Object feature processing method, device, electronic device and storage medium

Abstract

The present disclosure provides a method, apparatus, electronic device, and storage medium for processing object features, relating to the field of data processing, and more particularly, to the field of artificial intelligence technology. A specific implementation scheme comprises: obtaining object data of a target object, wherein the object data includes behavioral data and attribute data of the target object, and the target object's activity level is lower than the average activity level within a predetermined object range; and predicting an object feature vector of the target object based on the object data.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, in particular to the field of artificial intelligence technology, and specifically to an object feature processing method, device, electronic device and storage medium. Background Art

[0002] In related technologies, when recommending resources to user objects, the following approach is generally adopted: obtaining the user's resource operation history (e.g., search, click) to obtain the user's vector representation. Then, using the user's vector representation and comparing it with the vector representations of other users, similar users are found. Based on the resource usage of similar users, the current user's preferences are predicted and relevant resources are recommended. However, using this method for resource recommendation requires the current user to have a rich operation history. Otherwise, the vector representation corresponding to the user object is incomplete, and the similar users found based on the vector representation are also inaccurate, resulting in less accurate resource recommendations to the user and, consequently, poor recommendation results. Summary of the Invention

[0003] The present disclosure provides a method, apparatus, device, and storage medium for object feature processing.

[0004] According to one aspect of the present disclosure, an object feature processing method is provided, comprising: acquiring object data of a target object, wherein the object data includes behavior data of the target object and attribute data of the target object, and the activity of the target object is lower than the average activity within a predetermined object range; and predicting an object feature vector of the target object based on the object data.

[0005] Optionally, the above method also includes: determining a target similar object similar to the target object based on the object feature vector and similar feature vectors of similar objects; determining resource content to be pushed based on behavior data of the target similar object; and pushing the resource content to the target object.

[0006] Optionally, based on the object feature vector and the similar feature vectors of similar objects, determining the target similar object similar to the target object includes: obtaining the Euclidean distance between the object feature vector and the similar feature vector of the similar object; and selecting the similar object whose Euclidean distance is less than a predetermined distance threshold among the similar objects as the target similar object.

[0007] Optionally, based on the object data, predicting the object feature vector of the target object includes: based on the object data, using a meta-learning network model to predict the object feature vector of the target object, wherein the meta-learning network model is trained based on multiple groups of sample object data, the multiple groups of sample object data include the object data of the sample object and the feature vector of the sample object, and the activity of the sample object is higher than the average activity.

[0008] Optionally, the above method also includes: constructing an object resource graph, wherein the object resource graph is generated based on the travel routes of multiple objects, and the object resource graph includes: multiple object nodes, and resource nodes that have one or more layers of association relationships with the multiple object nodes; screening out the head node and long tail node for the object from the object resource graph, wherein the number of neighbor nodes of the head node is greater than or equal to a predetermined threshold, and the number of neighbor nodes of the long tail node is less than a predetermined threshold; and taking the objects corresponding to some or all of the nodes in the head node as sample objects.

[0009] Optionally, the object data of the sample object includes: behavior data of the sample object and attribute data of the sample object.

[0010] According to another aspect of the present disclosure, an object feature processing device is provided, including: an acquisition module for acquiring object data of a target object, wherein the object data includes behavior data of the target object and attribute data of the target object, and the activity of the target object is lower than the average activity within a predetermined object range; and a prediction module for predicting an object feature vector of the target object based on the object data.

[0011] Optionally, the above-mentioned device also includes: a first determination module, used to determine the target similar object similar to the target object based on the object feature vector and the similar feature vector of the similar object; a second determination module, used to determine the resource content to be pushed based on the behavior data of the target similar object; and a push module, used to push the resource content to the target object.

[0012] Optionally, the first determination module includes: an acquisition unit for acquiring the Euclidean distance between the object feature vector and the similar feature vectors of similar objects; and a selection unit for selecting similar objects whose Euclidean distance is less than a predetermined distance threshold as target similar objects.

