0168] Optionally, each correspondence information may record the identification information of a second object and a non-duplicate third object. For example, if the second object (user 1) has performed virtual resource interaction operations on the third objects (user 2, user 2, user 3, user 3), the correspondence information may be recorded as: user 1 corresponds to user 2 and user 3.

[0169] Based on this, step S21 may include: obtaining the identification information of the target object and the target operation time information, and obtaining the target correspondence information from the correspondence information between the second object and the third object according to the identification information of the target object.

[0170] In this embodiment, by obtaining the identification information of the second object that has performed virtual resource interaction operations, and determining the identification information of at least one third object corresponding to the second object, the second objects that meet the conditions can be pre-screened. Based on the identification information of the second object and the third object, the correspondence information between the second object and the third object can be constructed. The correspondence information between the second object and the third object that meet the conditions can be extracted and constructed from a large amount of object data. Furthermore, when applying the method, the target correspondence information can be obtained from the correspondence information between the second object and the third object based on the identification information of the target object. This can improve the processing efficiency of the target correspondence information and thus improve the overall processing speed of the multimedia sequence recommendation method.

[0171] In some embodiments, the correspondence information between the second object and the third object can be updated in real time. Specifically, when a second object performs a virtual resource interaction operation on a new object, the second object can be updated based on the new object to update the correspondence information between the second object and the third object. When an object that has never performed a virtual resource interaction operation performs one, the correspondence information between the second object and the third object is updated based on the newly performing virtual resource interaction operation and the object being interacted with. This allows the application to extract accurate target correspondence information based on the updated correspondence information between the second and third objects, thereby improving the accuracy of multimedia sequence recommendations.

[0172] Figure 7 This is a flowchart illustrating an operation prediction model training method according to an exemplary embodiment. The operation prediction model training method in this embodiment can train an operation prediction model used in a multimedia sequence recommendation method. Specifically, the operation prediction model training method may include the following steps:

[0173] In step S71, the sample correspondence and the sample operation time information of each sample correspondence are obtained.

[0174] In the embodiments of this specification, the sample correspondence can refer to the correspondence between at least one sample object and at least one first sample object. The sample operation time information is the time sequence information of the target sample object performing virtual resource interaction operations on at least one first sample object. The at least one first sample object is an object on which the target sample has performed virtual resource interaction operations. This time sequence information can be a sequence including at least two time information items. In practical applications, the correspondence between the target sample object and at least one first sample object can be a correspondence selected from a large amount of historical data. For example, it can be a selection of objects that have performed virtual resource interaction operations within the past six months, using these objects as target sample objects and the objects on which the target sample objects have performed virtual resource interaction operations within the past six months as first sample objects. Based on the selected target sample objects and their corresponding first sample objects, the correspondence between the target sample object and at least one first sample object is determined. Optionally, multiple sample correspondences can be obtained from pre-constructed correspondence information between second and third objects.

[0175] In the embodiments of this specification, the sample operation time information can be the time sequence information of the target sample object performing virtual resource interaction operations.

[0176] In step S72, the sample correspondence and sample operation time information are input into the initial operation prediction model for prediction processing to obtain sample prediction information.

[0177] In the embodiments of this specification, the initial operation prediction model can be a pre-defined neural network. Sample prediction information can characterize the probability that a target sample object will perform a virtual resource interaction operation within a target sample time period. The target sample time period can be a time period obtained based on time series information. For example, the time series information includes time t1 and time t2, and the target sample time period can be t2-t1.

[0178] In one specific embodiment, the initial operational prediction model may include: a feature extraction network, a preset distribution processing network, and a preset prediction information processing module. Based on this, Figure 8 This embodiment illustrates a method for inputting sample correspondence and sample operation time information into an initial operation prediction model for operation prediction processing to obtain sample prediction information. Specifically, the method may include:

[0179] In step S81, the sample correspondence is input into the feature extraction network for feature extraction processing to obtain the sample correspondence feature information.

[0180] In the embodiments of this specification, the feature extraction network can be a pre-trained neural network or a neural network to be trained. When the feature extraction network is a neural network to be trained, it can be trained together with a network that fits preset parameters.

[0181] In step S82, the sample correspondence feature information is input into a preset distribution processing network for distribution processing to obtain the sample time distribution of virtual resource interaction operations.

