A prediction method and device for business interaction network

Abstract

The present disclosure relates to the field of data processing technology, and in particular to a prediction method and device for a business interaction network, which solves the problem that the established business interaction network is incomplete and that potential business relationships cannot be predicted. The method is as follows: acquiring each collected business data, establishing an interaction network, and then establishing an adjacency matrix and a feature matrix, and then inputting the adjacency matrix and the feature matrix into a prediction model to obtain a business interaction prediction matrix, and then processing the elements in the business interaction prediction matrix to generate a business interaction network matrix, and determining the predicted business interaction network based on the business interaction network matrix. In this way, the interaction network generated by the business data is used to predict the business interaction network, avoiding the disadvantage that when establishing a predicted interaction network, it is necessary to retrieve and determine all interaction objects of the same business type, making the established business interaction network incomplete, and can predict interaction objects with potential interactions.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing technology, and in particular to a prediction method and device for a business interaction network. Background Art

[0002] At present, due to the development and interaction of the economy and society, close business dealings exist between various enterprises, and usually a business interaction network is established to determine the supply and demand relationship between the interaction objects.

[0003] Under the existing technology, when establishing a business interaction network, it is usually necessary to first build a business type framework from the perspective of business type, and then determine the interaction objects corresponding to different types of business, and then establish the business interaction network. For example, taking the construction of an industrial chain network as an example, it is usually necessary to build a commodity industrial chain framework from the perspective of supplied goods, and then obtain the enterprise information corresponding to the commodity chain of each link in the commodity industrial chain through retrieval and matching. After large-scale retrieval and matching, the enterprise industrial chain corresponding to the commodity industrial chain is improved.

[0004] However, in this way, the generation of the business interaction network depends on the retrieval and matching of custom rules, which cannot avoid the problem of missing information, making it impossible to fully list all the interaction objects existing on nodes of different business types. Moreover, the established business interaction network has poor flexibility and cannot predict potential business relationships, making it impossible to determine the potential supply and demand relationship between interaction objects based on the established business interaction network. Summary of the invention

[0005] The embodiments of the present disclosure provide a method and device for predicting a business interaction network, so as to solve the problems in the prior art that the established business interaction network is incomplete and potential business relationships cannot be predicted.

[0006] The specific technical solutions provided by the embodiments of the present disclosure are as follows:

[0007] Firstly, a prediction method of a business interaction network is proposed, comprising:

[0008] Acquire each collected business data, and establish a corresponding interaction network according to each interaction object associated with each business data and the identification information of each interaction object, wherein the interaction network is used to characterize the business relationship and business value between each interaction object;

[0009] According to the business relationship between each interactive object in the interactive network, a corresponding adjacency matrix is ​​established, and according to the business value of each interactive object for different businesses in the interactive network, a corresponding feature matrix is ​​established, wherein the element representing the existence of the business relationship in the adjacency matrix takes a first set value, and the element representing the absence of the business relationship takes a second set value, and the number of column elements in the feature matrix is ​​positively correlated with the total number of business types;

[0010] Inputting the adjacency matrix and the feature matrix into the trained prediction model to obtain a business interaction prediction matrix output by the prediction model, wherein the elements in the business interaction prediction matrix represent the probability of a business relationship between each interaction object in the predicted business interaction network;

[0011] The elements whose element values ​​in the business interaction prediction matrix are not lower than the set threshold value are set to the first set value, and the elements whose element values ​​are lower than the set threshold value are set to the second set value, to generate a business interaction network matrix, and to determine the predicted business interaction network based on the business interaction network matrix.

[0012] Optionally, after the corresponding interaction network is established and before the corresponding adjacency matrix is ​​established according to the business relationship between the interaction objects in the interaction network, the method further includes:

[0013] Merge the business data of the same business type between any two interaction objects in the interaction network, and for any one of the interaction objects included in the interaction network, obtain the first-category business values ​​associated with the interaction object when it is a service requester, and obtain the second-category business values ​​associated with the interaction object when it is a service provider; and, in the interaction network, retain the first-category business value and the second-category business value with the largest value when the quantity accounts for a set ratio value.