[0013] Optionally, the above-mentioned prediction module includes: a prediction unit, used to predict the object feature vector of the target object based on the object data using a meta-learning network model, wherein the meta-learning network model is trained based on multiple groups of sample object data, and the multiple groups of sample object data include the object data of the sample object and the feature vector of the sample object, and the activity of the sample object is higher than the average activity.

[0014] Optionally, the above-mentioned prediction module also includes: a construction unit for constructing an object resource graph, wherein the object resource graph is generated based on the wandering routes of multiple objects, and the object resource graph includes: multiple object nodes, and resource nodes that have one or more layers of association relationships with the multiple object nodes; a screening unit for screening out the head node and long-tail node for the object from the object resource graph, wherein the number of neighbor nodes of the head node is greater than or equal to a predetermined threshold, and the number of neighbor nodes of the long-tail node is less than a predetermined threshold; a processing unit for taking the objects corresponding to some or all of the nodes in the head node as sample objects.

[0015] Optionally, the object data of the sample object includes: behavior data of the sample object and attribute data of the sample object.

[0016] According to another aspect of the present disclosure, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute any one of the above methods.

[0017] According to yet another aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute any one of the above methods.

[0018] According to another aspect of the present disclosure, a computer program product is provided, comprising: a computer program, wherein the computer program implements any one of the above methods when executed by a processor.

[0019] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present disclosure, nor are they intended to limit the scope of the present disclosure. Other features of the present disclosure will become readily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings are provided to facilitate a better understanding of the present invention and do not constitute a limitation of the present disclosure.

[0021] Figure 1 is a flowchart of an object feature processing method provided according to an embodiment of the present disclosure;

[0022] Figure 2 is a schematic diagram of an object feature processing method provided according to an optional embodiment of the present disclosure;

[0023] Figure 3 is a schematic diagram of the first-order neighbors of the head node u;

[0024] Figure 4 It is a schematic diagram of the regression model structure;

[0025] Figure 5 is a structural block diagram of an object feature processing device provided according to an embodiment of the present disclosure;

[0026] Figure 6 It is a block diagram of an electronic device used to implement the object feature processing method according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0027] The following description of exemplary embodiments of the present disclosure is made in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding. These details should be considered as merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0028] Terminology

[0029] Meta-learning is the process of learning how to learn, rather than learning a specific task. When faced with a new task, an algorithm, armed with prior knowledge, can quickly learn using only a small amount of data.

[0030] u2u, User to User, emphasizes the establishment of connections and associations between users on social platforms.

[0031] User-based Collaborative Filtering (UCF) is an algorithm that makes recommendations based on a group of users with the same interests. It usually requires three steps, including: collecting information that can represent user interests, nearest neighbor search, and generating recommendation results.

[0032] Topk, the top k objects in a set of sorts.

[0033] Dropout, when training a large neural network, randomly excludes some neurons from the current training to prevent the model from learning too much noise and causing overfitting. That is, after the model is trained to a certain level, the test error obtained on the training set is much larger than the error obtained on the test set.

[0034] Fully connected (FC) network structure, a basic neural network or deep neural network layer, each node in the fully connected layer is connected to all nodes in the previous layer.

[0035] In an embodiment of the present disclosure, a method for processing object features is provided. Figure 1 is a flow chart of an object feature processing method provided according to an embodiment of the present disclosure, such as Figure 1As shown, the flowchart includes the following steps:

[0036] Step S102: acquiring object data of a target object, wherein the object data includes behavior data and attribute data of the target object, and the activity level of the target object is lower than the average activity level within a predetermined object range;

[0037] Step S104: predicting the object feature vector of the target object based on the object data.

[0038] By obtaining the target object's behavioral data and attribute data through the above steps, the characteristics of low-activity target objects can be more comprehensively obtained. Moreover, when predicting the target object, the object data used includes not only the object's behavioral data but also the object's attribute data. Therefore, the vector representation of low-activity objects (objects with activity levels lower than the average activity level within a predetermined object range) can be made more comprehensive, thereby enabling more accurate recommendations for these low-activity objects, solving the problem of inaccurate and poor recommendation results for low-activity objects.