[0182] In the embodiments of this specification, the sample time distribution may refer to the time distribution of sample objects performing virtual resource interaction operations on at least one sample object.

[0183] In a specific embodiment, the preset distribution processing network may include a preset parameter fitting network and a parameter processing module. Accordingly, inputting the sample correspondence into the feature extraction network for feature extraction to obtain sample correspondence feature information may include: inputting the sample correspondence feature information into the preset parameter fitting network for parameter fitting to obtain the sample estimated mean and sample estimated variance; and using the parameter processing module to perform distribution determination processing on the sample estimated mean and sample estimated variance to obtain the sample time distribution of virtual resource interaction operations. Here, the sample estimated mean and sample estimated variance can be parameters from a preset normal distribution.

[0184] In the embodiments of this specification, the preset normal distribution may refer to the time distribution of the target sample object performing virtual resource interaction operations on at least one first sample object.

[0185] Optionally, the preset parameter fitting network may include a preset number of fully connected layers and a preset number of sigmoid function processing modules. The sample correspondence feature information is input into the preset parameter fitting network, and the preset number of fully connected layers can perform parameter fitting processing on the sample correspondence feature information to obtain a first intermediate value and a second intermediate value. The first and second intermediate values ​​are then input into the preset number of sigmoid function processing modules for compression processing to obtain the sample predicted mean and sample predicted variance. For example, four fully connected layers can be set, with the number of network nodes in these four fully connected layers being [512, 256, 128, 2], thereby achieving a better fitting effect. A sigmoid function processing module can be connected to the total output of these four fully connected layers to improve the convergence speed of the preset neural network. For example, a 3*sigmoid function can be added, the formula of which is as follows:

[0186]

[0187] In the above formula, μ' and σ' are the inputs of the 3*sigmod function, μ(μ') is the predicted mean information obtained by the 3*sigmod function, and μ(μ') is the predicted variance information obtained by the 3*sigmod function. The 3*sigmod function can compress the overall output of the classification layer to the range of 0 to 3.

[0188] By using a preset number of fully connected layers to perform parameter fitting on the sample correspondence feature information system, the first and second intermediate values ​​can be effectively obtained. The first and second intermediate values ​​are then input into a preset number of S-shaped function processing modules for compression processing, which helps to improve the overall convergence effect of the preset parameter fitting network and obtain the sample predicted mean and sample predicted variance.

[0189] In step S83, the sample time distribution and sample time information are input into the preset prediction information processing module to obtain sample operation prediction information.

[0190] In the embodiments of this specification, the preset prediction information processing module can be used to fit a log-normal distribution. Specifically, in one example, the function formula used in the preset prediction information processing module is as follows:

[0191]

[0192] Where f(μ, σ; x) represents the intensity of the virtual resource interaction operation performed by the sample object on at least one first sample object, x represents the time span of the virtual resource interaction operation performed by the sample object, μ represents the mean in the log-normal distribution, σ represents the variance in the log-normal distribution, exp is an exponential function with the natural constant e as the base, and ln(x) is a logarithmic function with x as the independent variable.

[0193] This embodiment obtains relatively dense sample correspondence feature information by inputting the sample correspondence into a feature extraction network for feature extraction. The relatively dense sample correspondence feature information is then input into a preset parameter fitting network for parameter fitting to obtain the sample predicted mean and sample predicted variance. The sample predicted mean, sample predicted variance, and sample time information are then input into a preset prediction information processing module. Sample operation prediction information can be obtained using the sample correspondence that follows a log-normal distribution, thereby improving the accuracy of the sample operation prediction information.

[0194] In step S73, loss information is determined based on the sample operation prediction information and the operation statistics of the target sample object performing virtual resource interaction operations on at least one first sample object within the target sample time period.

[0195] In the embodiments of this specification, the loss information can be determined by calculating the maximum likelihood loss function corresponding to the sample operation prediction information and using the stochastic gradient descent method. The maximum likelihood loss function characterizes the magnitude of the loss between the sample operation prediction information and the operation statistics of the target sample object performing virtual resource interaction operations on at least one first sample object within the target sample time period.

[0196] In step S74, an initial operation prediction model is trained based on the loss information to obtain a target operation prediction model.