[0014] Optionally, establishing a corresponding adjacency matrix according to the business relationship between each interaction object in the interaction network includes:

[0015] Obtain each interactive object existing in the interactive network, sort the each interactive object in a random order, and establish an adjacency matrix representing the business relationship between the each interactive object based on the sorted each interactive object, wherein the elements representing two interactive objects having a business relationship in the adjacency matrix have the same value, and any row vector in the adjacency matrix represents the business relationship between an interactive object and each interactive object.

[0016] Optionally, in the adjacency matrix and the feature matrix, the interaction objects corresponding to the row vectors with the same relative position are the same.

[0017] Optionally, before inputting the adjacency matrix and the feature matrix into the trained industrial chain prediction model, the step further includes training the industrial chain prediction model, including:

[0018] Obtaining a training sample set, wherein a training sample includes an adjacency matrix sample and a feature matrix sample;

[0019] Read the training samples and perform the following operations respectively until the industrial chain prediction model meets the preset convergence conditions:

[0020] Inputting the adjacency matrix samples and the feature matrix samples into the constructed industrial chain prediction model to obtain the industrial chain prediction matrix output by the industrial chain prediction model, wherein the industrial chain prediction model is constructed based on the GVAE technology;

[0021] Using the preset cross entropy loss function and Gaussian distribution function, in accordance with the Gaussian distribution, the loss value of the industrial chain prediction model is calculated according to the difference in element values ​​between the prediction chain prediction matrix and the adjacency matrix samples, and based on the loss value, the parameters for generating the industrial chain prediction matrix are adjusted.

[0022] Optionally, further including:

[0023] For each row vector of the industrial chain prediction matrix, determine the probability that a business relationship exists between an interaction object represented by the row vector and each interaction object; and filter out each element in the row vector whose element value reaches a set threshold, and set the business relationship determined by each element as a strongly associated business relationship.

[0024] In a second aspect, a prediction device for a business interaction network is proposed, comprising:

[0025] An acquisition unit, used to acquire each collected business data, and establish a corresponding interaction network according to each interaction object associated with each business data and the identification information of each interaction object, wherein the interaction network is used to characterize the business relationship and business value between each interaction object;

[0026] An establishing unit is used to establish a corresponding adjacency matrix according to the business relationship between each interactive object in the interactive network, and to establish a corresponding feature matrix according to the business value of each interactive object for different businesses in the interactive network, wherein the element in the adjacency matrix representing the existence of a business relationship takes a first set value, and the element representing the absence of a business relationship takes a second set value, and the number of column elements in the feature matrix is ​​positively correlated with the total number of business types;

[0027] An input unit, used to input the adjacency matrix and the feature matrix into the trained prediction model to obtain a business interaction prediction matrix output by the prediction model, wherein the elements in the business interaction prediction matrix represent the probability of a business relationship between each interaction object in the predicted business interaction network;

[0028] The processing unit is used to set the elements in the business interaction prediction matrix whose element values ​​are not lower than the set threshold value to the first set value, and set the elements whose element values ​​are lower than the set threshold value to the second set value, generate a business interaction network matrix, and determine the predicted business interaction network based on the business interaction network matrix.

[0029] Optionally, after the corresponding interaction network is established and before the corresponding adjacency matrix is ​​established according to the business relationship between the interaction objects in the interaction network, the establishing unit is further configured to:

[0030] Merge the business data of the same business type between any two interaction objects in the interaction network, and for any one of the interaction objects included in the interaction network, obtain the first-category business values ​​associated with the interaction object when it is a service requester, and obtain the second-category business values ​​associated with the interaction object when it is a service provider; and, in the interaction network, retain the first-category business value and the second-category business value with the largest value when the quantity accounts for a set ratio value.