[0039] As an optional embodiment, after obtaining the predicted result of the target object's object feature vector, various operations can be performed. For example, the following operations can be taken: based on the object feature vector and similar feature vectors of similar objects, target similar objects similar to the target object are determined; based on the behavior data of the target similar objects, resource content to be pushed is determined; and resource content is pushed to the target object. In this way, by using the similarity of feature vectors to obtain high-activity users similar to low-activity users, and then recommending to low-activity users based on the behavior data of high-activity users, accurate recommendations can be made for low-activity users, thereby increasing their interest in usage.

[0040] As an optional embodiment, when determining a target similar object similar to the target object based on the object feature vector and the similar feature vector of the similar object, a variety of methods can be used. For example, the following method can be used: obtaining the Euclidean distance between the object feature vector and the similar feature vector of the similar object; selecting similar objects whose Euclidean distance is less than a predetermined distance threshold among the similar objects as target similar objects. By obtaining the Euclidean distance between feature vectors and screening out users whose Euclidean distance is less than a predetermined threshold as similar users of the target user, the Euclidean distance between feature vectors can quantitatively and accurately determine the similarity between users. Therefore, high-activity users similar to low-activity users can be accurately found, thereby recommending low-activity users based on the behavioral data of similar high-activity users, improving the accuracy of recommendations, and optimizing the effect of recommendations. It should be noted that using Euclidean distance to characterize the similarity between users is only an optional implementation method. In other embodiments of the present application, other methods of representing the similarity between users can also be used, such as cosine distance values.

[0041] As an optional embodiment, when predicting the object feature vector of the target object based on the object data, a variety of prediction methods can be used. For example, it can be implemented based on a prediction method of an artificial intelligence network model. For example, a meta-learning network model can be used to predict the object feature vector of the target object based on the object data, wherein the meta-learning network model is trained based on multiple groups of sample object data, and the multiple groups of sample object data include object data of the sample objects and feature vectors of the sample objects, and the activity of the sample objects is higher than the average activity. Since the meta-learning network model is trained with object data of sample objects with higher activity than the average activity and feature vectors of the sample objects, the model can learn richer features more efficiently. Using such a model to predict the feature vectors of low-activity users will produce more accurate results.

[0042] As an optional embodiment, multiple methods can be used when determining sample objects. For example, the following methods can be used: constructing an object resource graph, wherein the object resource graph is generated based on the travel routes of multiple objects, and the object resource graph includes: multiple object nodes, and resource nodes that have one or more layers of association with the multiple object nodes; filtering out the head node and long tail nodes for the object from the object resource graph, wherein the number of neighbor nodes of the head node is greater than or equal to a predetermined threshold, and the number of neighbor nodes of the long tail node is less than the predetermined threshold; and using the objects corresponding to some or all of the nodes in the head node as sample objects. Determining sample objects by the above method can ensure that all sample objects are highly active users, provide more behavioral data for the subsequent model learning process, thereby ensuring high efficiency and high accuracy of model learning, and further improve the accuracy of the prediction results of feature vectors for low-activity users. In addition, by constructing the object resource graph, the richness of the features corresponding to each object can be more systematically and comprehensively characterized. Moreover, based on the object resource graph, the head node and long tail nodes for the object are filtered out, wherein the number of neighbor nodes of the head node is greater than or equal to the predetermined threshold, and the number of neighbor nodes of the long tail node is less than the predetermined threshold. Based on the above-mentioned method of determining the head node and long-tail node of the object, the user's characteristics can be determined in a relatively quantitative and standard manner to a certain extent, so that the samples used to train the meta-learning network model are relatively accurate and efficient.