[0197] Specifically, in an exemplary embodiment, the maximum likelihood loss function can be formulated as follows:

[0198]

[0199] Where loss represents loss information, x represents the time span of the sample object performing virtual resource interaction operations, μ represents the mean in the log-normal distribution, σ represents the variance in the log-normal distribution, exp is the exponential function with the natural constant e as the base, and ln(x) is the logarithmic function with x as the independent variable.

[0200] Based on the loss determined by the maximum likelihood loss function formula, the parameters of the preset parameter fitting network can be updated by gradient propagation layer by layer in the reverse preset parameter fitting network, so that the parameters of the preset parameter fitting network gradually converge.

[0201] The above embodiments obtain multiple sample correspondences and sample operation time information for each of the multiple correspondences. The sample correspondences and sample operation time information that follow a log-normal distribution are input into the initial operation prediction model for operation prediction processing to obtain sample prediction information. Based on the sample operation prediction information, loss information is determined, and the initial operation prediction model is trained based on the loss information to obtain the target operation prediction model. In this way, a better operation prediction model can be trained, improving the accuracy of prediction information for virtual resource interaction operations.

[0202] Figure 9 This is a block diagram illustrating a multimedia sequence recommendation apparatus according to an exemplary embodiment. (Refer to...) Figure 9 The device may include:

[0203] The target information acquisition module 901 is configured to acquire target correspondence information and target operation time information of a target object; the target correspondence information stores the correspondence between the target object and at least one first object, the at least one first object being an object to which the target object has performed a virtual resource interaction operation; the virtual resource interaction operation is used for the target object to interact with the at least one first object to access virtual resources; the target operation time information is the operation time information of at least one operation time information that is less than a time threshold from the current time, the at least one operation time information being the operation time information corresponding to the target object performing the virtual resource interaction operation;

[0204] The target information prediction module 902 is configured to input the target correspondence information and the target operation time information into the operation prediction model to perform prediction processing of the virtual resource interaction operation, and obtain target prediction information corresponding to each of the at least one first object; the target prediction information represents the probability that the predicted target object will perform the virtual resource interaction operation on the at least one first object at a target time; the target time is a preset duration after the current time;

[0205] The multimedia sequence determination module 903 is configured to determine a target multimedia sequence matching the target object based on target prediction information corresponding to each of the at least one first object; the target multimedia sequence is at least one multimedia resource of the at least one first object; the at least one multimedia resource is associated with the virtual resource.

[0206] The push module 904 is configured to push the target multimedia sequence to the target object.

[0207] In the above embodiments, using the target correspondence information between the target object and at least one first object, as well as the target operation time information corresponding to the target object, as input to the operation prediction model can improve the correlation between the predicted probability of virtual resource interaction operations and the interaction time of the first object and historical virtual resource operations. The operation prediction model is used to predict the target correspondence information between the target object and at least one first object, as well as the target operation time information corresponding to the target object, to obtain target prediction information corresponding to each of the at least one first object. Combining the target prediction information corresponding to each of the at least one first object, the target multimedia sequence matching the target object is determined, which can improve the correlation between the target multimedia sequence and the virtual resource interaction operation. Pushing the target multimedia sequence with a higher correlation to the virtual resource interaction operation to the target object can optimize the overall performance of the target multimedia sequence in the actual recommendation scenario, improve the virtual resource conversion rate during multimedia sequence recommendation, and thus improve the overall benefit of multimedia sequence recommendation.

[0208] In one possible implementation, the target operation prediction model includes a feature extraction network, a distributed processing network, and a prediction information processing module, wherein the prediction information module includes:

[0209] The first feature extraction unit is configured to input the target correspondence information into the feature extraction network for feature extraction processing to obtain target correspondence feature information;

[0210] The first parameter processing unit is configured to input the target correspondence feature information into the distributed processing network for distributed processing to obtain the time distribution of virtual resource interaction operations; the time distribution of virtual resource interaction operations refers to the time distribution of the target object performing the virtual resource interaction operation on the at least one first object.

[0211] The target prediction information determination unit is configured to use the prediction information processing module to perform prediction processing on the target operation time information and the time distribution to obtain target prediction information corresponding to each of the at least one first object.