[0031] Optionally, when establishing a corresponding adjacency matrix according to the business relationship between each interaction object in the interaction network, the establishing unit is used to:

[0032] Obtain each interactive object existing in the interactive network, sort the each interactive object in a random order, and establish an adjacency matrix representing the business relationship between the each interactive object based on the sorted each interactive object, wherein the elements representing two interactive objects having a business relationship in the adjacency matrix have the same value, and any row vector in the adjacency matrix represents the business relationship between an interactive object and each interactive object.

[0033] Optionally, in the adjacency matrix and the feature matrix, the interaction objects corresponding to the row vectors with the same relative position are the same.

[0034] Optionally, before inputting the adjacency matrix and the feature matrix into the trained industrial chain prediction model, the step further includes training the industrial chain prediction model, wherein the input unit is used to:

[0035] Obtaining a training sample set, wherein a training sample includes an adjacency matrix sample and a feature matrix sample;

[0036] Read the training samples and perform the following operations respectively until the industrial chain prediction model meets the preset convergence conditions:

[0037] Inputting the adjacency matrix samples and the feature matrix samples into the constructed industrial chain prediction model to obtain the industrial chain prediction matrix output by the industrial chain prediction model, wherein the industrial chain prediction model is constructed based on the GVAE technology;

[0038] Using the preset cross entropy loss function and Gaussian distribution function, in accordance with the Gaussian distribution, the loss value of the industrial chain prediction model is calculated according to the difference in element values ​​between the prediction chain prediction matrix and the adjacency matrix samples, and based on the loss value, the parameters for generating the industrial chain prediction matrix are adjusted.

[0039] Optionally, the processing unit is further configured to:

[0040] For each row vector of the industrial chain prediction matrix, determine the probability that a business relationship exists between an interaction object represented by the row vector and each interaction object; and filter out each element in the row vector whose element value reaches a set threshold, and set the business relationship determined by each element as a strongly associated business relationship.

[0041] In a third aspect, an electronic device is provided, comprising:

[0042] A memory for storing executable instructions;

[0043] The processor is used to read and execute the executable instructions stored in the memory to implement any of the above-mentioned methods for predicting a business interaction network.

[0044] In a fourth aspect, a computer-readable storage medium is proposed. When instructions in the storage medium are executed by an electronic device, the electronic device is enabled to execute any of the above-mentioned methods for predicting a business interaction network.

[0045] The beneficial effects of the present disclosure are as follows:

[0046] In the embodiment of the present disclosure, each collected business data is obtained, and the business provider and business requester included in the business data are determined as interaction objects, and a corresponding interaction network is established according to the identification information of each interaction object, wherein the interaction network represents the business relationship and business value between the interaction objects, and then for any interaction object among the interaction objects in the interaction network, the corresponding business requesters when the interaction object is a business provider are determined, and the business values ​​of each business type provided by the interaction object are respectively determined; an adjacency matrix representing the business relationship between the business provider and the corresponding business requester is established, and according to the business values ​​of different business types provided by each interaction object, an adjacency matrix representing the business relationship between the business provider and the corresponding business requester is established. The corresponding feature matrix is ​​then input into the trained prediction model, and the business interaction prediction matrix output by the prediction model is obtained, wherein the elements in the business interaction prediction matrix represent the probability of the existence of a business relationship between the various interaction objects in the predicted business interaction network, and then the elements whose element values ​​in the business interaction prediction matrix are not lower than the set threshold value are set to the values ​​in the adjacency matrix representing the existence of the business relationship, and the elements whose element values ​​are lower than the set threshold value are set to the values ​​in the adjacency matrix representing the absence of the business relationship, thereby generating a business interaction network matrix, and determining the predicted business interaction network based on the business interaction network matrix. In this way, an interaction network is gradually constructed based on each interaction object included in the business data, and the business interaction network is predicted based on the adjacency matrix and feature matrix extracted from the interaction network. The interaction objects are used as the basis for building the interaction network, and then the business interaction network is predicted based on the interaction network. This avoids the disadvantage that when establishing a predicted interaction network, it is necessary to retrieve and determine all interaction objects of the same business type, which makes the established business interaction network incomplete. In addition, with the help of the prediction model, the characteristics of the adjacency matrix and the feature matrix can be integrated to generate a business interaction prediction matrix, and the probability of a business relationship between two interaction objects can be predicted respectively. Then, after processing based on the business interaction prediction matrix, a business interaction network matrix that can generate a corresponding business interaction network can be obtained. With the help of the prediction of the prediction model, interaction objects with potential interactions can be predicted. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] Figure 1 A schematic diagram of a process of predicting a service interaction network in an embodiment of the present disclosure;

