CN111382190A

CN111382190A - Object recommendation method and device based on intelligence and storage medium

Info

Publication number: CN111382190A
Application number: CN202010144737.6A
Authority: CN
Inventors: 陈思宏; 牟帅; 肖万鹏; 鞠奇
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-03-04
Filing date: 2020-03-04
Publication date: 2020-07-07
Anticipated expiration: 2040-03-04
Also published as: CN111382190B

Abstract

The embodiment of the invention discloses an object recommendation method, an object recommendation device and a storage medium based on intelligence; according to the embodiment of the invention, the object relation network can be generated according to the operation information of the user aiming at the object; according to a preset relationship mining model, extracting local association features of objects in an object relationship network to obtain initial feature information of the objects; according to the initial characteristic information of the object, carrying out global distribution characteristic extraction on the object in the object relation network to obtain global distribution characteristic information of the object; updating parameter information of a preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model; performing feature extraction on the object relationship network according to the target relationship mining model to obtain target feature information of the object; and recommending the object according to the target characteristic information. The scheme can more comprehensively and accurately mine the incidence relation between the objects, so that the objects can be more accurately recommended to the user.

Description

Object recommendation method and device based on intelligence and storage medium

Technical Field

The invention relates to the technical field of artificial intelligence, in particular to an object recommendation method and device based on intelligence and a storage medium.

Background

In recent years, with the popularization of networks, users are more and more involved in network activities, and therefore, the amount of operation information of users on different objects on the internet, such as operation information of browsing, commenting, sharing and the like on a certain commodity or song or a user, is increasing. At present, the operation information of the user for the object becomes more and more valuable, and the system can be helped to more comprehensively know the association relation between the objects by analyzing the operation information of the user for the object, so that the associated object is recommended to the user based on the operation of the user, and the service is better provided for the user.

At present, generally, an object relationship network can be generated according to the operation of a user on an object, and a trained relationship mining model can be adopted to mine the association relationship between objects, but the existing relationship mining model mines the relationship between the object and other objects from the local range of the object according to the object relationship network, and lacks the global distribution characteristics of the object in the object relationship network, so that the characteristics between the objects cannot be mined accurately and comprehensively, and in addition, because the training of the relationship mining model is black-boxed, the anti-noise capability in the relationship mining process is poor.

Disclosure of Invention

The embodiment of the invention provides an object recommendation method, an object recommendation device and a storage medium based on intelligence, which can more comprehensively and accurately mine the association relationship between objects, so that the objects can be more accurately recommended to users.

The embodiment of the invention provides an intelligent-based object recommendation method, which comprises the following steps:

generating an object relation network between objects according to operation information of a user aiming at the objects;

according to a preset relationship mining model, extracting local association features of objects in the object relationship network to obtain initial feature information of the objects;

according to the initial characteristic information of the object, carrying out global distribution characteristic extraction on the object in the object relation network to obtain global distribution characteristic information of the object;

updating the parameter information of the preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model;

extracting the characteristics of the object relationship network according to the target relationship mining model to obtain target characteristic information of the object;

and recommending the object according to the target characteristic information.

Correspondingly, an embodiment of the present invention further provides an object recommendation apparatus, including:

the generating unit is used for generating an object relation network between the objects according to the operation information of the user aiming at the objects;

the local feature extraction unit is used for extracting local associated features of the objects in the object relationship network according to a preset relationship mining model to obtain initial feature information of the objects;

the global distribution characteristic information extraction unit is used for carrying out global distribution characteristic extraction on the objects in the object relationship network according to the initial characteristic information of the objects to obtain the global distribution characteristic information of the objects;

the updating unit is used for updating the parameter information of the preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model;

the target characteristic information extraction unit is used for extracting the characteristics of the object relationship network according to the target relationship mining model to obtain the target characteristic information of the object;

and the recommending unit is used for recommending the object according to the target characteristic information.

Preferably, the updating unit includes a classifying subunit, an updating subunit and a determining subunit, and includes:

the classification subunit is configured to perform full-connection operation on the initial feature information according to a preset classification branch network to obtain class prediction information corresponding to the initial feature information;

the updating subunit is configured to update parameter information in the preset relationship mining model according to a preset loss function, the category prediction information, and the global distribution feature information, so as to obtain a current relationship mining model;

and the determining subunit is configured to determine whether the current relationship mining model is the target relationship mining model according to the global distribution feature information.

In addition, the embodiment of the present invention further provides a storage medium, where a plurality of instructions are stored, where the instructions are suitable for being loaded by a processor to perform the steps in any one of the intelligent object recommendation methods provided in the embodiment of the present invention.

According to the embodiment of the invention, an object relation network between the objects can be generated according to the operation information of the user aiming at the objects; according to a preset relationship mining model, extracting local association features of objects in the object relationship network to obtain initial feature information of the objects; according to the initial characteristic information of the object, carrying out global distribution characteristic extraction on the object in the object relation network to obtain global distribution characteristic information of the object; updating the parameter information of the preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model; extracting the characteristics of the object relationship network according to the target relationship mining model to obtain target characteristic information of the object; and recommending the object according to the target characteristic information. According to the scheme, the global distribution characteristic information of the objects in the object relation network is extracted, and the preset relation mining model is trained according to the global distribution characteristic information, so that the association relation between the objects can be mined more comprehensively and accurately, the objects can be recommended to users more accurately, and the anti-noise performance of the relation mining model during use can be improved.

Drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the description of the embodiments will be briefly introduced below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings based on these drawings without creative efforts.