[0043] As an optional embodiment, the object data of the sample objects may include multiple types of data, such as: behavioral data of the sample objects, such as user clicks, favorites, forwarding, and comments on resources; and attribute data of the sample objects, such as the user's gender, age, education level, and consumption level. This data can be used to comprehensively reflect the user's interests, preferences, and attribute characteristics, and thus efficiently and comprehensively train the prediction model using this data, making the model's prediction results more accurate and facilitating the identification of highly active users similar to low-activity users, thereby completing recommendations. This greatly improves the accuracy of recommendations for low-activity users and enhances the effectiveness of recommendations.

[0044] Based on the above embodiments and optional embodiments, an optional implementation is provided, which is described below.

[0045] For users with low activity levels, whether obtaining a vector representation of the user or characterizing the user based on their demographic attributes, the characterization of low-activity users is inaccurate and not suitable for application scenarios with a large number of low-activity users. This leads to inaccurate recommendations and poor recommendation results.

[0046] In the user collaborative filtering recall method of the related art (for example, recommending novels to users), a user click sequence or a walk-based graph model (i.e., the object resource graph mentioned above) is used to construct a vector representation of the user, and the cosine distance is calculated as the user similarity to obtain similar users, and resources are recalled based on the reading lists of similar users. However, this method works well for the vector representation of users with richer behaviors, but it works poorly for the vector representation of low-activity users with sparse behaviors (the offline evaluation u2u (User to User, abbreviated as u2u) correlation is weak, and when UCF (User-based Collaborative Filtering, abbreviated as UCF) is directly used for online recall, the experiment has no obvious benefits. In the novel vertical recommendation scenario, there are a large number of novel users whose behaviors are relatively sparse (about 83% of novel users have not downloaded more than 5 novel resources in the past month). Therefore, the user collaborative filtering recall in the related art cannot meet the resource recall needs of a large number of low-activity users in the novel scenario.

[0047] In the implementation of the above method, based on collaborative filtering recall, users' click preferences for items or resources are used to construct a user's historical reading list. A user vector representation is obtained from this historical reading list, and similarity between users is calculated using this vector. Based on this similarity, the top k nearest neighbor users are found. Based on the similarity weights of these neighboring users and their preferences for items or resources, the current user's favorite items or resources are predicted, and a ranked list of items is calculated for recommendation. This method is the most basic user recall method and is only suitable for recommendation scenarios with a large number of users with rich behaviors. It is not suitable for recommendation scenarios in vertical categories such as novels, where there are many users with sparse behaviors.

[0048] For example, for users with sparse behavior, a user-item graph model (the object resource graph mentioned above) is often introduced. A vector representation of the user is obtained based on composition walks, and then the user-based collaborative filtering recall method is used to recommend items or resources. However, in practice, this method has been found to be inaccurate for users with sparse behavior, leading to biased user resource recall and making it unsuitable for novel recommendations.

[0049] However, for users with sparse behavior, clustering and recall can also be performed based on their demographic attributes. However, clustering and recalling based solely on their demographic attributes provides a relatively rough description of user behavior, and a large number of sparse users will affect the clustering effect.

[0050] Based on the above situation, in an optional embodiment of the present disclosure, an object feature processing solution is provided. Figure 2is a flow chart of an object feature processing method provided according to an optional embodiment of the present disclosure, such as Figure 2 As shown, the process includes the following processing:

[0051] (1) Using novel reading history to construct a user-resource isomorphism graph (same as the object-resource graph mentioned above, here taking users as the object as an example), selecting users and resources with high node degrees as head nodes, and offline training of the meta-learning network model can include the following steps:

[0052] Building a basic model: Using novel users' reading history as training samples, we constructed a map walk and used the word2vec algorithm to obtain vector representations of users and resources as the basic model.

[0053] Filter head nodes: Based on the above original samples, build a user-resource isomorphism graph, filter head nodes (number of neighbors greater than 5, including users and resources), and obtain the head node vector produced by the above model;

[0054] Obtaining user demographic attributes: To address the problem of sparse user vector representation, we adopt the idea of ​​knowledge distillation and introduce user demographic attributes (gender, age, education level, and consumption level).