[0212] In one possible implementation, the multimedia sequence determination module includes:

[0213] A conversion rate acquisition unit is configured to acquire conversion rate information corresponding to the target object; the conversion rate information represents the proportion of multimedia resources in the total multimedia resources pushed to the target object in which the target object performs virtual resource interaction operations; each multimedia resource in the total multimedia resources is associated with the virtual resource;

[0214] The multimedia sequence determination unit is configured to determine the target multimedia sequence based on the conversion rate information and the target prediction information corresponding to each of the at least one first object.

[0215] In one possible implementation, the multimedia sequence determination module includes:

[0216] The priority information determination unit is configured to determine the priority information of the at least one first object based on the target prediction information corresponding to each of the at least one first object;

[0217] A multimedia resource acquisition unit is configured to acquire the target multimedia resources of the at least one first object;

[0218] The sorting unit is configured to sort the target multimedia resources based on the priority information to obtain the target multimedia sequence.

[0219] In one possible implementation, the device further includes:

[0220] The identification information acquisition module is configured to acquire identification information of a second object that has performed the virtual resource interaction operation; the second object includes the target object;

[0221] The identification information determination module is configured to determine the identification information of at least one third object corresponding to the second object; the third object is an object that the second object has performed the virtual resource interaction operation on, and the at least one third object includes the at least one first object;

[0222] The correspondence information construction module is configured to construct correspondence information between the second object and the third object based on the identification information of the second object and the identification information of the third object.

[0223] The target information acquisition module includes:

[0224] The first information acquisition unit is configured to acquire the identification information of the target object and the target operation time information;

[0225] The target correspondence information acquisition unit is configured to acquire the target correspondence information from the correspondence information between the second object and the third object based on the identification information of the target object.

[0226] In one possible implementation, the target information prediction module includes:

[0227] The parameter fitting unit is configured to input the target correspondence feature information into the parameter fitting network for parameter fitting processing to obtain the target prediction mean and target prediction variance of the time distribution.

[0228] The time distribution determination unit is configured to use the parameter processing module to perform distribution determination processing on the target estimated mean and the target estimated variance to obtain the time distribution of the virtual resource interaction operation.

[0229] Figure 10 This is a block diagram illustrating an operational prediction model training apparatus according to an exemplary embodiment. (Refer to...) Figure 9 The device may include:

[0230] The sample information acquisition module 1001 is configured to acquire sample correspondence relationships and sample operation time information for each of the sample correspondence relationships; the sample correspondence relationship is the correspondence relationship between a target sample object and at least one first sample object, and the sample operation time information is the time sequence information of at least one sample object performing virtual resource interaction operations on at least one first sample object; the at least one first sample object is an object that the target sample object has performed the virtual resource interaction operation on.

[0231] The sample information prediction module 1002 is configured to input the sample correspondence and the sample operation time information into an initial operation prediction model for operation prediction processing to obtain sample prediction information; the sample prediction information represents the probability of predicting that the target sample object will perform the virtual resource interaction operation during the target sample time period; the target sample time period is a time period obtained based on the time series information;

[0232] The loss determination module 1003 is configured to determine loss information based on the sample operation prediction information and the operation statistics of the target sample object performing the virtual resource interaction operation on the at least one first sample object within the target sample time period.

[0233] Training module 1004 is configured to train the initial operation prediction model based on the loss information to obtain the target operation prediction model.

[0234] The above embodiments obtain multiple sample correspondences and sample operation time information for each of the multiple correspondences. The sample correspondences and sample operation time information that follow a log-normal distribution are input into the initial operation prediction model for operation prediction processing to obtain sample prediction information. Based on the sample operation prediction information, loss information is determined, and the initial operation prediction model is trained based on the loss information to obtain the target operation prediction model. In this way, a better operation prediction model can be trained, improving the accuracy of prediction information for virtual resource interaction operations.

[0235] In one possible implementation, the initial operation prediction model includes: a feature extraction network, a preset parameter fitting network, and a preset prediction information processing module; the sample information prediction module includes:

[0236] The second feature extraction unit is configured to input the sample correspondence into the feature extraction network for feature extraction processing to obtain sample correspondence feature information.