[0048] Figure 2a A schematic diagram of connections between different interactive objects created based on business data in an embodiment of the present disclosure;

[0049] Figure 2b This is a schematic diagram of simplified connection lines between different interactive objects in an embodiment of the present disclosure;

[0050] Figure 2c is a schematic diagram of an interactive network after processing in an embodiment of the present disclosure;

[0051] Figure 3a A schematic diagram of a service relationship determined according to an interactive network in an embodiment of the present disclosure;

[0052] Figure 3b A schematic diagram of digitizing a business relationship schematic diagram in an embodiment of the present disclosure;

[0053] Figure 3c This is a schematic diagram after assigning specific numerical values ​​to the business relationship schematic diagram in the embodiment of the present disclosure;

[0054] Figure 4 A schematic diagram of a feature matrix established in an embodiment of the present disclosure;

[0055] Figure 5 It is a schematic diagram of the internal processing of the prediction matrix in the embodiment of the present disclosure;

[0056] Figure 6 A schematic diagram of a process for predicting an industrial chain network in an embodiment of the present disclosure;

[0057] Figure 7 A schematic diagram of the logical structure of the business interaction network prediction in the embodiment of the present disclosure;

[0058] Figure 8 A schematic diagram of the entity structure of the business interaction network prediction in the embodiment of the present disclosure. DETAILED DESCRIPTION

[0059] In order to make the purpose, technical solution and beneficial effects of the present disclosure more clear, the present disclosure is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present disclosure and are not used to limit the present disclosure.

[0060] Those skilled in the art will appreciate that the embodiments of the present disclosure may be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure may be specifically implemented in the following forms, namely: complete hardware, complete software (including firmware, resident software, microcode, etc.), or a combination of hardware and software.

[0061] In order to solve the problems in the prior art that the established business interaction network is incomplete and the potential business relationship cannot be predicted, the present disclosure specifically proposes a business interaction network prediction method and device, specifically, obtaining each collected business data, determining the business provider and business requester included in the business data as interaction objects, and establishing a corresponding interaction network according to the identification information of each interaction object, wherein the interaction network represents the business relationship and business value between the interaction objects, and then for any one of the interaction objects in the interaction network, determining each corresponding business requester when the interaction object is a business provider, and respectively determining the business value of each business type provided by the interaction object; establishing a business relationship between the business provider and the corresponding business requester An adjacency matrix of the relationship, and a corresponding feature matrix are established according to the business values ​​of different business types provided by each interactive object. Then, the adjacency matrix and the feature matrix are input into the trained prediction model to obtain the business interaction prediction matrix output by the prediction model, wherein the elements in the business interaction prediction matrix represent the probability of the existence of a business relationship between each interactive object in the predicted business interaction network, and then the elements in the business interaction prediction matrix whose element values ​​are not lower than the set threshold value are set to the values ​​in the adjacency matrix representing the existence of the business relationship, and the elements whose element values ​​are lower than the set threshold value are set to the values ​​in the adjacency matrix representing the absence of the business relationship, thereby generating a business interaction network matrix, and determining the predicted business interaction network based on the business interaction network matrix.

[0062] It should be noted that the prediction of the business interaction network defined in the present disclosure may be repeated periodically, such as on a monthly basis, to obtain business data generated in a past period, and to predict the business interaction network based on the obtained business data.