FIG. 1 is a scene diagram of an intelligent object recommendation method according to an embodiment of the present invention;

fig. 2a is a schematic flowchart of object relationship mining in the intelligent-based object recommendation method according to the embodiment of the present invention;

fig. 2b is a schematic flowchart of object recommendation performed in the intelligent-based object recommendation method according to the embodiment of the present invention; (ii) a

FIG. 3 is a schematic structural diagram of an object recommendation apparatus according to an embodiment of the present invention;

fig. 4 is a schematic structural diagram of a network device according to an embodiment of the present invention;

fig. 5 is a schematic diagram of updating a preset relationship mining model according to an embodiment of the present invention.

Fig. 6a is a schematic structural diagram of a data sharing system when an object relationship mining apparatus provided in an embodiment of the present invention is used as a node of the data sharing system;

FIG. 6b is a block chain and block structure diagram of the data sharing system shown in FIG. 6 a;

fig. 6c is a block generation flow diagram of the block chain shown in fig. 6 b.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

The embodiment of the invention provides an object recommendation method, an object recommendation device and a storage medium based on intelligence.

The object recommendation method in the embodiment of the invention relates to an Artificial Intelligence technology, wherein Artificial Intelligence (AI) is a theory, a method, a technology and an application system which simulate, extend and expand human Intelligence by using a digital computer or a machine controlled by the digital computer, sense the environment, acquire knowledge and use the knowledge to obtain the best result. In other words, artificial intelligence is a comprehensive technique of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence is the research of the design principle and the realization method of various intelligent machines, so that the machines have the functions of perception, reasoning and decision making. The solution provided by the embodiment of the present invention relates to technologies such as Computer Vision (CV) and Machine Learning (ML) of artificial intelligence, and will be specifically described with reference to the following embodiments.

The object recommendation method of the invention also relates to Cloud technology, wherein Cloud technology refers to a hosting technology for unifying series resources such as hardware, software, network and the like in a wide area network or a local area network to realize calculation, storage, processing and sharing of data. The cloud technology is a general term of network technology, information technology, integration technology, management platform technology, application technology and the like applied based on a cloud computing business model, can form a resource pool, is used as required, and is flexible and convenient. Cloud computing technology will become an important support. Background services of the technical network system require a large amount of computing and storage resources, such as video websites, picture-like websites and more web portals. With the high development and application of the internet industry, each article may have its own identification mark and needs to be transmitted to a background system for logic processing, data in different levels are processed separately, and various industrial data need strong system background support and can only be realized through cloud computing. The scheme provided by the embodiment of the invention relates to the basic technologies of cloud computing, cloud storage, cloud internet of things and the like of the cloud technology, and the following embodiment is specifically used for explanation.

The object recommendation device in the present invention may be specifically integrated in a network device, such as a terminal or a server, and the object recommendation device in the present invention may include an object relationship mining device and a recommendation device. In addition, the object relationship mining device and the recommending device can be integrated in the same network equipment, and can also be integrated in two network equipment connected through a network.

In some embodiments, referring to fig. 6a, the terminal and the server may be one node in a data sharing system, where the data sharing system refers to a system for performing data sharing between nodes, the data sharing system may include a plurality of nodes, and the plurality of nodes may refer to respective network devices in the data sharing system. Each node stores one same block chain, and the object relationship mining device can store the target characteristic information of the object into the block chain, so that the data sharing is performed with other network devices, for example, the target characteristic information of the object can be shared to the network device for object recommendation through the block chain.

For example, referring to fig. 1, an object relationship network between objects may be generated according to operation information of a user for the objects; according to a preset relationship mining model, extracting local association features of objects in the object relationship network to obtain initial feature information of the objects; according to the initial characteristic information of the object, carrying out global distribution characteristic extraction on the object in the object relation network to obtain global distribution characteristic information of the object; updating the parameter information of the preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model; extracting the characteristics of the object relationship network according to the target relationship mining model to obtain target characteristic information of the object; and recommending the object according to the target characteristic information.

According to the scheme, the global distribution characteristic information of the objects in the object relation network is extracted, and the preset relation mining model is trained according to the global distribution characteristic information, so that the association relation between the objects can be mined more comprehensively and accurately, the objects can be recommended to users more accurately, and the anti-noise performance of the relation mining model during use can be improved.

The following are detailed below. The order of the following examples is not intended to limit the preferred order of the examples.

The first embodiment,

In this embodiment, description will be made from the perspective of an object relationship mining apparatus, which may be specifically integrated in a network device, such as a terminal or a server. In the following embodiments, the object relationship mining apparatus is integrated in a server, and a detailed description will be given.

The embodiment of the invention provides an object relation mining method, which comprises the following steps: generating an object relation network between objects according to operation information of a user aiming at the objects; according to a preset relationship mining model, extracting local association features of objects in the object relationship network to obtain initial feature information of the objects; according to the initial characteristic information of the object, carrying out global distribution characteristic extraction on the object in the object relation network to obtain global distribution characteristic information of the object; updating the parameter information of the preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model; and extracting the characteristics of the object relationship network according to the target relationship mining model to obtain the target characteristic information of the object.

As shown in fig. 2a, the specific process of the object relationship mining method may be as follows:

101. and generating an object relation network between the objects according to the operation information of the user aiming at the objects.

The object refers to an object for a user to perform operations, and may be represented as a commodity, a song, a contact person and the like browsed, commented and shared by the user. In different scenarios, the object may be represented as any one of a commodity, a song, a contact, and the like.

The operation type of the user for the object may include (but is not limited to): browsing, praise, forward, comment, share, delete, collect, etc.