[0237] The first sample time distribution determination unit is configured to input the sample correspondence feature information into the preset distribution processing network for distribution processing to obtain the sample time distribution of virtual resource interaction operations; the sample time distribution refers to the time distribution of the sample object performing the virtual resource interaction operation on the at least one first sample object.

[0238] The second information prediction unit is configured to input the sample time distribution and the sample time information into the preset prediction information processing module to obtain the sample operation prediction information.

[0239] In one possible implementation, the preset distribution processing network includes a preset parameter fitting network and a parameter processing module; the first sample time distribution determination unit includes:

[0240] The sample parameter fitting unit is configured to input the sample correspondence feature information into the preset parameter fitting network for parameter fitting processing to obtain the sample estimated mean and sample estimated variance.

[0241] The second sample time distribution determination unit is configured to use the parameter processing module to perform distribution determination processing on the estimated mean and the estimated variance of the samples to obtain the sample time distribution of the virtual resource interaction operation.

[0242] In one possible implementation, the preset parameter fitting network includes a preset number of fully connected layers and the preset number of sigmoid function processing modules; the second parameter processing unit includes:

[0243] The intermediate value determination unit is configured to perform parameter fitting processing on the sample correspondence feature information using a preset number of fully connected layers to obtain a first intermediate value and a second intermediate value.

[0244] The sample prediction value determination unit is configured to input the first intermediate value and the second intermediate value into the preset number of S-shaped function processing modules for compression processing to obtain the sample prediction mean and the sample prediction variance.

[0245] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0246] Figure 11 This is a block diagram illustrating an electronic device for a multimedia sequence recommendation method and / or an operational prediction model training method, according to an exemplary embodiment. The electronic device may be a server, and its internal structure diagram may be as follows: Figure 11 As shown, the electronic device includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage medium. The network interface is used to communicate with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for constructing virtual objects.

[0247] Those skilled in the art will understand that Figure 11The structure shown is merely a block diagram of a portion of the structure related to the present disclosure and does not constitute a limitation on the electronic device to which the present disclosure is applied. A specific electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0248] In an exemplary embodiment, an electronic device is also provided, including: a processor; and a memory for storing processor-executable instructions; wherein the processor is configured to execute the instructions to implement the multimedia sequence recommendation method and / or the operation prediction model training method as described in the embodiments of this disclosure.

[0249] In an exemplary embodiment, a computer-readable storage medium is also provided, which, when executed by a processor of an electronic device, enables the electronic device to perform the multimedia sequence recommendation method and / or operation prediction model training method of the embodiments of this disclosure. The computer-readable storage medium may be a ROM, random access memory (RAM), CD-ROM, magnetic tape, floppy disk, or optical data storage device, etc.

[0250] In an exemplary embodiment, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to perform the multimedia location information processing method of the present disclosure embodiments.

[0251] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. This computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), RAMbus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0252] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the following claims.

[0253] It should be understood that this disclosure is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of this disclosure is limited only by the appended claims.

Claims

1. A multimedia sequence recommendation method, characterized in that, include: Obtain target object correspondence information and target operation time information; The target correspondence information indicates that the target object has performed a virtual resource interaction operation on the virtual resources of at least one first object; the target operation time information is the time information of the target object performing the virtual resource interaction operation on the virtual resources of each first object, and the time since the current time is less than a time threshold; the virtual resource interaction operation refers to the operation of purchasing goods in the multimedia resources of the first object; The target correspondence information and the target operation time information are input into the operation prediction model to perform prediction processing of the virtual resource interaction operation, so as to obtain the target prediction information corresponding to each of the at least one first object. The target prediction information represents the predicted probability that the target object will perform the virtual resource interaction operation on the at least one first object at a target time; The target time is a preset duration after the current time; Based on the target prediction information corresponding to each of the at least one first object, a target multimedia sequence matching the target object is determined; the target multimedia sequence is at least one multimedia resource of the at least one first object. The at least one multimedia resource is associated with a virtual resource; The target multimedia sequence is pushed to the target object.

2. The method according to claim 1, characterized in that, The operation prediction model includes a feature extraction network, a distributed processing network, and a prediction information processing module. The step of inputting the target correspondence information and the target operation time information into the operation prediction model to perform prediction processing for the virtual resource interaction operation, and obtaining the target prediction information corresponding to each of the at least one first object, includes: The target correspondence information is input into the feature extraction network for feature extraction processing to obtain the target correspondence feature information; The target correspondence feature information is input into the distributed processing network for distributed processing to obtain the time distribution of virtual resource interaction operations; the time distribution of virtual resource interaction operations refers to the time distribution of the target object performing the virtual resource interaction operation on the at least one first object. The prediction information processing module is used to perform prediction processing on the target operation time information and the time distribution to obtain the target prediction information corresponding to each of the at least one first object.

3. The method according to claim 1, characterized in that, The step of determining the target multimedia sequence matching the target object based on the target prediction information corresponding to each of the at least one first object includes: Obtain the conversion rate information corresponding to the target object; the conversion rate information represents the proportion of multimedia resources in the total multimedia resources pushed to the target object in which the target object performs virtual resource interaction operations; each multimedia resource in the total multimedia resources is associated with the virtual resource; The target multimedia sequence is determined based on the conversion rate information and the target prediction information corresponding to each of the at least one first object.

4. The method according to claim 1, characterized in that, The step of determining the target multimedia sequence matching the target object based on the target prediction information corresponding to each of the at least one first object includes: Based on the target prediction information corresponding to each of the at least one first object, the priority information of the at least one first object is determined; Acquire the target multimedia resources of the at least one first object; The target multimedia resources are sorted based on the priority information to obtain the target multimedia sequence.

5. The method according to claim 1, characterized in that, The method further includes: Obtain the identification information of a second object that has performed the virtual resource interaction operation; the second object includes the target object; Determine the identification information of at least one third object corresponding to the second object; the third object is an object that the second object has performed the virtual resource interaction operation on, and the at least one third object includes the at least one first object; Based on the identification information of the second object and the identification information of the third object, construct the correspondence information between the second object and the third object; The step of obtaining the target correspondence information between the target object and at least one first object and the target operation time information of the target object includes: Obtain the identification information of the target object and the target operation time information; The target correspondence information is obtained from the correspondence information between the second object and the third object based on the identification information of the target object.

6. The method according to claim 2, characterized in that, The distributed processing network includes a parameter fitting network and a parameter processing module; the step of inputting the target correspondence feature information into the distributed processing network for distributed processing to obtain the time distribution of virtual resource interaction operations includes: The target correspondence feature information is input into the parameter fitting network for parameter fitting processing to obtain the target prediction mean and target prediction variance of the time distribution. The parameter processing module is used to perform distribution determination processing on the estimated mean and variance of the target to obtain the time distribution of the virtual resource interaction operation.

7. A method for training an operational prediction model, characterized in that, The method includes: Obtain sample correspondences and sample operation time information for each of the sample correspondences; the sample correspondences represent that the target sample object has performed virtual resource interaction operations on the virtual resources of at least one first sample object, and the sample operation time information is the time sequence information of the target sample object performing the virtual resource interaction operations on each first sample object; the virtual resource interaction operation refers to the operation of purchasing goods in the multimedia resources of the first sample object. The sample correspondence and the sample operation time information are input into the initial operation prediction model for operation prediction processing to obtain sample prediction information; the sample prediction information represents the probability that the target sample object will perform the virtual resource interaction operation during the target sample time period; the target sample time period is a time period obtained based on the time series information; Based on the sample prediction information and the operation statistics of the target sample object performing the virtual resource interaction operation on the at least one first sample object during the target sample time period, the loss information is determined. The initial operation prediction model is trained based on the loss information to obtain the operation prediction model.

8. The method according to claim 7, characterized in that, The initial operation prediction model includes: a feature extraction network, a preset distribution processing network, and a preset prediction information processing module; the step of inputting the sample correspondence and the sample operation time information into the initial operation prediction model for operation prediction processing to obtain sample prediction information includes: The sample correspondence is input into the feature extraction network for feature extraction processing to obtain sample correspondence feature information; The sample correspondence feature information is input into the preset distribution processing network for distribution processing to obtain the sample time distribution of virtual resource interaction operations; the sample time distribution refers to the time distribution of the sample object performing the virtual resource interaction operation on the at least one first sample object. The sample time distribution and the sample operation time information are input into the preset prediction information processing module to obtain the sample prediction information.