[0063] In some embodiments of the present disclosure, it can be applied to the scenario where the tax authorities establish an industrial chain network prediction based on the collected tax data. The tax authorities determine the transaction types and transaction data between the various companies included in the tax data based on the collected tax data, and establish an interactive network based on the transaction data and transaction types of each company, and extract the adjacency matrix representing the transaction relationship between the companies and the feature matrix representing the transaction data between the companies from the interactive network, so as to predict the potential transaction relationship between the companies. Specifically, the tax data within the past cycle can be obtained on a monthly basis, and the industrial chain network between different companies can be predicted based on the obtained tax data.

[0064] The preferred implementation of the embodiment of the present disclosure is further described in detail below in conjunction with the accompanying drawings:

[0065] The following is combined with Figure 1, in the implementation of the present disclosure, the prediction process of the service interaction network by the processing device based on the obtained service data is described:

[0066] Step 101: The processing device obtains various collected service data, determines the service provider and service requester included in the service data as interaction objects, and establishes a corresponding interaction network according to the identification information of each interaction object.

[0067] Specifically, the processing device obtains the collected business data, and determines the business provider and business requester included in the business data as interaction objects, and establishes a corresponding interaction network based on the identification information of each interaction object, wherein one business data contains at least two interaction objects, namely the business requester and the business provider, and the interaction network is used to characterize the business relationship and business value between the interaction objects, the business relationship and business value are determined by the business data, the business relationship characterizes the interaction relationship between the business requester and the business provider, and the business value is used to value the business provided by the business provider.

[0068] It should be noted that, considering that there are multiple types of services between different interactive objects, and the same type of services between two interactive objects may be represented by multiple business data, in order to simplify the interactive network, the processing device merges the business data of the same business type between any two interactive objects in the interactive network, and for any interactive object among the interactive objects included in the interactive network, obtains the first-class business values ​​associated with the interactive object when it is a service requester, and obtains the second-class business values ​​associated with the interactive object when it is a service provider; and, in the interactive network, the first-class business value and the second-class business value with the largest value are retained when the quantity accounts for the set ratio value.

[0069] Specifically, after the processing device preliminarily establishes an interactive network based on the business data, it merges the business data between any two interactive objects with the same business type in the interactive network, so that between any two interactive objects in the interactive network, for the same business type, there is only one connection line, that is, it corresponds to only one merged business value, wherein the business type is used to limit the specific content of the transmission between the two interactive objects, and determines the business requester and the business provider.

[0070] For example, see Figure 2a As shown, it is assumed that according to the acquired business data, it is determined that there are three types of business interactions between the interaction object A and the interaction object B, and there are two business data corresponding to business type 1, three business data corresponding to business type 2, and two business data corresponding to business type 3. Therefore, after merging the business data of the same business type, it can be obtained Figure 2bFurther, after the above simplification of the interactive network, we can get Figure 2c The interactive network shown in the figure, wherein the direction of the arrow in the interactive network represents the business relationship. Taking the direction of the arrow between interactive object A and interactive object B as an example, the arrow from interactive object A to interactive object B illustrates the interactive relationship when interactive object A acts as a service provider and interactive object B acts as a service requester under a certain type of business.

[0071] For another example, taking the prediction scenario of the industrial chain network as an example, the processing equipment of the tax authority obtains various tax data, and establishes an interactive network based on the various companies associated with each tax data and the identification information of each company, which can be the taxpayer identification number. In the interactive network, any two enterprises that actually trade are connected in the form of directed edges, and the directed edges point from the seller enterprise to the buyer enterprise, and the business value is the amount of the actual transaction. Then, according to the content of the transaction between the enterprises, the directed edges representing the transaction of the same items between the two enterprises are merged to simplify the interactive network.

[0072] Furthermore, for any one of the interactive objects in the interactive network, a first service value associated with the interactive object when it serves as a service requester is determined, and a second service value associated with the interactive object when it serves as a service provider is determined, and in the interactive network, the first service value and the second service value with the largest value whose quantity ratio is a set ratio are retained, wherein the quantity ratio is specifically the ratio of the retained quantity to the total number of the same type. For example, assuming that the quantity ratio is 85%, and the total number of connection lines when the interactive object in the interactive network serves as a service provider is 100, the first 85 connection lines with the largest second service value are selected for storage.