The object relation network is a mesh graph used for representing object association relation, the mesh graph is composed of nodes and edges connecting the two nodes, the nodes represent the objects, when the two nodes are connected through the edges, the two corresponding objects are associated, and the edges connecting the two nodes represent the association degree between the two objects.

In an embodiment, the step "generating an object relationship network between objects according to operation information of a user for the objects" may specifically include: acquiring a user behavior sequence based on an operation time sequence according to operation information of a user aiming at an object; and generating an object relation network between the objects according to the user behavior sequence. Specifically, referring to fig. 5, adjacent objects in the user behavior sequence of the same user are associated, that is, adjacent nodes in the object relationship network are connected by edges, and the direction of the edges may also be determined according to the sequence of the user behavior sequence for the object operation, so as to generate the object relationship network.

Preferably, the edge weight may also be calculated from attribute information of the object. The attribute information of the object is information for representing the object, for example, when the object is represented as a commodity, the attribute information of the object may include a brand of the commodity, a manufacturer, and a commodity identifier (the commodity identifier may include a picture, a video, a text description, and the like of the commodity).

The edge weight is a parameter for measuring the magnitude of the association degree, and in the object relationship network, the edge weight can be represented by the thickness and length of the edge.

102. And according to a preset relationship mining model, extracting local association features of the objects in the object relationship network to obtain initial feature information of the objects.

The preset relationship mining model is used for extracting incidence relationships in the object relationship network, and mapping graph data (usually a high-dimensional dense matrix) into an algorithm model of low-dimensional dense initial characteristic information according to incidence relationships between objects and other objects.

Wherein, the local association means: relationships or associations between an object and other objects in the vicinity in the object relationship network, local association features refer to information representing local associations.

The initial feature information is information that represents the low-dimensional density of data representing the features of the object, and includes local association features between the object and other objects and features of the object itself. And may be in the form of a feature vector or a feature matrix, etc.

The method adopts a preset relationship mining model to extract the characteristics of an object relationship network, and relates to a Computer Vision technology in an artificial intelligence technology, wherein the Computer Vision technology (Computer Vision, CV) is a science for researching how to make a machine see, and further means that a camera and a Computer are used for replacing human eyes to perform machine Vision such as identification, tracking, measurement and the like on a target, and further, image processing is performed, so that the Computer processing becomes an image which is more suitable for human eyes to observe or is transmitted to an instrument to detect. As a scientific discipline, computer vision research-related theories and techniques attempt to build artificial intelligence systems that can capture information from images or multidimensional data. The processing of the object relation network in the invention relates to image processing, image recognition, image semantic understanding and other technologies in the computer vision technology.

In this embodiment, the preset relationship mining model may include a sequence extraction submodel and a mapping submodel, and the step of "extracting local association features of an object in the object relationship network according to the preset relationship mining model to obtain initial feature information of the object" may specifically include:

adopting a sequence extraction sub-model in the preset relationship mining model to determine adjacent objects of each object, and generating an object sequence according to the adjacent objects of each object;

training hidden layer weight information in an initial mapping sub-model in the preset relationship mining model according to the object sequence to obtain a mapping sub-model in the preset relationship mining model;

and mapping the object to initial characteristic information of the object according to the hidden layer weight information in the mapping sub-model.

The process of mining the model according to the pre-trained preset relationship to obtain the initial characteristic information of the object actually uses the idea of word embedding (word vector). The basic processing element of word embedding is a word, and the basic processing element of the present embodiment is a node (i.e., an object) in the object relationship network; the word embedding is to analyze a word sequence constituting one sentence, and in the present embodiment, the mapping sub-model is to analyze an object sequence. The object sequence is a sequence formed by objects forming a path in the object relationship network, and comprises a plurality of objects arranged according to the adjacent relationship in the path.

In this embodiment, the sequence extraction submodel is used to obtain the object sequence, and the sequence extraction submodel may generate the object sequence through a walking algorithm, where the walking algorithm may start from a specific target node in the object relationship network, determine all edges connected to the target node, move from the target node to the next vertex along the connected edges, and repeat the process continuously until all adjacent nodes of the target node are found, and then determine all edges connected to the adjacent nodes, but repeat the process until all adjacent edges of the adjacent nodes are found, and repeat the process continuously, thereby finally forming a plurality of paths that run through the object relationship network. All nodes passed by each path form an object sequence according to the passing order. And accessing the adjacent node of the target node from the target node, then accessing the adjacent node of the adjacent node by the adjacent node until the adjacent node has no adjacent node, and generating the object sequence according to the access sequence.

For example, in comparison with a method for generating a sequence by random walk by deep walk, the LINE algorithm models first-order similarity and second-order similarity of nodes, and performs sampling training on edges according to weights, so as to improve accuracy of a final result, wherein the first-order similarity of the nodes can be calculated by adopting the following formula:

wherein, w_ijIs the edge weight between nodes i and j, and w is the sum of the edge weights of the entire object relationship network. If the node i and the node j are not connected by an "edge", the similarity between the two is 0.

The second-order similarity of the nodes can be calculated by adopting the following formula:

wherein, w_ijIs the edge weight between nodes i, j, where d_iIs the out degree of node i (i.e., the number of outgoing edges of node i). If there is no common neighboring node between node i and node j, the similarity between the two is 0.

In one embodiment, the mapping sub-model may be represented as a work2vec model, where a work2vec model is a group of related models used to generate object feature vectors. The models are shallow and double-layer neural networks, and after training, the objects can be mapped to initial feature information of the objects according to hidden layer weight information in a mapping sub-model, wherein the mapping sub-model is a model based on the neural networks.