9. The method according to claim 8, characterized in that, The preset distribution processing network includes a preset parameter fitting network and a parameter processing module; the step of inputting the sample correspondence feature information into the preset distribution processing network for distribution processing to obtain the sample time distribution of virtual resource interaction operations includes: The sample correspondence feature information is input into the preset parameter fitting network for parameter fitting processing to obtain the sample predicted mean and sample predicted variance. The parameter processing module is used to perform distribution determination processing on the estimated mean and variance of the samples to obtain the sample time distribution of the virtual resource interaction operation.

10. The method according to claim 9, characterized in that, The preset parameter fitting network includes a preset number of fully connected layers and a preset number of S-shaped function processing modules; The step of inputting the sample correspondence feature information into the preset parameter fitting network for parameter fitting processing to obtain the sample predicted mean and sample predicted variance includes: A preset number of fully connected layers are used to perform parameter fitting on the sample correspondence feature information to obtain a first intermediate value and a second intermediate value. The first intermediate value and the second intermediate value are input into the preset number of S-shaped function processing modules for compression processing to obtain the sample estimated mean and the sample estimated variance.

11. A multimedia sequence recommendation device, characterized in that, include: The target information acquisition module is configured to acquire target correspondence information and target operation time information of the target object; The target correspondence information indicates that the target object has performed a virtual resource interaction operation on the virtual resources of at least one first object; the target operation time information is the time information of the target object performing the virtual resource interaction operation on the virtual resources of each first object, and the time since the current time is less than a time threshold; the virtual resource interaction operation refers to the operation of purchasing goods in the multimedia resources of the first object; The target information prediction module is configured to input the target correspondence information and the target operation time information into the operation prediction model to perform prediction processing of the virtual resource interaction operation, so as to obtain the target prediction information corresponding to each of the at least one first object. The target prediction information represents the predicted probability that the target object will perform the virtual resource interaction operation on the at least one first object at a target time; The target time is a preset duration after the current time; A multimedia sequence determination module is configured to determine a target multimedia sequence matching the target object based on target prediction information corresponding to each of the at least one first object; the target multimedia sequence is at least one multimedia resource of the at least one first object. The at least one multimedia resource is associated with a virtual resource; The push module is configured to push the target multimedia sequence to the target object.

12. The apparatus according to claim 11, characterized in that, The operation prediction model includes a feature extraction network, a distributed processing network, and a prediction information processing module. The information prediction module includes: The first feature extraction unit is configured to input the target correspondence information into the feature extraction network for feature extraction processing to obtain target correspondence feature information; The first parameter processing unit is configured to input the target correspondence feature information into the distributed processing network for distributed processing to obtain the time distribution of virtual resource interaction operations; the time distribution of virtual resource interaction operations refers to the time distribution of the target object performing the virtual resource interaction operation on the at least one first object. The information prediction unit is configured to use the prediction information processing module to perform prediction processing on the target operation time information and the time distribution to obtain target prediction information corresponding to each of the at least one first object.

13. The apparatus according to claim 11, characterized in that, The multimedia sequence determination module includes: A conversion rate acquisition unit is configured to acquire conversion rate information corresponding to the target object; the conversion rate information represents the proportion of multimedia resources in the total multimedia resources pushed to the target object in which the target object performs virtual resource interaction operations; each multimedia resource in the total multimedia resources is associated with the virtual resource; The multimedia sequence determination unit is configured to determine the target multimedia sequence based on the conversion rate information and the target prediction information corresponding to each of the at least one first object.

14. The apparatus according to claim 11, characterized in that, The multimedia sequence determination module includes: The priority information determination unit is configured to determine the priority information of the at least one first object based on the target prediction information corresponding to each of the at least one first object; A multimedia resource acquisition unit is configured to acquire the target multimedia resources of the at least one first object; The sorting unit is configured to sort the target multimedia resources based on the priority information to obtain the target multimedia sequence.

15. The apparatus according to claim 11, characterized in that, The device further includes: The identification information acquisition module is configured to acquire identification information of a second object that has performed the virtual resource interaction operation; the second object includes the target object; The identification information determination module is configured to determine the identification information of at least one third object corresponding to the second object; the third object is an object that the second object has performed the virtual resource interaction operation on, and the at least one third object includes the at least one first object; The correspondence information construction module is configured to construct correspondence information between the second object and the third object based on the identification information of the second object and the identification information of the third object. The target information acquisition module includes: The first information acquisition unit is configured to acquire the identification information of the target object and the target operation time information; The target correspondence information acquisition unit is configured to acquire the target correspondence information from the correspondence information between the second object and the third object based on the identification information of the target object.