[0055] Construct a meta-learning task: randomly sample the first-order neighbors of the head node and perform dropout (no more than 5 neighbors are sampled for each node), such as Figure 3 As shown, Figure 3 This is a schematic diagram of the first-order neighbor nodes of the head node u. The meta-learning task is constructed, including constructing a test set with the node itself and a training set with the node's neighbors. The model learns how to use the node's neighbors to represent the node's own knowledge (for example, it can be represented by a 3-layer fully connected (FC) network [512, 128, 32]). In this optional implementation, the main purpose is to optimize the feature vectors of long-tail users. Therefore, only the user head node is taken from the training set and the test set, and all the neighbor nodes used in the training process are head nodes. The Euclidean distance between each node output vector and the basic model vector in the test set and the training set is summed, as shown in Figure 2. Figure 4 The training samples are constructed as shown, including the input feature of the node's first-order neighbor node aggregation vector (average pooling) spliced ​​with the user's demographic attribute features. The model fitting target is the vector output by the above model, and the loss function can be modulo, Figure 4 This is a schematic diagram of the structure of the meta-learning network model in an optional embodiment of the present disclosure;

[0056] Network parameter update: Iterative process meta-learning: In each training session, a sub-network is constructed based on the meta-network parameters. The loss of the training nodes is first calculated and the gradient is returned to update the sub-network parameters. Then, the test loss is calculated using the updated sub-network. This process is repeated k times (k=5). Finally, the test loss is gradient-optimized and the meta-network parameters are updated.

[0057] (2) Using the meta-learning network model trained in (1), offline prediction of long-tail user node vector representations can be performed using the following steps:

[0058] Filter long-tail user nodes: Filter out long-tail node users from the above user-resource isomorphism graph (for example, the total number of clicks on novel resources within 30 days does not exceed 5);

[0059] Construct a meta-test task: Obtain a vector representation of a long-tail user node: Construct a meta-test task in a similar manner to the training set. The test set is the long-tail user node to be predicted, and the training set is the node's neighbors. The calculation process is similar to that during training, and finally the vector representation of the user node is obtained.

[0060] (3) To construct a meta-UCF recall pathway online, the following steps can be taken:

[0061] Merge the updated long-tail node vector with the user vector generated by the original model, and select users with a click history of >5 as core users to construct a u2u similarity matrix;

[0062] New online meta-collaboration (metaUCF) recall: Each user recalls the top 50 most similar users, and uses a voting algorithm to recall novel resources clicked by similar users.

[0063] The above-mentioned implementation method provides an effective method for low-activity user vector representation and user recommendation for vertical scenarios with a large number of low-activity users. This method combines meta-learning and recommendation business scenarios, proving the feasibility of transfer learning in recommendation scenarios.

[0064] In the embodiment of the present disclosure, an object feature processing device is also provided. Figure 5 is a structural block diagram of an object feature processing device provided according to an embodiment of the present disclosure, such as Figure 5 As shown, the device includes: an acquisition module 51 and a prediction module 52. The device is described below.

[0065] An acquisition module 51 is used to acquire object data of a target object, wherein the object data includes behavioral data and attribute data of the target object, and the activity of the target object is lower than the average activity within a predetermined object range; a prediction module 52 is connected to the acquisition module 51 and is used to predict an object feature vector of the target object based on the object data.

[0066] As an optional embodiment, the above device further includes: a first determination module, a second determination module and a push module. The device is described below.

[0067] The first determination module is connected to the above-mentioned prediction module and is used to determine the target similar object similar to the target object based on the object feature vector and the similar feature vector of the similar object; the second determination module is connected to the above-mentioned first determination module and is used to determine the resource content to be pushed based on the behavior data of the target similar object; the push module is connected to the above-mentioned second determination module and is used to push the resource content to the target object.

[0068] As an optional embodiment, the first determination module includes: an acquisition unit, used to obtain the Euclidean distance between the object feature vector and the similar feature vector of the similar object; and a selection unit, used to select similar objects whose Euclidean distance is less than a predetermined distance threshold among the similar objects as target similar objects.

[0069] As an optional embodiment, the above-mentioned prediction module may include: a prediction unit, used to predict the object feature vector of the target object based on the object data using a meta-learning network model, wherein the meta-learning network model is trained based on multiple groups of sample object data, and the multiple groups of sample object data include the object data of the sample object and the feature vector of the sample object, and the activity of the sample object is higher than the average activity.

[0070] As an optional embodiment, the above prediction module may further include: a construction unit, a screening unit and a processing unit. The prediction module is described below.

[0071] A construction unit is used to construct an object resource graph, wherein the object resource graph is generated based on the wandering routes of multiple objects, and the object resource graph includes: multiple object nodes, and resource nodes that have one or more layers of association relationships with the multiple object nodes; a screening unit is connected to the above-mentioned construction unit, and is used to screen out head nodes and long-tail nodes for objects from the object resource graph, wherein the number of neighbor nodes of the head node is greater than or equal to a predetermined threshold, and the number of neighbor nodes of the long-tail node is less than a predetermined threshold; a processing unit is connected to the above-mentioned screening unit, and is used to use objects corresponding to some or all of the nodes in the head node as sample objects.

[0072] As an optional embodiment, the object data of the sample object includes: behavior data of the sample object and attribute data of the sample object.

[0073] In the technical solutions disclosed herein, the acquisition, storage, and application of user personal information involved comply with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0074] According to an embodiment of the present disclosure, the present disclosure also provides an electronic device, a readable storage medium, and a computer program product.

[0075] Figure 6 A schematic block diagram of an example electronic device 600 that can be used to implement embodiments of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital assistants, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are provided as examples only and are not intended to limit the implementation of the present disclosure described and / or claimed herein.

[0076] like Figure 6 As shown, the electronic device (or device 600) includes a computing unit 601, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 602 or a computer program loaded from a storage unit 608 into a random access memory (RAM) 603. Various programs and data required for the operation of the device 600 can also be stored in the RAM 603. The computing unit 601, ROM 602, and RAM 603 are connected to each other via a bus 604. An input / output (I / O) interface 605 is also connected to the bus 604.

[0077] Various components in device 600 are connected to I / O interface 605, including an input unit 606, such as a keyboard, mouse, etc.; an output unit 607, such as various types of displays, speakers, etc.; a storage unit 608, such as a magnetic disk, optical disk, etc.; and a communication unit 609, such as a network card, modem, wireless communication transceiver, etc. The communication unit 609 allows device 600 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0078] The computing unit 601 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of the computing unit 601 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 601 performs the various methods and processes described above, such as the object feature processing method. For example, in some embodiments, the object feature processing method can be implemented as a computer software program tangibly embodied in a machine-readable medium, such as the storage unit 608. In some embodiments, part or all of the computer program can be loaded and / or installed onto the device 600 via the ROM 602 and / or the communication unit 609. When the computer program is loaded into the RAM 603 and executed by the computing unit 601, one or more steps of the object feature processing method described above can be performed. Alternatively, in other embodiments, the computing unit 601 can be configured to perform the object feature processing method by any other suitable means (e.g., via firmware).

[0079] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0080] The program code for implementing the method of the present disclosure can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device so that when the program code is executed by the processor or controller, the functions / operations specified in the flow chart and / or block diagram are implemented. The program code can be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0081] In the context of the present disclosure, a machine-readable medium can be a tangible medium that can contain or store a program for use by or in conjunction with an instruction execution system, device or equipment. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium can include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0082] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0083] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer having a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network (LAN), a wide area network (WAN), and the Internet.

[0084] A computer system may include a client and a server. The client and server are generally remote from each other and typically interact through a communication network. The client-server relationship arises through computer programs running on the respective computers and having a client-server relationship with each other. The server may be a cloud server, a server in a distributed system, or a server integrated with a blockchain.

[0085] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in this disclosure can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved. This is not a limitation herein.

[0086] The above specific embodiments do not constitute a limitation on the scope of protection of this disclosure. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this disclosure shall be included within the scope of protection of this disclosure.