The word 2vec model comprises an input layer, a hidden layer and an output layer, before input, one-hot coding can be performed on objects in an object sequence, assuming that n objects are total, each object can be represented by an n-dimensional vector, only one position of the n-dimensional vector is 1, and the rest positions are 0, in the object sequence, a separator is added between the one-hot codes of the objects, the word 2vec model can be trained according to the one-hot coding of adjacent objects, the local association relation between the objects in the object sequence is extracted, and weight information in the hidden layer is updated according to the local association relation, the hidden layer comprises a plurality of neurons, the number of the neurons is consistent with the number of the objects, and elements of generated object feature vectors are consistent, and assuming that n objects exist, the generated object feature vectors are in an n-vector dimension.

103. And according to the initial characteristic information of the object, carrying out global distribution characteristic extraction on the object in the object relation network to obtain the global distribution characteristic information of the object.

In an embodiment, the step of performing global distribution feature extraction on the object in the object relationship network according to the initial feature information of the object to obtain the global distribution feature information of the object may specifically include:

according to the number of users in the object relationship network, carrying out clustering operation on the initial characteristic information of the object, and determining a category label corresponding to the initial characteristic information;

and taking the category label corresponding to the initial characteristic information as the global distribution characteristic information of the object.

The clustering operation may be performed by using a k-means clustering algorithm, and the step of obtaining the category label according to the clustering algorithm may specifically include: and dividing the initial characteristic information of the objects into K groups according to the number of users, randomly selecting K objects as initial clustering centers, calculating the distance between each object and each seed clustering center, and allocating each object to the nearest clustering center. The cluster centers and the objects assigned to them represent a cluster. The cluster center of a cluster is recalculated for each sample assigned based on the objects existing in the cluster. This process will be repeated until some termination condition is met. The termination condition may be that no (or minimum number) objects are reassigned to different clusters, no (or minimum number) cluster centers are changed again, and the sum of squared errors is locally minimal.

In this embodiment, the initial feature information of the object is grouped according to the number of users, and the finally obtained category label can be actually expressed as the identification information of the user.

The user corresponding to the object can be used for representing the global distribution characteristics of the object in the object relationship network.

In an embodiment, the hidden layer weight information in the mapping sub-model may also be used to perform a clustering operation, where the hidden layer weight information in the mapping sub-model may actually be represented as a matrix, and the weight matrix may be operated with the one-hot code of the input object to obtain initial feature information of the object, where the weight matrix is actually a vector representation of all objects, and each row of the weight matrix is actually a vector representation of one object, and the weight vectors are subjected to a clustering operation, so as to obtain a clustering result that is the same as the initial feature information of the vectors.

104. And updating the parameter information of the preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model.

The step relates to a machine learning technology in an artificial intelligence technology, Machine Learning (ML) is a multi-field cross subject, and relates to a plurality of subjects such as probability theory, statistics, approximation theory, convex analysis and algorithm complexity theory. The special research on how a computer simulates or realizes the learning behavior of human beings so as to acquire new knowledge or skills and reorganize the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, is the fundamental approach for computers to have intelligence, and is applied to all fields of artificial intelligence. The invention relates to a semi-supervised machine learning method, which uses a category label as a pseudo label to supervise the parameter updating of a preset relationship mining model so as to enable the preset relationship mining model to learn global distribution characteristics.

In an embodiment, the step of "updating the parameter information of the preset relationship mining model according to the global distribution feature information to obtain the target relationship mining model" may include:

performing full-connection operation on the initial characteristic information according to a preset classification branch network to obtain category prediction information corresponding to the initial characteristic information;

updating parameter information in the preset relationship mining model according to a preset loss function, the category prediction information and the global distribution characteristic information to obtain a current relationship mining model;

and determining whether the current relationship mining model is the target relationship mining model or not according to the global distribution characteristic information.

Full connection layer: the learned features can be mapped to a sample label space, which mainly functions as a "classifier" in the whole convolutional neural network, and each node of the fully-connected layer is connected to all nodes output by the previous layer (e.g., the down-sampling layer in the convolutional layer), wherein one node of the fully-connected layer is called one neuron in the fully-connected layer, and the number of neurons in the fully-connected layer can be determined according to the requirements of the practical application. Optionally, in the fully-connected layer, a non-linear factor may also be added by adding an activation function, for example, an activation function sigmoid (S-type function) may be added.

The fully-connected layer generally further includes a softmax function, and the softmax function is used for mapping the category prediction score calculated by the fully-connected layer to a probability with a numerical value between (0, 1).

The step of determining whether the current relationship mining model is the target relationship mining model according to the global distribution feature information may specifically include: extracting the characteristics of the object relationship network by adopting the current relationship mining model to obtain current initial characteristic information;

according to the number of users in the object relationship network, performing clustering operation on the current initial characteristic information, determining a category label corresponding to the current initial characteristic information, and taking the category label corresponding to the current initial characteristic information as global distribution characteristic information of the object;

acquiring standard mutual information according to global distribution characteristic information obtained by two adjacent clustering operations;

when the standard mutual information reaches a preset threshold value, taking the current relationship mining model as a target relationship mining model;

and when the standard mutual information is not more than a preset threshold value, repeating the steps until the standard mutual information reaches the preset threshold value, and taking the current relationship mining model as a target relationship mining model.