[0073] In this way, with the help of the collected business data, an interaction network that represents the business relationships and business values ​​between various interaction objects can be established, avoiding the drawbacks of determining interaction objects through retrieval under the existing technology and ensuring the effectiveness of the subsequent business interaction network prediction process.

[0074] Step 102: The processing device determines, for any interactive object among the interactive objects in the interactive network, the corresponding service requesting parties when the interactive object acts as a service provider, and respectively determines the service value of each service type provided by the interactive object; establishes an adjacency matrix based on the service relationship between each interactive object and the corresponding service requesting party, and establishes a corresponding feature matrix according to the service values ​​of different service types provided by each interactive object.

[0075] After the processing device establishes the interactive network, the processing device determines, for any interactive object among the interactive objects in the interactive network, the corresponding service requesters when the interactive object is a service provider, and respectively determines the service value of each service type provided by the interactive object. Then, based on the service relationship between each interactive object and the corresponding service requester, an adjacency matrix is ​​established.

[0076] Specifically, the processing device obtains each interactive object existing in the interactive network, sorts the interactive objects in a random order, and then, for any interactive object among the interactive objects after the sorting is completed, obtains each business requester corresponding to the interactive object when the interactive object is used as a business provider, and establishes row elements based on the business relationship between the interactive object and the business requesters, and then arranges the row elements corresponding to the interactive objects according to the sorting and combines them into an adjacency matrix.

[0077] For example, continue with step 101 Figure 2c Taking an example, there are five interaction objects in the interaction network, namely interaction objects AE. When interaction object A acts as a service provider, interaction object B is the corresponding service requester; when interaction object B acts as a service provider, interaction object A and interaction object E act as service requesters; when interaction object C acts as a service provider, interaction object B and interaction object D act as service requesters; when interaction object D acts as a service provider, interaction object A acts as a service requester; when interaction object E acts as a service provider, interaction object D acts as a service requester.

[0078] See also Figure 3a-3c Indicated, such as Figure 3a As shown, assuming that the order of the interaction objects AE is followed, when the interaction objects AE are respectively used as service providers, the existing service relationships can be specifically shown in the black box part. For further information, refer to Figure 3b As shown, after the business relationship is numerically post-processed, the black frame part representing the existence of the business relationship is represented by the first set value, and the blank part representing the non-existence of the business relationship is represented by the second set value, wherein the first set value and the second set value are configured according to the actual situation, and the values ​​of the first set value and the second set value are different. When the first set value is set to 1 and the second set value is set to 0, it can be obtained Figure 3c The adjacency matrix shown.

[0079] Furthermore, the processing device constructs a one-hot feature matrix according to the service values ​​of different service types provided by each interactive object.

[0080] Specifically, the processing device establishes a feature matrix according to different types of business values ​​provided by each interactive object. The processing device obtains the corresponding business value of any interactive object among the objects when the interactive object acts as an object provider and an object requester, and establishes a row vector corresponding to the interactive object according to the correspondence between the interactive object and the business values ​​under different business types; and then combines the row vectors corresponding to the interactive objects to generate a feature matrix.

[0081] In the disclosed embodiment, the row vectors with the same relative position in the adjacency matrix and the feature matrix correspond to the same interaction objects, and the number of columns in the feature matrix is ​​positively correlated with the total number of service types that can be provided by each interaction object, specifically twice the total number of types, wherein the number is set to twice so that the service values ​​involved when the interaction object acts as a service provider and a service requester can be simultaneously reflected in the feature matrix.

[0082] It should be noted that when the present disclosure is applied to the prediction scenario of the industrial chain network, the number of columns of the generated feature matrix is ​​twice the total number of various commodity tax codes and service tax classification codes, which is used to characterize the corresponding transaction data when the company acts as the requester and provider. The commodity tax codes and service tax classification codes are uniformly formulated by the state.