Claims

1. A method for processing object features, comprising: Acquiring object data of a target object, wherein the object data includes behavior data of the target object and attribute data of the target object, and the activity level of the target object is lower than an average activity level within a predetermined object range; Predicting an object feature vector of the target object based on the object data; Among them, predicting the object feature vector of the target object based on the object data includes: based on the object data, using a meta-learning network model to predict the object feature vector of the target object, wherein the meta-learning network model is trained based on multiple groups of sample object data, and the multiple groups of sample object data include the object data of the sample object and the feature vector of the sample object, and the activity of the sample object is higher than the average activity.

2. The method according to claim 1, wherein The method further comprises: Determining a target similar object similar to the target object based on the object feature vector and similar feature vectors of similar objects; Determining resource content to be pushed based on the behavior data of the target similar object; Push the resource content to the target object.

3. The method according to claim 2, wherein: The determining of a target similar object similar to the target object based on the object feature vector and similar feature vectors of similar objects includes: Obtaining the Euclidean distance between the feature vector of the object and similar feature vectors of similar objects; A similar object whose Euclidean distance is less than a predetermined distance threshold among the similar objects is selected as the target similar object.

4. The method according to claim 1, wherein The method further comprises: Constructing an object resource graph, wherein the object resource graph is generated based on the wandering routes of the plurality of objects, and the object resource graph includes: a plurality of object nodes, and resource nodes having one or more layers of association relationships with the plurality of object nodes; Filtering a head node and a long-tail node for an object from the object resource graph, wherein the number of neighbor nodes of the head node is greater than or equal to a predetermined threshold, and the number of neighbor nodes of the long-tail node is less than the predetermined threshold; The objects corresponding to some or all of the nodes in the head node are used as the sample objects.

5. The method according to any one of claims 1 to 4, wherein The object data of the sample object includes: behavior data of the sample object and attribute data of the sample object.

6. An object feature processing device, comprising: an acquisition module, configured to acquire object data of a target object, wherein the object data includes behavior data of the target object and attribute data of the target object, and the activity level of the target object is lower than an average activity level within a predetermined object range; A prediction module, configured to predict an object feature vector of the target object based on the object data; In which, the prediction module includes: a prediction unit, used to predict the object feature vector of the target object based on the object data using a meta-learning network model, wherein the meta-learning network model is trained based on multiple groups of sample object data, and the multiple groups of sample object data include the object data of the sample object and the feature vector of the sample object, and the activity of the sample object is higher than the average activity.

7. The device according to claim 6, wherein The device further comprises: A first determining module is configured to determine a target similar object similar to the target object based on the object feature vector and similar feature vectors of similar objects; A second determining module is used to determine the resource content to be pushed based on the behavior data of the target similar object; A push module is used to push the resource content to the target object.

8. The device according to claim 7, wherein The first determining module includes: an acquiring unit, configured to acquire a Euclidean distance between the feature vector of the object and similar feature vectors of similar objects; A selection unit is configured to select a similar object whose Euclidean distance is less than a predetermined distance threshold among the similar objects as the target similar object.

9. The device according to claim 6, wherein The prediction module also includes: A construction unit is configured to construct an object resource graph, wherein the object resource graph is generated based on the wandering routes of multiple objects, and the object resource graph includes: multiple object nodes, and resource nodes having one or more layers of association relationships with the multiple object nodes; a screening unit, configured to screen out a head node and a long-tail node for an object from the object resource graph, wherein the number of neighbor nodes of the head node is greater than or equal to a predetermined threshold, and the number of neighbor nodes of the long-tail node is less than the predetermined threshold; The processing unit is configured to use objects corresponding to some or all of the nodes in the head node as the sample objects.

10. The device according to any one of claims 6 to 9, wherein The object data of the sample object includes: behavior data of the sample object and attribute data of the sample object.

11. An electronic device comprising: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 5.

12. A non-transitory computer-readable storage medium storing computer instructions, wherein: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 5.

13. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1 to 5.

Images