The above process is described in conjunction with fig. 5:

after the relationship mining model is preset, the relationship mining model comprises a clustering branch network and a classification branch network, the clustering branch network can perform clustering operation on the initial characteristic information of the object according to the number of users in the object relationship network to determine a category label corresponding to the initial characteristic information, and the classification branch network can perform full-connection operation on the initial characteristic information to obtain category prediction information corresponding to the initial characteristic information: the global distribution feature information obtained by clustering operation cannot be reversely propagated to the preset relationship mining model to guide parameter updating, so that the parameters in the relationship mining model are updated by reversely propagating the classification branch network and the category labels to the preset relationship mining model, so that the relationship mining model can learn the global distribution features among the objects, and the method specifically comprises the following steps: acquiring a current loss value according to a preset loss function, the category prediction information and the global distribution characteristic information; updating the weight information of the mapping sub-model in the preset relationship mining model according to the current loss value to obtain a current mapping sub-model; and extracting a sub-model and the current mapping sub-model according to a sequence in a preset relationship mining model to obtain a current relationship mining model.

The loss function can be flexibly set according to the actual application requirement, for example, the loss function J can be selected as the cross entropy as follows:

wherein C is the number of categories,

for the output class prediction value, y_kIndicates whether the category predicted value and the category label are in the same category. And continuously training by reducing the error between the network category predicted value and the category label to adjust the weight to a proper value, so as to obtain the current mapping sub-model.

Mutual Information (Mutual Information) is a useful Information measure in Information theory, and the standard Mutual Information can be calculated by the following formula:

the results obtained by two adjacent clustering operations are expressed as the probability that a certain object corresponds to a certain class of labels. I is a mutual information calculation formula, and H is an entropy calculation formula. The NMI range is [0, 1], and the higher the correlation between A and B, the higher the NMI result. The weight information in the mapping sub-model is updated according to the global distribution characteristic information, so that the mapping sub-model can learn and master the global distribution characteristic of the object, and the object characteristic vector obtained by mapping can more accurately and comprehensively represent the object.

In an embodiment, after the standard mutual information satisfies the preset threshold, a parameter in the sequence extraction submodel may be further adjusted, so that the walking algorithm may extract the object sequence more accurately, which may specifically include the following steps: performing parameter adjustment on the sequence extraction submodel in the target relation mining model according to the target characteristic information to obtain an adjusted sequence extraction submodel;

combining the adjusted sequence extraction model and a target mapping sub-model in the target relation mining model to obtain an adjusted target relation mining model;

and extracting the characteristics of the object relationship network according to the adjusted target relationship mining model to obtain the adjusted target characteristic information of the object.

The step of performing parameter adjustment on the sequence extraction submodel in the target relationship mining model according to the target characteristic information to obtain an adjusted sequence extraction submodel includes:

extracting a sub-model by adopting a sequence in the preset relation mining model to obtain target local correlation information among the target characteristic information of the object;

and adjusting parameters in the sequence extraction submodel by adopting a preset loss function and the target local correlation information to obtain an adjusted sequence extraction submodel.

Taking the sequence extraction submodel as an example, the target local association information can be expressed as the similarity between target characteristic information, the target local association information is used for guiding and adjusting parameters in the sequence extraction submodel, specifically, the similarity between nodes in the object relationship network is calculated, then the similarity between the target characteristic information is calculated, the two similarities are substituted into a loss function to obtain a loss value, and the parameters are updated according to the loss value until the sequence extraction submodel converges. Here, convergence refers to a case of the sequence extraction submodel in which the loss value approaches infinity.

In an embodiment, taking the sequence extraction sub-model as an example to obtain the object sequence by using the ilene walking algorithm, the similarity may be calculated by using the following formula:

first order similarity between nodes:

first-order similarity between target feature information:

wherein, w_ijIs the edge weight between node i and node j, and w is the weight sum of the whole graph.

And u_jRespectively are target characteristic vectors of the nodes i and j, phi is a clustering regulation factor, if the nodes i and the nodes j belong to the same category in the category label, phi is 1, otherwise, phi is a parameter in the range of (0, 1).

Second-order similarity between nodes:

second-order similarity between target feature information:

whereinWherein d is_iIs the out degree of node i. And | V | is the number of nodes in the object relationship network.

105. And extracting the characteristics of the object relationship network according to the target relationship mining model to obtain the target characteristic information of the object.

In an embodiment, the target relationship mining model comprises a sequence extraction sub-model and a mapping sub-model updated based on global feature information, and the target relationship mining model is adopted to obtain the target feature information of the object, so that the target relationship mining model is more accurate due to the fact that the target relationship mining model comprises the global distribution feature of the object.

In an embodiment, after the feature extraction is performed on the object relationship network according to the object relationship mining model to obtain the object feature information of the object, parameters in the sequence extraction submodel are further adjusted to obtain an adjusted sequence extraction submodel, and finally, the feature extraction is performed on the object relationship network according to the adjusted object relationship mining model by using the adjusted object relationship mining model to obtain the adjusted object feature information of the object. And taking the adjusted target characteristic information as a final representation of the object. The adjusted target relation mining model comprises an adjusted sequence extraction model and a target mapping sub-model in the target relation mining model.

As can be seen from the above, the object relationship network between the objects may be generated according to the operation information of the user for the objects; according to a preset relationship mining model, extracting local association features of objects in the object relationship network to obtain initial feature information of the objects; according to the initial characteristic information of the object, carrying out global distribution characteristic extraction on the object in the object relation network to obtain global distribution characteristic information of the object; updating the parameter information of the preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model; and extracting the characteristics of the object relationship network according to the target relationship mining model to obtain the target characteristic information of the object. According to the scheme, the global distribution characteristic information of the objects in the object relation network can be extracted, and the preset relation mining model is trained according to the global distribution characteristic information, so that the association relation between the objects can be mined more comprehensively and accurately.