[0083] For example, see Figure 4 As shown, N is the total number of interactive objects in the interactive network, M is the total number of business types, and X is the generated feature matrix. The total number of types refers to all objectively existing business types, including business types that exist between each interactive object and those that do not currently exist.

[0084] In this way, by establishing the adjacency matrix and the feature matrix, the business relationships existing in the interaction network and the business values ​​under different business relationships can be reflected in numerical form, realizing a matrix description of the interaction characteristics, and providing input data for the subsequent prediction model processing, so that the GVAE model can be used to predict the business interaction network.

[0085] Step 103: The processing device inputs the adjacency matrix and the feature matrix into the trained prediction model to obtain the business interaction prediction matrix output by the prediction model.

[0086] The processing device uses the trained prediction model to train the prediction model before processing the adjacency matrix and the feature matrix. The training process of the prediction model is described below:

[0087] A training sample set is obtained, wherein a training sample includes an adjacency matrix sample and a feature matrix sample.

[0088] Specifically, the processing device establishes a historical interaction network based on business data in different historical periods, and then determines the historical adjacency matrix and historical feature matrix corresponding to the historical interaction network, and uses the obtained historical adjacency matrix as an adjacency matrix sample, and uses the corresponding historical feature matrix as a feature matrix sample.

[0089] Furthermore, the processing device reads the training samples and performs the following operations respectively until the prediction model meets the preset convergence conditions:

[0090] S1: The processing device inputs the adjacency matrix samples and the feature matrix samples into the constructed prediction model to obtain the business interaction prediction matrix output by the prediction model, wherein the prediction model is constructed based on the Graph Variational Autoencoder (GVAE) technology.

[0091] See also Figure 5 In the processing process shown in the figure, the processing device inputs the feature matrix X and the adjacency matrix A into the prediction model, so that inside the prediction model, it passes through the encoder and decoder successively, and finally reconstructs the business interaction network. Among them, the encoder completes the encoding of the interactive object, and the decoder needs to generate edges based on the decoding of the interactive object. The encoder uses a multi-layer graph convolutional neural network (Graph Convolutional Network, GCN), which extracts the data features on the graph through GCN, and introduces Gaussian distribution as a distribution constraint to fine-tune the generated latent variable matrix Z. The decoder completes the model calculation by minimizing the objective function loss through the dot product of the node features and the activation function, and finally realizes the prediction function of the prediction model.

[0092] Specifically, when building a prediction model, the process of building an encoder is as follows: use a multi-layer GCN model to extract data features on the constructed network, complete the GCN modeling of the mean μ and standard deviation δ of the interactive object in the node feature matrix dimension, and finally implement the construction of the latent variable matrix Z by comprehensive random sampling (sample). Among them, GCN is a neural network layer, and its propagation method between layers is shown in the following formula:

[0093]

[0094] in, I is the identity matrix, A is the adjacency matrix, yes The degree matrix (degreeematrix), H is the feature of each layer, for the input layer, H = X, σ is the nonlinear activation function.

[0095] It should be noted that in the disclosed embodiment, the role of GCN is to extract the main features that determine the business relationship between the interactive objects, and eliminate redundant features, obtain reasoning ability, obtain clearer rules, and at the same time, the interactive object features can be integrated with other interactive object features. For example, taking the scenario of predicting the industrial chain network as an example, the upstream related enterprises of the enterprise selling photovoltaic cells tend to be enterprises in related industries selling crystal materials. After GCN predicts the link, for a directed edge, it must predict the downstream enterprises, and then potentially map the upstream and downstream industrial chain relationship of the enterprise.

[0096] The output Z of the encoder is a comprehensive latent variable matrix of the mean and standard deviation of each interactive object in the business type dimension under the probability distribution, that is, the standard deviation is multiplied by the randomly generated variable ε, and then added to the mean to obtain the result after Gaussian distribution upsampling. The specific processing process is detailed in the following formula:

[0097] μ=GCN μ (X,A)

[0098] σ=GCN σ (X,A)

[0099] Z=μ+ε×σ

[0100] