CN114462461A

CN114462461A - Community characteristic determination method based on relational network and related device

Info

Publication number: CN114462461A
Application number: CN202011139656.3A
Authority: CN
Inventors: 陈信欢; 李友焕; 胡雨松; 陈守志
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2020-10-22
Filing date: 2020-10-22
Publication date: 2022-05-10

Abstract

The embodiment of the application discloses a community characteristic determination method based on a relational network, which comprises the following steps: carrying out supervision training on the neural network model through a first training sample, and determining a community feature extraction model based on a hidden layer of the trained neural network model, wherein the first training sample comprises a community identifier of a community to which a sample node belongs in a sample relation network and an account number type of the sample node; according to community identifications of a plurality of communities to which the nodes to be identified belong, community characteristics corresponding to the nodes to be identified are determined through a community characteristic extraction model, and the community characteristics are used for identifying network relationships of the nodes to be identified in a relationship network. According to the method and the device, the community feature extraction model can be determined from the neural network model obtained through training, so that the community feature extracting model for extracting the nodes to be identified can have a data structure suitable for the classification model and carry more information related to the account number type.

Description

Community characteristic determination method based on relational network and related device

Technical Field

The present application relates to the field of data processing, and in particular, to a method and a related apparatus for determining community characteristics based on a relationship network.

Background

With the popularization of the internet, users can realize the requirements of social contact, shopping, entertainment, fund exchange and the like through network behaviors on the internet. The network behaviors of the users can connect the users, and the relationship networks are in various forms, for example, a social network can be formed through friend relationships, and a fund network is formed through transaction relationships.

The network account used by the user can be used as a node in the relational network, and the connection line between the nodes is used for showing the network behavior between the nodes so as to mark the correlation degree between the nodes. The relationship network can be divided into a plurality of communities (communities) according to the degree of association among the nodes, wherein the communities are associated node sets with compact topological structures on the relationship network, namely cluster-type clusters of unstructured graph data. Moreover, often the same node belongs to multiple communities.

Therefore, the network relationship of the account corresponding to the node can be reflected by a plurality of communities to which the node belongs, and if community characteristics used for expressing the network relationship of the account can be accurately obtained based on the communities, the method can play a remarkable role in application scenes such as account identification and account classification.

Disclosure of Invention

In order to solve the technical problem, the application provides a method for determining community characteristics based on a relational network, a neural network model for judging the account number type of a node according to the community identification of the node can be obtained through supervision training, and a community characteristic extraction model is determined based on a hidden layer of the neural network model, so that the community characteristics of the node to be identified extracted by using the community characteristic extraction model can have a data structure suitable for a classification model and carry more information related to the account number type.

The embodiment of the application discloses the following technical scheme:

in a first aspect, an embodiment of the present application provides a method for determining community characteristics based on a relationship network, where the method includes:

determining a neural network model and a corresponding first training sample, wherein the first training sample comprises a community identification of a community to which a sample node belongs in a sample relationship network and an account number type of the sample node;

performing supervised training on the neural network model through the first training sample, and determining a community feature extraction model based on a hidden layer of the trained neural network model;

determining a plurality of communities to which nodes to be identified belong in the relational network;

and determining community characteristics corresponding to the node to be identified through the community characteristic extraction model according to the community identifications of the communities, wherein the community characteristics are used for identifying the network relationship of the node to be identified in the relationship network.

In a second aspect, an embodiment of the present application provides a community feature determination device based on a relationship network, where the device includes a first determination unit, a first training unit, a second determination unit, and a third determination unit:

the first determining unit is used for determining a neural network model and a corresponding first training sample, wherein the first training sample comprises a community identifier of a community to which a sample node belongs in a sample relationship network and an account type of the sample node;

the first training unit is used for carrying out supervised training on the neural network model through the first training sample and determining a community feature extraction model based on a hidden layer of the trained neural network model;

the second determining unit is configured to determine multiple communities to which nodes to be identified belong in the relationship network;

the third determining unit is configured to determine, according to the community identifiers of the multiple communities, a community feature corresponding to the node to be identified through the community feature extraction model, where the community feature is used to identify a network relationship that the node to be identified has in the relationship network.

In a possible implementation manner, the fourth determining unit is specifically configured to:

determining input data of a classification model according to the account characteristics corresponding to the node to be identified and the community characteristics;

and determining the account type of the node to be identified according to the input data through the classification model.

In a possible implementation manner, the apparatus further includes a fifth determining unit, a sixth determining unit, and a second training unit:

the fifth determining unit is configured to determine a sample community feature of the sample node according to a first training sample and the community feature extraction model;

the sixth determining unit is configured to determine a classification model and a corresponding second training sample, where the second training sample includes a sample account feature and a sample community feature of the sample node, and an account type of the sample node;

and the second training unit is used for carrying out supervision training on the classification model through the second training sample to obtain the classification model.

In a possible implementation manner, the relationship network is a feature value relationship network generated based on feature value transfer, and the fourth determining unit is specifically configured to:

and determining the account type of the node to be identified as a normal account or an abnormal account according to the account characteristics corresponding to the node to be identified and the community characteristics.

In a third aspect, an embodiment of the present application provides a computer device, where the device includes a processor and a memory:

the memory is used for storing program codes and transmitting the program codes to the processor;

the processor is configured to execute the method for determining community characteristics based on a relationship network according to the first aspect according to instructions in the program code.

In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium for storing a computer program for executing the method for determining community characteristics based on a relationship network in the first aspect.

According to the technical scheme, when community characteristics corresponding to the node to be recognized need to be determined, the neural network model can be supervised and trained through the first training sample, the first training sample comprises the community identification of the community to which the sample node belongs in the sample relation network and the account number type of the sample node, in the training process, the community characteristics corresponding to the sample node can be extracted through the hidden layer of the neural network model according to the community identification of the sample node input by the input layer, then the community characteristics are transmitted to the output layer to recognize the account number type, model training is conducted under the supervision of the account number type of the first training sample, and the trained neural network model has the capability of judging the node account number type according to the community identification of the community to which the node belongs. Based on the training, the accuracy and the performance of extracting the community features by the hidden layer of the neural network model can be improved. The method comprises the steps of determining a community feature extraction model according to a hidden layer of a trained neural network model, when community features of nodes to be recognized need to be extracted, determining a plurality of communities to which the nodes to be recognized belong in a relational network firstly, then determining the community features corresponding to the nodes to be recognized through the community feature extraction model according to community identifications of the communities, and enabling the community features to identify network relationships of the nodes to be recognized in the relational network, namely reserving community features of the nodes to be recognized in the relational network. Meanwhile, the neural network model belongs to one of classification models, so that the community features obtained through the community feature extraction model are closer to classification services, and the data structure suitable for the classification model is provided.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative efforts.