A computing platform recommended by an object is built according to a cloud technology, wherein cloud computing (cloud computing) is a computing mode, computing tasks are distributed on a resource pool formed by a large number of computers, and various application systems can acquire computing power, storage space and information service according to needs. The network that provides the resources is called the "cloud". Resources in the "cloud" appear to the user to be infinitely expandable and to be available at any time, available on demand, expandable at any time, and pay per use.

As a basic capability provider of cloud computing, a cloud computing resource pool (called a cloud platform for short, generally referred to as laaS (Infrastructure as a Service) platform is established, and multiple types of virtual resources are deployed in the resource pool and are selectively used by external clients.

According to the logic function division, a Platform as a Service (PaaS) layer can be deployed on an laaS (Infrastructure as a Service) layer, a Software as a Service (SaaS) layer is deployed on the PaaS layer, and the SaaS can be directly deployed on the laaS. PaaS is a platform on which software runs, such as a database, a web container, etc. SaaS is a variety of business software, such as web portal, sms, and mass texting. Generally, SaaS and PaaS are upper layers relative to laaS.

According to the cloud technology, the obtained target characteristic information is stored in a Database, wherein the Database (Database), in short, can be regarded as an electronic file cabinet, namely a place for storing electronic files, and a user can add, query, update, delete and the like to the data in the files. A "database" is a collection of data that is stored together in a manner that can be shared with multiple users, has as little redundancy as possible, and is independent of the application.

The above process also relates to a Cloud internet of things technology, wherein a Cloud internet of things (Cloud IOT) aims to connect information sensed by sensing devices in a traditional internet of things and received instructions into the internet, so as to really realize networking, and realize mass data storage and operation through a Cloud computing technology.

In an embodiment, the object relationship mining method further includes storing target feature information of the object as input information to a blockchain.

Referring to fig. 6a, the network device integrated with the object relationship mining apparatus is a node in the data sharing system, and each node in the data sharing system may receive input information during normal operation, and maintain data in the data sharing system based on the received input information. In order to ensure information intercommunication in the data sharing system, information connection can exist between each node in the data sharing system, and information transmission can be carried out between the nodes through the information connection. For example, when an arbitrary node in the data sharing system receives input information, other nodes in the data sharing system acquire the input information according to a consensus algorithm, and store the input information as data in shared data, so that the data stored on all the nodes in the data sharing system are consistent.

Each node in the data sharing system has a node identifier corresponding thereto, and each node in the data sharing system may store a node identifier of another node in the data sharing system, so that the generated block is broadcast to the other node in the data sharing system according to the node identifier of the other node in the following. Each node may maintain a node identifier list as shown in the following table, and store the node name and the node identifier in the node identifier list correspondingly. The node identifier may be an IP (Internet Protocol) address and any other information that can be used to identify the node. For example, when a terminal or a server integrated with the object relationship mining device performs video anomaly identification on a video to be identified to obtain target feature information of an object, the target feature information of the object is broadcasted to a node identifier list, and a network device in a data sharing system corresponding to the node identifier is obtained. The following table is only illustrative of IP addresses.

Node name	Node identification
		Node
1	117.114.151.174
		Node 2	117.116.189.145
…	…
		Node N	119.123.789.258

Each node in the data sharing system stores one identical blockchain. The block chain is composed of a plurality of blocks, as shown in fig. 6b, the block chain is composed of a plurality of blocks, the starting block includes a block header and a block main body, the block header stores an input information characteristic value, a version number, a timestamp and a difficulty value, and the block main body stores input information; the next block of the starting block takes the starting block as a parent block, the next block also comprises a block head and a block main body, the block head stores the input information characteristic value of the current block, the block head characteristic value of the parent block, the version number, the timestamp and the difficulty value, and the like, so that the block data stored in each block in the block chain is associated with the block data stored in the parent block, and the safety of the input information in the block is ensured. In this embodiment, the target feature information of the object may be stored into the tile body.

When each block in the block chain is generated, referring to fig. 6c, when the node where the block chain is located receives the input information, the input information is verified, after the verification is completed, the input information is stored in the memory pool, and the hash tree for recording the input information is updated; and then, updating the updating time stamp to the time when the input information is received, trying different random numbers, and calculating the characteristic value for multiple times, so that the calculated characteristic value can meet the following formula:

SHA256(SHA256(version+prev_hash+merkle_root+ntime+nbits+x))<TARGET

wherein, SHA256 is a characteristic value algorithm used for calculating a characteristic value; version is version information of the relevant block protocol in the block chain; prev _ hash is a block head characteristic value of a parent block of the current block; merkle _ root is a characteristic value of the input information; ntime is the update time of the update timestamp; nbits is the current difficulty, is a fixed value within a period of time, and is determined again after exceeding a fixed time period; x is a random number; TARGET is a feature threshold, which can be determined from nbits.

Therefore, when the random number meeting the formula is obtained through calculation, the information can be correspondingly stored, and the block head and the block main body are generated to obtain the current block. And then, the node where the block chain is located respectively sends the newly generated blocks to other nodes in the data sharing system where the newly generated blocks are located according to the node identifications of the other nodes in the data sharing system, the newly generated blocks are verified by the other nodes, and the newly generated blocks are added to the block chain stored in the newly generated blocks after the verification is completed.

Example II,

According to the method described in the foregoing embodiment, a case of performing object recommendation based on the target feature information will be described in further detail below by way of example. In this embodiment, a case where the recommendation apparatus is specifically integrated in the network device will be described as an example.

As shown in fig. 2b, based on the trained model, the specific flow of the vehicle identification method may be as follows:

201. when detecting the operation of a user for a target object, the network equipment acquires the target characteristic information of the object.