16. The apparatus according to claim 12, characterized in that, The first parameter processing unit includes: The parameter fitting unit is configured to input the target correspondence feature information into the parameter fitting network for parameter fitting processing to obtain the target prediction mean and target prediction variance of the time distribution. The time distribution determination unit is configured to use the parameter processing module to perform distribution determination processing on the target estimated mean and the target estimated variance to obtain the time distribution of the virtual resource interaction operation.

17. A training device for an operational prediction model, characterized in that, The device includes: The sample information acquisition module is configured to acquire sample correspondence relationships and sample operation time information for each of the sample correspondence relationships; the sample correspondence relationship indicates that the target sample object has performed a virtual resource interaction operation on the virtual resources of at least one first sample object, and the sample operation time information is the time sequence information of the target sample object performing the virtual resource interaction operation on each first sample object; the virtual resource interaction operation refers to the operation of purchasing goods in the multimedia resources of the first sample object. The sample information prediction module is configured to input the sample correspondence and the sample operation time information into an initial operation prediction model for operation prediction processing to obtain sample prediction information; the sample prediction information represents the probability of the target sample object performing the virtual resource interaction operation during the target sample time period; the target sample time period is a time period obtained based on the time series information; The loss determination module is configured to determine loss information based on the sample prediction information and the operation statistics of the target sample object performing the virtual resource interaction operation on the at least one first sample object within the target sample time period. The training module is configured to train the initial operation prediction model based on the loss information to obtain the operation prediction model.

18. The apparatus according to claim 17, characterized in that, The initial operation prediction model includes: a feature extraction network, a preset distribution processing network, and a preset prediction information processing module; the sample information prediction module includes: The second feature extraction unit is configured to input the sample correspondence into the feature extraction network for feature extraction processing to obtain sample correspondence feature information. The first sample time distribution determination unit is configured to input the sample correspondence feature information into the preset distribution processing network for distribution processing to obtain the sample time distribution of virtual resource interaction operations; the sample time distribution refers to the time distribution of the sample object performing the virtual resource interaction operation on the at least one first sample object. The second information prediction unit is configured to input the sample time distribution and the sample operation time information into the preset prediction information processing module to obtain the sample prediction information.

19. The apparatus according to claim 18, characterized in that, The preset distribution processing network includes a preset parameter fitting network and a parameter processing module; The first sample time distribution determination unit includes: The sample parameter fitting unit is configured to input the sample correspondence feature information into the preset parameter fitting network for parameter fitting processing to obtain the sample estimated mean and sample estimated variance. The second sample time distribution determination unit is configured to use the parameter processing module to perform distribution determination processing on the estimated mean and the estimated variance of the samples to obtain the sample time distribution of the virtual resource interaction operation.

20. The apparatus according to claim 19, characterized in that, The preset parameter fitting network includes a preset number of fully connected layers and a preset number of S-shaped function processing modules; The second parameter processing unit includes: The intermediate value determination unit is configured to perform parameter fitting processing on the sample correspondence feature information using a preset number of fully connected layers to obtain a first intermediate value and a second intermediate value. The sample prediction value determination unit is configured to input the first intermediate value and the second intermediate value into the preset number of S-shaped function processing modules for compression processing to obtain the sample prediction mean and the sample prediction variance.

21. An electronic device, characterized in that, include: processor; Memory used to store the processor's executable instructions; The processor is configured to execute the instructions to implement the multimedia sequence recommendation method or the operation prediction model training method as described in any one of claims 1 to 10.

22. A computer-readable storage medium, characterized in that, When the instructions in the computer-readable storage medium are executed by the processor of the electronic device, the electronic device is able to perform the multimedia sequence recommendation method or the operation prediction model training method as described in any one of claims 1 to 10.

23. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instructions are executed by the processor, they implement the multimedia sequence recommendation method or operation prediction model training method as described in any one of claims 1 to 10.

Images