Fig. 1 is a schematic diagram of a method for determining community characteristics based on a relationship network in an actual application scenario according to an embodiment of the present application;

fig. 2 is a flowchart of a community characteristic determination method based on a relationship network according to an embodiment of the present application;

fig. 3 is a schematic diagram of a method for determining community features based on a relational network in an actual application scenario according to an embodiment of the present application;

fig. 4 is a schematic diagram of a method for determining community features based on a relationship network in an actual application scenario according to an embodiment of the present application;

fig. 5 is a block diagram of a structure of a community characteristic determining apparatus based on a relationship network according to an embodiment of the present application;

fig. 6 is a block diagram of a computer device according to an embodiment of the present disclosure;

fig. 7 is a block diagram of a server according to an embodiment of the present disclosure.

Detailed Description

Embodiments of the present application are described below with reference to the accompanying drawings.

With the continuous improvement of the complexity of the relational network, the difficulty of accurately analyzing the nodes in the relational network is increasing. The community characteristics are one of important reference standards for analyzing the nodes in the relational network. However, in the related art, the community feature extraction complexity is too high, and the extracted community feature accuracy is poor.

It is to be understood that the method may be applied to a processing device, which is a processing device having a feature extraction function, and may be, for example, a terminal device or a server having a feature extraction function. The method can be independently executed by the terminal equipment or the server, and can also be applied to a network scene of communication between the terminal equipment and the server, and the terminal equipment and the server are matched for operation. The terminal device may be a mobile phone, a desktop computer, a Personal Digital Assistant (PDA for short), a tablet computer, or the like. The server may be understood as an application server, or may also be a Web server, and in actual deployment, the server may be an independent physical server, or may be a server cluster or a distributed system formed by a plurality of physical servers. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the application is not limited herein.

In addition, the present application relates to Artificial Intelligence (AI) technology. Artificial intelligence is a theory, method, technique and application system that uses a digital computer or a machine controlled by a digital computer to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use the knowledge to obtain the best results. 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 artificial intelligence technology is a comprehensive subject and relates to the field of extensive technology, namely the technology of a hardware level and the technology of a software level. The artificial intelligence infrastructure generally includes technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other major directions, and the application mainly relates to the machine learning technology.

Machine Learning (ML) is a multi-domain cross discipline, and relates to a plurality of disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory and the like. 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. Machine learning and deep learning generally include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and formal education learning.

In the embodiment of the present application, models involved in neural network models, classification models, and the like may be trained through machine learning techniques.

In order to facilitate understanding of the technical scheme of the present application, a method for determining community characteristics based on a relationship network provided in the embodiment of the present application will be introduced in combination with an actual application scenario.

Referring to fig. 1, fig. 1 is a schematic diagram of a community feature determination method based on a relationship network in an actual application scenario provided by the embodiment of the present application. In this practical application scenario, the processing device is a server 101, and the neural network model is a classification model having three layers, i.e., an input layer, a hidden layer, and an output layer.

In order to incorporate the community features into the judgment of the node account types, the community features capable of reflecting the account types need to be obtained first. In the practical application scenario, a first training sample may be used to perform supervised training on the neural network model, where the first training sample includes a community identifier of a community to which a sample node belongs in the sample relationship network, and an account type of the sample node. After the server 101 obtains the first training sample, the first training sample may be input into an input layer of the neural network model, the input layer may transmit data in the first training sample to a hidden layer, and the hidden layer may perform community feature extraction on a community identifier in the first training sample, so that the extracted feature data may carry more effective data for embodying the account type of the sample node, for example, a community characteristic of the sample node in a sample relational network. After the characteristics are extracted, the hidden layer can transmit the obtained characteristic data to the output layer, the output layer classifies the characteristic data, and the account type is determined and output. In the training process, the neural network model can compare the account number type output by the output layer with the account number type of the sample node in the first training sample, and parameters in the model are corrected according to the comparison result, so that the neural network model capable of accurately judging the node account number type through community identification is obtained through training.

It can be understood that, since the hidden layer has a function of extracting the community features from the community identifiers, after training is completed, the server 101 may determine a community feature extraction model for extracting the community features based on the hidden layer. In this practical application scenario, the server 101 may extract model parameters of an input layer and a hidden layer in the neural network model, and determine the model parameters as the community feature extraction model, so that after a community identifier corresponding to a node is input into the community feature extraction model, the community feature corresponding to the node can be obtained. The community identification of the attributive community can identify the community to which the node belongs, and the neural network model is also used for judging the account number type of the node, so that the community characteristics extracted through the hidden layer can embody the network relationship of the node in the relationship network and carry related information for judging the account number type.

When applying the community feature extraction model, the server 101 may first obtain a relationship network including a plurality of nodes, and according to the relationship network, the server 101 may determine a community to which each node belongs. When determining the community characteristics of a certain node to be identified, the server 101 may first determine, according to the relationship network, an attribution community of the node to be identified corresponding to the node to be identified, and then determine a community identifier corresponding to the attribution community. The server 101 may obtain the community feature corresponding to the node to be identified through the community identifier and the community feature extraction model, where the community feature may carry related information for determining the account type of the node to be identified. Therefore, when the account type of the node to be identified is judged through the community characteristics, the obtained judgment result can be more accurate.

Meanwhile, because the neural network model belongs to one of the classification models, the data processed by the neural network model generally has a data structure suitable for the classification model, that is, the community features extracted by the hidden layer are suitable for other classification models. When the community characteristics are applied to other classification models for classification judgment, the classification models can more quickly and accurately identify and process the node account, so that the classification efficiency and accuracy are further improved.

Next, a method for determining community characteristics based on a relationship network according to an embodiment of the present application will be described with reference to the drawings.

Referring to fig. 2, fig. 2 is a flowchart of a method for determining community characteristics based on a relational network according to an embodiment of the present application, where the method includes:

s201: a neural network model and a corresponding first training sample are determined.

In order to blend the community features into various application scenarios of the pair of nodes, the processing device needs to be able to obtain more accurate community features corresponding to each node. The node is used for embodying a user identifier of a user in a relationship network, and may be, for example, a network account used by the user in the relationship network; the relational network is a network formed by connection relations among a plurality of nodes; the community is a node set formed by nodes with a compact topological structure on the relational network, and the community characteristics are used for embodying the network relationship of the nodes in the relational network. By adding the community characteristics, the processing equipment can better combine the network relationship of the node in the relationship network when analyzing the node, so that the analysis result is more effective and accurate.

The relationship network may include a variety of types, including, for example, a social network generated based on user interaction with social software, a funding network generated based on funding transactions conducted between users, and the like. In the funding network, nodes may be funding accounts used by users to manage funds.

It can be understood that, in the application scenarios related to the nodes, the proportion of the application scenarios in which the account types corresponding to the nodes are classified and analyzed is large. The account type is used to identify a type of the node in the relationship network, and may include, for example, whether the node belongs to an abnormal node, whether the node belongs to a security node, and the like. Therefore, in order to more accurately judge the account type corresponding to the node in the application scenes, the processing device can enable the extracted node community characteristics to carry more effective data capable of representing the node account type. Meanwhile, in the application scenario, various classification models are generally adopted for classifying and analyzing the account types, and data suitable for the classification models often have unique data structures, so that the classification models can more accurately utilize the community features in order to further improve the applicability of the community features in the classification models, and the processing equipment can extract the community features as much as possible by using the data structures of the classification models.

As one of the classical classification models, the neural network model has the capability of rapidly and efficiently performing classification analysis on samples based on certain sample characteristics. Therefore, in the embodiment of the present application, in order to obtain the community features with the above characteristics, the processing device may first train to obtain a neural network model, where the trained neural network model is used to extract the community features of the nodes according to the relevant information of the nodes, and classify the account types of the nodes according to the community features. Therefore, in the working process of the trained neural network model, part of the model can have the function of extracting the community features according to the relevant information of the nodes. Because the trained neural network model is a classification model for classifying the account types, the community features extracted by the model not only meet the data structure with the classification model, but also can carry effective data related to the account types in the extraction process, so that the requirement on the community features in the application is met.

To obtain the trained neural network model, the processing device may first determine an initial neural network model and a corresponding first training sample. The first training sample comprises a community identification of a community to which a sample node belongs in the sample relationship network and an account type of the sample node. The community identifier may be determined by the processing device according to an attribution relationship of the sample node in the sample network, and may be used to identify a community to which the sample node belongs in the sample relationship network, where the attribution relationship indicates that the sample node is one of the nodes forming the sample community; the sample relational network can be any one type of relational network, and the sample node is a node with a known account number type in the sample relational network.

The community identification can be used for identifying communities in the relational network, and can also carry other related community information for the convenience of determining community characteristics of the model, and the community information can embody partial community characteristics of the community. For example, when the relationship network is a fund network, the community may be a community formed by user fund accounts with close fund transactions, and the community identifier of the community may carry information such as transaction frequency and transaction time of users in the community, and a proportion of abnormal nodes in the community, where the abnormal nodes refer to nodes with abnormal fund operations, such as nodes that have performed a large number of asset transactions within a short period. Therefore, the neural network model can extract more accurate community characteristics based on the community characteristics, and further can accurately judge the account number type of the sample node.

S202: and carrying out supervision training on the neural network model through the first training sample, and determining a community feature extraction model based on a hidden layer of the trained neural network model.

After determining the first training sample, the processing device may perform supervised training on the neural network model through the first training sample. In the embodiment of the present application, the input may be a community identifier of a community to which the sample node belongs, and the expected output may be an account type of the sample node.

It is understood that a neural network model is generally composed of an input layer, a hidden layer and an output layer, wherein the input layer is used for receiving data and transmitting the data to the hidden layer; the hidden layer is used for processing the data, so that the processed data has a classification function more suitable for the neural network model; the output layer is used for classifying the processed data and outputting a classification result. In the training process of the embodiment of the application, the input layer of the neural network model can receive the first training sample and transmit the first training sample to the hidden layer; the hidden layer can extract community characteristics of the sample nodes suitable for account type classification according to the community identification of the community to which the sample nodes in the first training sample belong, and transmit the community characteristics to the output layer; the output layer can classify and analyze the account types corresponding to the sample nodes according to the community characteristics and output the account types corresponding to the sample nodes. In the supervised training mode, the processing device may compare the account type of the sample node in the first training sample with the account type output by the neural network model, and feed back the comparison result to the neural network model, so that the neural network model can correct the model parameter thereof based on the comparison result, and the output account type is more suitable for the actual account type of the sample node.

Therefore, when the neural network model is trained in the above manner, the hidden layer of the neural network model can gradually have the capability of accurately extracting the community features of the sample nodes according to the community identifications of the communities to which the sample nodes belong, and in order to enable the output layer to accurately judge the account number types of the sample nodes, the community features can also have more effective data for embodying the account number types of the sample nodes. Meanwhile, the neural network model is a classification model, and the hidden layer performs data processing on data input by the input layer based on the data structure of the classification model, so that the community features extracted by the hidden layer can have the data structure suitable for the classification model. Therefore, the community features extracted by the hidden layer of the trained neural network model can meet the requirement of the application on the community features.

Based on this, in order to obtain a form for accurately extracting the community features suitable for account type classification in a classification model data structure, the processing device may split the trained neural network model, and determine the community feature extraction model based on the hidden layer of the trained neural network model. The method is equivalent to stripping an original output layer from a neural network model, and a hidden layer of the neural network model is used as the output layer of the community feature model, so that the community feature extraction model can be used for extracting the community features corresponding to the nodes according to the community identifications of the communities to which the nodes belong.

S203: and determining a plurality of communities to which the nodes to be identified belong in the relational network.

After the community feature extraction model is obtained, the processing device may apply the community feature extraction model to community feature extraction of the node. For example, when the processing device needs to perform classification analysis on the account type of a certain node to be identified in the relationship network based on the community characteristics of the node to be identified, a plurality of communities to which the node to be identified belongs in the relationship network may be determined first. In a possible implementation manner, since the relationship network includes a plurality of nodes and the communities are formed by some or all of the plurality of nodes, the processing device may determine, according to the node to be identified, the plurality of communities to which the node to be identified belongs by using information of the nodes included in each community.

In order to enable the community feature extraction model to extract community features more accurately, the processing device can select the relationship network in a targeted manner. For example, in different types of relationship networks, the content contained in the community identifier may be different, for example, the community identifier in the fund network may include a fund flow situation in the community, the community identifier in the social network may include a node chat frequency situation in the community, and the like. Because the neural network model is obtained by performing supervised training based on the sample relational network, the community feature extraction model may be more suitable for determining the community features based on the community identification of the type of the sample relational network.

Based on this, in order to further improve the accuracy of the community feature, when using the community feature extraction model, the processing device may select a node in a relationship network of the same or similar type as the sample relationship network as the node to be identified. For example, when the sample relationship network is a fund network, the relationship network where the node to be identified is located may also be a fund network or a virtual currency network, etc.

Meanwhile, the Method for determining the attributive community may include multiple types, which may be generally divided into two types, one type is a heuristic Algorithm, and the community to which the node belongs is obtained by using the transmissibility, the complete subgraph characteristics and the Matrix decomposition of the complex network, such as a community monitoring (SLPA) Algorithm, a Label-transfer-based overlapping community discovery (COPRA) Algorithm, a Cluster filtering Method (CPM) Algorithm, a non-negative Matrix decomposition (NMF) Algorithm, and the like; the other type is a Graph neural Network (GCN) algorithm, which embeds nodes and edges in a Network into a low-dimensional space to be converted into structured data, and then combines the structured data with a traditional non-Graph Clustering algorithm, such as a Network vector (Network Embedding) and a K-means Clustering algorithm (K-means Clustering algorithm) Based on a Graph neural Network and a Density-Based Clustering of Applications with Noise (DBSCAN) algorithm.

In the embodiment of the present application, a partitioning algorithm improved based on COPRA may be adopted to determine the home community.

In addition, because through different modes, the node attribution communities determined may also be distinguished, so that in order to enable the community feature extraction model to determine the community features of the nodes to be identified more accurately, the processing equipment can determine the community to which the nodes to be identified belong in the same mode as the mode of determining the community to which the sample nodes belong, thereby ensuring that the attribution community can be attached to the model features of the community feature extraction model as much as possible.

S204: and determining community characteristics corresponding to the nodes to be identified through a community characteristic extraction model according to the community identifications of the plurality of communities.

After determining the multiple communities to which the nodes to be identified belong, the processing device may determine community identifiers of the multiple communities, and then use the community identifiers as input data of the community feature extraction model. The processing device can determine the community characteristics corresponding to the node to be identified through the community characteristic extraction model, the community characteristics are used for identifying the network relationship of the node to be identified in the relationship network, and the network relationship can reflect the association degree of the node to be identified and other nodes in the relationship network and the association degree of the node to be identified and each community in the relationship network.

Through the community characteristics, the processing equipment can analyze the recessive characteristic of the node to be identified, the recessive characteristic refers to the characteristic that the dominant information of the node cannot be directly obtained, and the dominant information refers to the node information which does not need to be processed and can be directly obtained. For example, when the relationship network is a fund network, the explicit information may be the fund transfer-out amount, the frequency, and the like of the node to be identified, and through the explicit information, the processing device may analyze some explicit characteristics of the node to be identified, for example, through analyzing the fund transfer-out amount of the node to be identified, the processing device may analyze whether there is an abnormal fund operation in the node to be identified.

However, it may be difficult to accurately determine the node to be identified only by means of the explicit characteristic, and some information of the node to be identified may not have an abnormal condition, for example, in a fund network, the number of money transfers and the frequency of money transfers of a certain node to be identified may be in a normal range, but the node to be identified may often have fund exchange with some abnormal nodes or some abnormal communities, and therefore, in an actual situation, the node to be identified is also highly likely to belong to an abnormal node. At this moment, the processing device may analyze the implicit characteristic of the node to be recognized, for example, may analyze the implicit characteristic such as the community characteristic of the node to be recognized by determining the community characteristic corresponding to the node to be recognized, so that when the community characteristic is applied to other application scenarios, the node to be recognized may be analyzed and judged by integrating the implicit characteristic of the node to be recognized.

It will be appreciated that the number of hidden layers of the neural network model may vary depending on the accuracy requirements for data processing. In one possible implementation, the neural network model may include an input layer, an output layer, and k hidden layers located between the belonging layer and the output layer. Wherein the value of k can be adjusted according to the data accuracy requirement of the model. After the input layer receives the related data, the data can be transmitted to the first hidden layer, the first hidden layer can extract the community characteristics of the nodes from the data, the community characteristics are transmitted to each subsequent hidden layer to be subjected to layer-by-layer iteration processing, and finally a more accurate community characteristic with more prominent characteristics is obtained. Therefore, a partial model composed of any one of the k hidden layers and a model layer located in front of the hidden layer has a function of extracting community features corresponding to the node according to community identification of the community to which the node belongs.

Based on this, in one possible implementation, the processing device may determine a target hidden layer from the k hidden layers, where the target hidden layer may be any one of the k hidden layers. Then, the processing device may generate the community feature extraction model based on a model structure of the neural network model between the input layer and the target hidden layer, and an output layer of the community feature extraction model is the target hidden layer. Therefore, in the community feature extraction model, after the community identification of the community to which the node to be identified belongs is input, the output layer of the model can output the community feature corresponding to the node.

The types of the Neural Network model may also include multiple types, for example, a Recurrent Neural Network (RNN) model, a Long Short-Term Memory (LSTM) model, a Variational self-encoder (VAE) model, and the like. In a possible implementation manner, the neural network model may be a multi-Layer Perceptron (MLP) model, and the k hidden layers of the MLP model may perform Layer-by-Layer iterative processing on the extracted community features in the above manner. As mentioned above, each hidden layer can process the data output by the previous layer of model more accurately, that is, the closer to the hidden layer of the output layer, the higher the accuracy of the community features obtained by the processing, so that the output layer can judge the account number type of the node based on a high-accuracy community feature.

Therefore, in order to enable the community feature extraction model determined based on the hidden layer of the MLP model to output the community features with high accuracy, the processing device may use the hidden layer closest to the output layer among the k hidden layers as the target hidden layer, so that the community feature extraction model can output the community features of the node to be identified after the iteration processing of the k hidden layers.

In addition to improving the accuracy of the model, the processing device may also improve from the data input to the model in order to further improve the accuracy of the community features. It can be understood that as network technology continues to grow in popularity, users using the network continue to increase. The complexity of the relationship network corresponding to each user is also rapidly increasing, so that each node in the relationship network may belong to more communities. For example, when the relationship network is a message exchange network established based on chat software, the node may be a user account of a user in the chat software, and the user may add a plurality of group chats and exchange messages with a plurality of users, where each group chat and each user may belong to a different community, and when determining the home community, because the node has direct message exchange with the user nodes in the group chats and other user nodes, the node corresponding to the user may be included in node information included in the plurality of communities, that is, a plurality of communities to which the node belongs may be determined. Even if the node has too little communication with a certain node and the communication frequency is low, the node may be determined to belong to a community.