The network device may be a terminal or a server.

The network device may obtain the target feature information of the object or the adjusted target feature information from the blockchain.

Wherein, the objects refer to all objects used for training, and the target object refers to the object currently operated by the user.

202. And the network equipment acquires the similarity between the target object and other objects according to the target characteristic information of the object.

The target feature information of the object is generally expressed in the form of a feature vector, and an inner product of the target object feature vector and other object feature vectors can be calculated and used as the similarity.

203. And the network equipment determines the object to be recommended associated with the target object from the other objects according to the similarity.

The target object with the largest similarity can be selected as the object to be recommended.

204. And the network equipment recommends the object to be recommended to the user.

The network device may recommend the object to be recommended to the user, for example, push information such as a link, a picture, an id, and the like of the object to be recommended to the user.

Therefore, the object feature information of the object is extracted more accurately, so that the object which is possibly interested by the user can be recommended to the user more accurately.

Example III,

In order to better implement the method, the embodiment of the invention further provides an object recommendation device, and the vehicle identification device can be specifically integrated in network equipment, such as a terminal or a server.

For example, as shown in fig. 3, the object recommendation apparatus includes a generation unit 301, a local feature extraction unit 302, a global distribution feature information extraction unit 303, an update unit 304, a target feature information extraction unit 305, and a recommendation unit 306, as follows:

(1) a generating unit 301, configured to generate an object relationship network between objects according to operation information of a user for the objects;

(2) the local feature extraction unit 302 performs local associated feature extraction on the object in the object relationship network according to a preset relationship mining model to obtain initial feature information of the object.

In an embodiment, the local feature extraction unit 302 may specifically be configured to:

training an initial mapping sub-model in the preset relationship mining model according to the object sequence to obtain a mapping sub-model in the preset relationship mining model;

(3) A global distribution feature information extraction unit 303, configured to perform global distribution feature extraction on the object in the object relationship network according to the initial feature information of the object, to obtain global distribution feature information of the object.

In an embodiment, the global distribution feature information extracting unit 303 may specifically be configured to:

(4) An updating unit 304, configured to update parameter information of the preset relationship mining model according to the global distribution feature information, so as to obtain a target relationship mining model.

In an embodiment, the updating unit 304 may specifically include an updating unit including a classification subunit, an updating subunit, and a determination subunit, as follows:

Wherein the determining sub-unit may specifically be configured to:

extracting the characteristics of the object relationship network by adopting the current relationship mining model to obtain current initial characteristic information;

In an embodiment, the updating unit 304 may further include an adjusting subunit, configured to perform parameter adjustment on the sequence extraction submodel in the target relationship mining model according to the target feature information, so as to obtain an adjusted sequence extraction submodel.

The adjustment subunit may be specifically configured to:

(5) And a target feature information extraction unit 305, configured to perform feature extraction on the object relationship network according to the target relationship mining model, so as to obtain target feature information of the object.

In an embodiment, after the target feature information extracting unit 305, an adjusting unit is further included, configured to combine the adjusted sequence extraction model and a target mapping sub-model in the target relationship mining model to obtain an adjusted target relationship mining model; and extracting the characteristics of the object relationship network according to the adjusted target relationship mining model to obtain the adjusted target characteristic information of the object.

(6) And a recommending unit 306, configured to recommend an object according to the target feature information.

In an embodiment, the recommending unit 306 may specifically be configured to:

when the operation of a user for a target object is detected, acquiring target characteristic information of the object;

acquiring the similarity between the target object and other objects according to the target characteristic information of the object;

determining an object to be recommended associated with the target object from the other objects according to the similarity;

and recommending the object to be recommended to a user.

In a specific implementation, the above units may be implemented as independent entities, or may be combined arbitrarily to be implemented as the same or several entities, and the specific implementation of the above units may refer to the foregoing method embodiments, which are not described herein again.

As can be seen from the above, the object relationship network between the objects may be generated according to the operation information of the user for the objects; according to a preset relationship mining model, extracting local association features of objects in the object relationship network to obtain initial feature information of the objects; according to the initial characteristic information of the object, carrying out global distribution characteristic extraction on the object in the object relation network to obtain global distribution characteristic information of the object; updating the parameter information of the preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model; extracting the characteristics of the object relationship network according to the target relationship mining model to obtain target characteristic information of the object; and recommending the object according to the target characteristic information. According to the scheme, the global distribution characteristic information of the objects in the object relation network is extracted, and the preset relation mining model is trained according to the global distribution characteristic information, so that the association relation between the objects can be mined more comprehensively and accurately, the objects can be recommended to users more accurately, and the anti-noise performance of the relation mining model during use can be improved.

Example four,

The embodiment of the invention also provides network equipment, which can be equipment such as a server or a terminal and integrates any object recommendation device provided by the embodiment of the invention. Fig. 4 is a schematic diagram illustrating a network device according to an embodiment of the present invention, specifically:

the network device may include components such as a processor 401 of one or more processing cores, memory 402 of one or more computer-readable storage media, a power supply 403, and an input unit 404. Those skilled in the art will appreciate that the network device architecture shown in fig. 4 does not constitute a limitation of network devices and may include more or fewer components than shown, or some components may be combined, or a different arrangement of components. Wherein:

the processor 401 is a control center of the network device, connects various parts of the entire network device by using various interfaces and lines, and performs various functions of the network device and processes data by running or executing software programs and/or modules stored in the memory 402 and calling data stored in the memory 402, thereby performing overall monitoring of the network device. Optionally, processor 401 may include one or more processing cores; preferably, the processor 401 may integrate an application processor, which mainly handles operating systems, user interfaces, application programs, etc., and a modem processor, which mainly handles wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 401.