Therefore, communities to which the node belongs are determined directly according to the attribution relationship possibly more, and the main community to which the node belongs in the relationship network is difficult to highlight, so that when the community identifications of the communities are directly applied to the community feature extraction model, the determined community features are not obvious enough to highlight the community features of the node. Based on this, in order to further highlight the community characteristics of the node to be identified and find the main community link of the node to be identified in the relationship network, after determining multiple communities to which the node to be identified belongs, the processing device may first determine the attribution probabilities that the node to be identified respectively corresponds to the multiple communities, where the attribution probabilities can reflect the closeness of the link between the node to be identified and the communities.

It can be understood that, when the degree of closeness of the connection between the node to be identified and a certain community is higher, the community can embody the community characteristics of the node to be identified in the relationship network. Therefore, the processing device may set a threshold condition for the attribution probability, and then determine, according to the attribution probability, a target community with an attribution probability greater than the threshold condition from the multiple communities, where the target community is a community with a closer relationship with the node to be identified in the multiple attribution communities. The processing device can determine the community characteristics corresponding to the node to be identified through the community characteristic extraction model according to the community identification of the target community.

For example, in the above-mentioned relationship network established based on the chat software, the attribution probability may be determined by the message exchange frequency between the node to be identified and the plurality of communities. The processing device can determine, through the judgment of the threshold condition, the communities with higher message communication frequency with the node to be identified as target communities, which are main communities for the user to perform message communication in the relationship network, so that the network relationship of the node to be identified in the relationship network can be further highlighted according to the community characteristics determined by the target communities, and the community characteristics of the node to be identified can be more obviously reflected.

After the community features corresponding to the node to be identified are obtained, the processing device may apply the community features in a plurality of application scenarios. In a possible implementation manner, the processing device may determine, by using the community feature, an account type corresponding to the node to be identified. In order to further improve the accuracy of determining the account type, the processing device may further analyze the account type by combining with other characteristics related to the node to be recognized on the basis of the community characteristics. For example, the processing device may determine an account characteristic corresponding to the node to be recognized, where the account characteristic is used to represent an account characteristic of the node to be recognized in the relationship network, and the account characteristic may be represented without passing through a community relationship of the node to be recognized. For example, when the relationship network is a fund network, the node to be identified may be a fund account of the user in the fund network, and the processing device may monitor the fund account, analyze whether there is an abnormal behavior of the user, for example, whether there is an abnormal place to log in, an abnormal amount of fund to transfer out, and determine an account characteristic corresponding to the node to be identified based on the monitoring result.

Meanwhile, when the types of the relationship networks are different, the content included in the account feature may also be different. For example, when the relationship network is a social network generated based on chat software, the account characteristics may include a login location of the account, a chat time, a message sending mode, and the like. Therefore, when the account type of the node is determined based on the account characteristics, the determined account type can be more attached to the specific function of the relationship network, so that the account type is more effective and the accuracy is higher.

The processing equipment can determine the account number type of the node to be identified according to the account number characteristic and the community characteristic corresponding to the node to be identified, so that the obtained account number type can be integrated with the account number characteristic of the node to be identified and the community characteristic in the relationship network, the dominant characteristic of the node to be identified can be shown, namely the account number characteristic which can be directly obtained by monitoring the behavior of the node to be identified can be shown, the recessive characteristic of the node to be identified can be shown, and namely the community characteristic which is obtained by analyzing the community relationship of the node to be identified can be shown.

For example, in the above-mentioned fund network, the explicit characteristic may be a characteristic obtained by monitoring a behavior of the fund account itself, and the behavior may be an amount of transfer funds, a login behavior, or the like; the invisible characteristic may be some characteristics that cannot be directly obtained only by means of information of the account number, for example, a behavior of a certain fund account number may not be abnormal, but a plurality of abnormal fund account numbers exist in a certain community to which the fund account number belongs, at this time, the account number type determined by the method can be integrated with the community characteristic of the fund account number, and the fund account number is determined as an account number with a certain abnormal probability, so that the invisible characteristic of the fund account number can be mined, and the determined account number type is more accurate.

When the account type of the node to be identified is determined according to the above conditions, the determination method may also include multiple determination methods. As mentioned above, the community features obtained by the community feature extraction model have a data structure suitable for the classification model, so that, in order to make the determination of the account type more efficient and accurate, in a possible implementation manner, the processing device may determine the account type of the node to be identified through the classification model.

The processing device may determine the input data of the classification model according to the account number feature and the community feature of the node to be identified. For example, the community feature and the account feature can be embodied in a vector mode, and a vector (embedding) is a multi-dimensional feature representation of data, and can reduce the dimension of complex non-European data to low-dimensional data to obtain a feature vector with low dimension and rich information quantity. In the community feature extraction model, the community identification of a plurality of communities input by the processing equipment can be converted into the community feature vector of the low latitude through data processing by the model, so that the community characteristic of the node to be recognized can be reserved, and the model can be applied to the classification model more conveniently and effectively. The processing device may combine the account feature vector of the node to be identified with the community feature vector, and determine the combined vector as the input data of the classification model.

The processing device may determine the account type of the node to be identified according to the input data through the classification model. It can be understood that, because the types of the relationship network are rich, the types of account types corresponding to the nodes to be identified may also be different in different relationship network types, for example, when the relationship network is a social network established based on social software, the account types may include an account for normal login and an account for abnormal login, and the like; when the relationship network is a fund network, the account types may include an account for normal fund operation, an account for abnormal fund operation, and the like.

In one possible implementation, the relationship network may be a feature value relationship network generated based on feature value transfer, and the feature value may include various types, such as a virtual currency feature value in a game, a real currency feature value in a fund software, and the like. In order to ensure the asset security of each node in the eigenvalue relationship network, the processing device may determine the account type corresponding to the node.

In the characteristic value relationship network, the account types may include a normal account and an abnormal account, the normal account refers to an account type with a low risk of characteristic value transfer operation related to the node, and the abnormal account refers to an account type with a high risk of characteristic value transfer operation related to the node. The processing device can determine the account type of the node to be identified as a normal account or an abnormal account according to the account type and the community characteristics corresponding to the node to be identified, so that the node can be correspondingly processed according to the account type. For example, when the processing device determines that the account type of the node is an abnormal account, the feature value transfer function of the node may be suspended, thereby preventing the asset loss of the user.

In order to accurately determine the account type of the node to be recognized through the classification model, the processing device can perform more detailed model training on the classification model to improve the model precision. In order to improve the training effect, the processing device can train by using the community features determined by the community feature extraction model to obtain the classification model.

In one possible implementation manner, the processing device may determine the sample community feature of the sample node according to the first training sample and the community feature extraction model. The processing device may then determine a classification model and a corresponding second training sample that includes the sample account features and the sample community features of the sample node, and the account type of the sample node.

The processing device can perform supervision training on the classification model through the second training sample to obtain the classification model, and the classification model is used for determining the account type corresponding to the node to be recognized based on the account characteristics and the community characteristics corresponding to the node to be recognized. In the process of supervision training, the processing equipment can input the second training sample into the classification model, analyze the account number type output by the classification model, compare whether the account number type is consistent with the account number type of the sample node in the second training sample, and feed back the comparison result to the classification model, so that the classification model can adjust the parameters in the model in a targeted manner, and the output account number type is close to the actual account number type of the sample node as much as possible. After the training is finished, the trained classification model can be used as the classification model to determine the account type.

In the related art, there are actually other determination methods of community characteristics. For example, the communities to which the nodes belong may be encoded by using a bag-of-words model (one-hot), the communities belonging to a certain community are encoded as 1, otherwise the communities are encoded as 0, and the method is simple, but causes dimension explosion and feature sparseness, and millions of communities may be generated in a relational network comprising hundreds of millions of nodes, which means millions of dimensions, and is not preferable. In addition, there are also Community algorithms for directly researching Community feature vectors (Community embedding), but these algorithms need to combine or unify Node vector (Node embedding) algorithms and clustering algorithms, and these algorithms are coupled together to generate high complexity, excessive parameters and poor practicability.

As shown in the following table, the following table provides a comparison of various community feature extraction methods:

as can be seen from the table, the algorithm provided by the embodiment of the application can better reflect the community characteristics of the nodes, can avoid the problem of complex iteration caused by node vectors, and can obtain a community feature extraction model with a good effect only through supervision training by using a mature neural network model technology.

Next, a method for determining community characteristics based on a relationship network provided in the embodiment of the present application will be introduced in combination with an actual application scenario.

Referring to fig. 3, fig. 3 is a schematic diagram of a community feature determination method based on a relationship network in an actual application scenario according to an embodiment of the present application. In the practical application scenario, the neural network model may be an MLP model with three hidden layers, and the processing device is a server capable of performing feature extraction. Fig. 3 shows a training process of the community feature extraction model and the classification model for account type classification in the practical application scenario. First, the server may determine a sample node identifier as a sample, and then determine, based on the sample node identifier, a home community corresponding to the sample node by using a community algorithm such as SLPA, Copra, or the like.

Subsequently, the server may determine N communities with higher attribution probability from the multiple attribution communities, and determine community identifications of the communities, which are denoted as Top1_ comID and Top1_ comID. The server can normalize the community identifications, then takes the account number type corresponding to the sample node as a training label, takes the normalized community identifications as input, and supervises training to obtain a community feature extraction model. The output layer of the community feature extraction model is the last hidden layer of the original MLP model, the input layer of the original MLP model is the input layer of the original MLP model, and the community feature extraction model comprises a model part between the input layer of the original MLP and the last hidden layer.

After obtaining the community feature extraction model, the server may determine the community features corresponding to the sample node by using the model. Subsequently, the server may use the community feature and other features corresponding to the sample node, such as account features, and the account type corresponding to the sample node, as training samples, and perform supervised training on the classification model to obtain a classification model for determining the account type of the node according to a plurality of features.

Part of the training formula in the model is as follows:

i∈NVector_community＝NN_MLP(Vector_input)

InputFeature_{classification}＝Vector_community||Vector_{other_feature}

wherein, Vector_inputRepresenting the input feature vector, Norm (Top _ Communnyids) representing the normalized vector of the respective community identity,

indicating that normalized community identifications are determined based on each community identification in TopN communities and the maximum value in the belonged community identification, wherein N is the number of the TopN communities, i is any one of the N communities, and Vector_communityRepresenting characteristics of communities, NN_MLP(Vector_input) Indicating the processing of input feature vectors by the MLP model, Inputfeature_{classification}Input vectors, representing classification models_community||Vector_{other_feature}And the joint vector representing the community feature vector and other features, wherein the other features can be account features and the like.

After the community feature extraction model and the classification model are obtained through training, in the application process, as shown in fig. 4, a server can obtain a node to be recognized in a relationship network with the same type as a sample node network, a community belonging is obtained through a community algorithm based on the node to be recognized, then the community feature of the node to be recognized is determined through the community feature extraction model, and finally the community feature, other features and the node to be recognized of the node to be recognized are input into the classification model to determine the account type corresponding to the node to be recognized.

The method for determining the account type through the community characteristics is compared with other methods, and the comparison result is shown in the following table:

Method	AUC value	Accuracy of measurement	Recall rate	F1 value	Accuracy of
						A	91.15％	72.17％	92.40％	81.04％	79.59％
B	75.30％	54.45％	87.34％	67.08％	59.52％
						C	75.39％	62.78％	84.81％	72.15％	69.08％
D	74.87％	52.66％	92.94％	67.23％	57.21％
						A+B	91.64％	72.35％	92.76％	81.29％	79.84％
A+C	93.04％	78.26％	91.13％	83.85％	84.21％
						A+D	93.35％	79.15％	91.32％	84.54％	84.80％

The method A is to determine the account type by using account characteristics of the nodes, the method B is to determine the account type by using node vectors, the method C is to determine the account type by performing weighted aggregation through a community algorithm and the node vectors, and the method D is to determine the account type by combining the community algorithm and a neural network model. According to the table, all data of the method A + D are the highest or higher data in all methods, namely parameters such as account type accuracy and the like determined by combining the community characteristics obtained by the technical scheme of the application and the account characteristics of the nodes are at a higher level, and therefore the community characteristics obtained by the technical scheme of the application can be better utilized in an application scene needing account type analysis.

Based on the method for determining community characteristics based on the relationship network described in the foregoing embodiment, the present application further provides a device for determining community characteristics based on the relationship network, as shown in fig. 5, fig. 5 is a block diagram of a device 500 for determining community characteristics based on the relationship network provided in the embodiment of the present application, where the device 500 includes a first determining unit 501, a first training unit 502, a second determining unit 503, and a third determining unit 504:

a first determining unit 501, configured to determine a neural network model and a corresponding first training sample, where the first training sample includes a community identifier of a community to which a sample node belongs in a sample relationship network, and an account type of the sample node;

a first training unit 502, configured to perform supervised training on the neural network model through the first training sample, and determine a community feature extraction model based on a hidden layer of the trained neural network model;

a second determining unit 503, configured to determine multiple communities to which nodes to be identified belong in the relationship network;

a third determining unit 504, configured to determine, according to the community identifiers of the multiple communities, a community feature corresponding to the node to be identified through the community feature extraction model, where the community feature is used to identify a network relationship that the node to be identified has in the relationship network.

In a possible implementation manner, the neural network model includes an input layer, an output layer, and k hidden layers located between the input layer and the output layer, and the first training unit 502 is specifically configured to:

determining a target hidden layer from the k hidden layers;

generating the community feature extraction model based on a model structure of the neural network model between the input layer and the target hidden layer, wherein an output layer of the community feature extraction model is the target hidden layer.

In a possible implementation manner, the neural network model is a multi-layer perceptron model, and the first training unit 502 is specifically configured to:

and taking the hidden layer closest to the output layer in the k hidden layers as the target hidden layer.

In a possible implementation manner, the third determining unit 504 is specifically configured to:

determining the attribution probability of the nodes to be identified corresponding to the plurality of communities respectively;

determining a target community with the attribution probability larger than a threshold condition from the plurality of communities according to the attribution probability;

and determining the community characteristics corresponding to the node to be identified through the community characteristic extraction model according to the community identification of the target community.

In one possible implementation, the apparatus 500 further includes a fourth determining unit:

and the fourth determining unit is used for determining the account type of the node to be identified according to the account characteristics corresponding to the node to be identified and the community characteristics.

In one possible implementation, the apparatus 500 further includes a fifth determining unit, a sixth determining unit, and a second training unit:

a fifth determining unit, configured to determine a sample community feature of the sample node according to the first training sample and the community feature extraction model;

a sixth determining unit, configured to determine a classification model and a corresponding second training sample, where the second training sample includes a sample account feature and a sample community feature of the sample node, and an account type of the sample node;

The embodiment of the present application further provides a computer device, which is described below with reference to the accompanying drawings. Referring to fig. 6, an embodiment of the present application provides a device, which may also be a terminal device, where the terminal device may be any intelligent terminal including a mobile phone, a tablet computer, a Personal Digital Assistant (PDA), a Point of Sales (POS), a vehicle-mounted computer, and the terminal device is taken as the mobile phone as an example:

fig. 6 is a block diagram illustrating a partial structure of a mobile phone related to a terminal device provided in an embodiment of the present application. Referring to fig. 6, the handset includes: a Radio Frequency (RF) circuit 610, a memory 620, an input unit 630, a display unit 640, a sensor 650, an audio circuit 660, a wireless fidelity (WiFi) module 670, a processor 680, and a power supply 690. Those skilled in the art will appreciate that the handset configuration shown in fig. 6 is not intended to be limiting and may include more or fewer components than those shown, or some components may be combined, or a different arrangement of components.

The following describes each component of the mobile phone in detail with reference to fig. 6:

the RF circuit 610 may be used for receiving and transmitting signals during information transmission and reception or during a call, and in particular, receives downlink information of a base station and then processes the received downlink information to the processor 680; in addition, the data for designing uplink is transmitted to the base station. In general, RF circuit 610 includes, but is not limited to, an antenna, at least one Amplifier, a transceiver, a coupler, a Low Noise Amplifier (LNA), a duplexer, and the like. In addition, the RF circuitry 610 may also communicate with networks and other devices via wireless communications. The wireless communication may use any communication standard or protocol, including but not limited to Global System for Mobile communication (GSM), General Packet Radio Service (GPRS), Code Division Multiple Access (CDMA), Wideband Code Division Multiple Access (WCDMA), Long Term Evolution (LTE), email, Short Message Service (SMS), and the like.

The memory 620 may be used to store software programs and modules, and the processor 680 may execute various functional applications and data processing of the mobile phone by operating the software programs and modules stored in the memory 620. The memory 620 may mainly include a storage program area and a storage data area, wherein the storage program 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 (such as audio data, a phonebook, etc.) created according to the use of the cellular phone, and the like. Further, the memory 620 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.

The input unit 630 may be used to receive input numeric or character information and generate key signal inputs related to user settings and function control of the cellular phone. Specifically, the input unit 630 may include a touch panel 631 and other input devices 632. The touch panel 631, also referred to as a touch screen, may collect touch operations of a user (e.g., operations of the user on the touch panel 631 or near the touch panel 631 by using any suitable object or accessory such as a finger or a stylus) thereon or nearby, and drive the corresponding connection device according to a preset program. Alternatively, the touch panel 631 may include two parts of a touch detection device and a touch controller. The touch detection device detects the touch direction of a user, detects a signal brought by touch operation and transmits the signal to the touch controller; the touch controller receives touch information from the touch sensing device, converts the touch information into touch point coordinates, sends the touch point coordinates to the processor 680, and can receive and execute commands sent by the processor 680. In addition, the touch panel 631 may be implemented using various types, such as resistive, capacitive, infrared, and surface acoustic wave. The input unit 630 may include other input devices 632 in addition to the touch panel 631. In particular, other input devices 632 may include, but are not limited to, one or more of a physical keyboard, function keys (such as volume control keys, switch keys, etc.), a trackball, a mouse, a joystick, and the like.

The display unit 640 may be used to display information input by the user or information provided to the user and various menus of the mobile phone. The Display unit 640 may include a Display panel 641, and optionally, the Display panel 641 may be configured in the form of a Liquid Crystal Display (LCD), an Organic Light-Emitting Diode (OLED), or the like. Further, the touch panel 631 can cover the display panel 641, and when the touch panel 631 detects a touch operation thereon or nearby, the touch panel is transmitted to the processor 680 to determine the type of the touch event, and then the processor 680 provides a corresponding visual output on the display panel 641 according to the type of the touch event. Although in fig. 6, the touch panel 631 and the display panel 641 are two independent components to implement the input and output functions of the mobile phone, in some embodiments, the touch panel 631 and the display panel 641 may be integrated to implement the input and output functions of the mobile phone.

The handset may also include at least one sensor 650, such as a light sensor, motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor that adjusts the brightness of the display panel 641 according to the brightness of ambient light, and a proximity sensor that turns off the display panel 641 and/or the backlight when the mobile phone is moved to the ear. As one of the motion sensors, the accelerometer sensor can detect the magnitude of acceleration in each direction (generally, three axes), can detect the magnitude and direction of gravity when stationary, and can be used for applications of recognizing the posture of a mobile phone (such as horizontal and vertical screen switching, related games, magnetometer posture calibration), vibration recognition related functions (such as pedometer and tapping), and the like; as for other sensors such as a gyroscope, a barometer, a hygrometer, a thermometer, and an infrared sensor, which can be configured on the mobile phone, further description is omitted here.

Audio circuitry 660, speaker 661, and microphone 662 can provide an audio interface between a user and a cell phone. The audio circuit 660 may transmit the electrical signal converted from the received audio data to the speaker 661, and convert the electrical signal into an audio signal through the speaker 661 and output the audio signal; on the other hand, the microphone 662 converts the collected sound signals into electrical signals, which are received by the audio circuit 660 and converted into audio data, which are processed by the audio data output processor 680 and then transmitted via the RF circuit 610 to, for example, another cellular phone, or output to the memory 620 for further processing.

WiFi belongs to short-distance wireless transmission technology, and the mobile phone can help a user to receive and send e-mails, browse webpages, access streaming media and the like through the WiFi module 670, and provides wireless broadband Internet access for the user. Although fig. 6 shows the WiFi module 670, it is understood that it does not belong to the essential constitution of the handset, and can be omitted entirely as needed within the scope not changing the essence of the invention.

The processor 680 is a control center of the mobile phone, and connects various parts of the entire mobile phone by using various interfaces and lines, and performs various functions of the mobile phone and processes data by operating or executing software programs and/or modules stored in the memory 620 and calling data stored in the memory 620, thereby performing overall monitoring of the mobile phone. Optionally, processor 680 may include one or more processing units; preferably, the processor 680 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 processor 680.

The handset also includes a power supply 690 (e.g., a battery) for powering the various components, which may preferably be logically connected to the processor 680 via a power management system, such that the power management system may be used to manage charging, discharging, and power consumption.

Although not shown, the mobile phone may further include a camera, a bluetooth module, etc., which are not described herein.

In this embodiment, the processor 680 included in the terminal device further has the following functions:

Referring to fig. 7, fig. 7 is a block diagram of a server 700 provided in this embodiment, and the server 700 may have a relatively large difference due to different configurations or performances, and may include one or more Central Processing Units (CPUs) 722 (e.g., one or more processors) and a memory 732, and one or more storage media 730 (e.g., one or more mass storage devices) storing an application program 742 or data 744. Memory 732 and storage medium 730 may be, among other things, transient storage or persistent storage. The program stored in the storage medium 730 may include one or more modules (not shown), each of which may include a series of instruction operations for the server. Further, the central processor 722 may be configured to communicate with the storage medium 730, and execute a series of instruction operations in the storage medium 730 on the server 700.

The server 700 may also include one or more power supplies 726, one or more wired or wireless network interfaces 750, one or more input-output interfaces 758, and/or one or more operating systems 741, such as Windows Server, Mac OS XTM, UnixTM, LinuxTM, FreeBSDTM, and so forth.

The steps performed by the server in the above embodiments may be based on the server structure shown in fig. 7.

The embodiment of the present application further provides a computer-readable storage medium, configured to store a computer program, where the computer program is configured to execute any one implementation of the relationship network-based community characteristic determination method described in the foregoing embodiments.

Those of ordinary skill in the art will understand that: all or part of the steps for realizing the method embodiments can be completed by hardware related to program instructions, the program can be stored in a computer readable storage medium, and the program executes the steps comprising the method embodiments when executed; and the aforementioned storage medium may be at least one of the following media: various media that can store program codes, such as read-only memory (ROM), RAM, magnetic disk, or optical disk.

It should be noted that, in the present specification, each embodiment is described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus and system embodiments, since they are substantially similar to the method embodiments, they are described in a relatively simple manner, and reference may be made to some of the descriptions of the method embodiments for related points. The above-described embodiments of the apparatus and system are merely illustrative, and the units described as separate parts may or may not be physically separate, and the parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment. One of ordinary skill in the art can understand and implement it without inventive effort.

The above description is only one specific embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present application should be covered by the scope of the present application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.

Claims

1. A community characteristic determination method based on a relationship network is characterized by comprising the following steps:

2. The method of claim 1, wherein the neural network model comprises an input layer, an output layer, and k hidden layers between the input layer and the output layer, and wherein determining the community feature extraction model based on the trained hidden layers of the neural network model comprises:

determining a target hidden layer from the k hidden layers;

3. The method of claim 2, wherein the neural network model is a multi-layer perceptron model, and wherein determining a target hidden layer from the k hidden layers comprises:

4. The method according to claim 1, wherein the determining, according to the community identifiers of the plurality of communities, the community features corresponding to the node to be identified through the community feature extraction model includes:

5. The method according to any one of claims 1-4, further comprising:

and determining the account type of the node to be identified according to the account characteristics corresponding to the node to be identified and the community characteristics.

6. The method according to claim 5, wherein the determining the account type of the node to be identified according to the account characteristic corresponding to the node to be identified and the community characteristic comprises:

7. The method of claim 6, further comprising:

determining sample community characteristics of the sample nodes according to a first training sample and the community characteristic extraction model;

determining a classification model and a corresponding second training sample, wherein the second training sample comprises sample account number characteristics and sample community characteristics of the sample nodes, and account number types of the sample nodes;

and carrying out supervision training on the classification model through the second training sample to obtain the classification model.

8. The method according to claim 5, wherein the relationship network is a feature value relationship network generated based on feature value transfer, and determining the account type of the node to be identified according to the account feature corresponding to the node to be identified and the community feature comprises:

9. A community feature determination device based on a relationship network is characterized by comprising a first determination unit, a first training unit, a second determination unit and a third determination unit:

10. The apparatus of claim 9, wherein the neural network model comprises an input layer, an output layer, and k hidden layers between the input layer and the output layer, and wherein the first training unit is specifically configured to:

determining a target hidden layer from the k hidden layers;

11. The apparatus of claim 10, wherein the neural network model is a multi-layered perceptron model, and wherein the first training unit is specifically configured to:

12. The apparatus according to claim 9, wherein the third determining unit is specifically configured to:

13. The apparatus according to any of claims 9-12, characterized in that the apparatus further comprises a fourth determination unit:

14. A computer device, the device comprising a processor and a memory:

the processor is configured to execute the method for determining community characteristics based on the relationship network according to any one of claims 1 to 8 according to instructions in the program code.

15. A computer-readable storage medium for storing a computer program for executing the method for determining community characteristics based on a relationship network according to any one of claims 1 to 8.