The memory 402 may be used to store software programs and modules, and the processor 401 executes various functional applications and data processing by operating the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application program required by at least one function (such as a sound playing function, an image playing function, etc.), and the like; the storage data area may store data created according to use of the network device, and the like. Further, the memory 402 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid state storage device. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 access to the memory 402.

The network device further includes a power supply 403 for supplying power to each component, and preferably, the power supply 403 is logically connected to the processor 401 through a power management system, so that functions of managing charging, discharging, and power consumption are implemented through the power management system. The power supply 403 may also include any component of one or more dc or ac power sources, recharging systems, power failure detection circuitry, power converters or inverters, power status indicators, and the like.

The network device may also include an input unit 404, where the input unit 404 may be used to receive input numeric or character information and generate keyboard, mouse, joystick, optical or trackball signal inputs related to user settings and function control.

Although not shown, the network device may further include a display unit and the like, which are not described in detail herein. Specifically, in this embodiment, the processor 401 in the network device loads the executable file corresponding to the process of one or more application programs into the memory 402 according to the following instructions, and the processor 401 runs the application program stored in the memory 402, thereby implementing various functions as follows:

and recommending the object according to the target characteristic information.

The above operations can be implemented in the foregoing embodiments, and are not described in detail herein.

As can be seen from the above, the object relationship network between the objects may be generated according to the operation information of the user for the objects; according to a preset relationship mining model, extracting local association features of objects in the object relationship network to obtain initial feature information of the objects; according to the initial characteristic information of the object, carrying out global distribution characteristic extraction on the object in the object relation network to obtain global distribution characteristic information of the object; updating the parameter information of the preset relationship mining model according to the global distribution characteristic information to obtain a target relationship mining model; extracting the characteristics of the object relationship network according to the target relationship mining model to obtain target characteristic information of the object; and recommending the object according to the target characteristic information.

Example V,

It will be understood by those skilled in the art that all or part of the steps of the methods of the above embodiments may be performed by instructions or by associated hardware controlled by the instructions, which may be stored in a computer readable storage medium and loaded and executed by a processor.

To this end, embodiments of the present invention provide a storage medium having stored therein a plurality of instructions, which can be loaded by a processor to perform the steps of any one of the intelligent object recommendation methods provided by the embodiments of the present invention. For example, the instructions may perform the steps of:

and recommending the object according to the target characteristic information.

Wherein the storage medium may include: read Only Memory (ROM), Random Access Memory (RAM), magnetic or optical disks, and the like.

Since the instructions stored in the storage medium may execute the steps in any one of the intelligent object recommendation methods provided in the embodiments of the present invention, beneficial effects that can be achieved by any one of the intelligent object recommendation methods provided in the embodiments of the present invention may be achieved, which are detailed in the foregoing embodiments and will not be described herein again.

The foregoing describes in detail an object recommendation method, an object recommendation device, and a storage medium based on intelligence according to embodiments of the present invention, and specific examples are applied herein to explain the principles and implementations of the present invention, and the descriptions of the foregoing embodiments are only used to help understanding the method and the core ideas of the present invention; meanwhile, for those skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.

Claims

1. An intelligent-based object recommendation method is characterized by comprising the following steps:

and recommending the object according to the target characteristic information.

2. The intelligent object recommendation method according to claim 1, wherein the performing global distribution feature extraction on the object in the object relationship network according to the initial feature information of the object to obtain the global distribution feature information of the object comprises:

3. The intelligent object recommendation method according to claim 1, wherein the updating the parameter information of the preset relationship mining model according to the global distribution feature information to obtain a target relationship mining model comprises:

4. The intelligent object recommendation method according to claim 3, wherein said determining whether a current relationship mining model is the target relationship mining model according to the global distribution feature information comprises:

and when the standard mutual information does not meet a preset threshold value, repeating the steps until the standard mutual information reaches the preset threshold value, and taking the current relationship mining model as a target relationship mining model.

5. The intelligent object recommendation method according to any one of claims 1-4, wherein the extracting local association features of the object in the object relationship network according to a preset relationship mining model to obtain initial feature information of the object comprises:

6. The intelligent object recommendation method according to claim 5, wherein the updating parameter information in the preset relationship mining model according to a preset loss function, the category prediction information and the global distribution feature information to obtain a current relationship mining model comprises:

acquiring a current loss value according to a preset loss function, the category prediction information and the global distribution characteristic information;

updating the weight information of the mapping sub-model in the preset relationship mining model according to the current loss value to obtain a current mapping sub-model;

and extracting a sub-model and the current mapping sub-model according to a sequence in a preset relationship mining model to obtain a current relationship mining model.

7. The intelligent object recommendation method according to claim 6, wherein after the extracting features of the object relationship network according to the target relationship mining model to obtain the target feature information of the object, the method further comprises:

performing parameter adjustment on the sequence extraction submodel in the target relation mining model according to the target characteristic information to obtain an adjusted sequence extraction submodel;

8. The intelligent object recommendation method according to claim 7, wherein the performing parameter adjustment on the sequence extraction submodel in the target relationship mining model according to the target feature information to obtain an adjusted sequence extraction submodel comprises:

9. The intelligent object recommendation method according to claim 1, wherein the recommending objects according to the target feature information comprises:

and recommending the object to be recommended to a user.

10. An object recommendation apparatus, comprising:

11. A storage medium storing a plurality of instructions adapted to be loaded by a processor to perform the steps of the intelligent object recommendation method according to any one of claims 1 to 9.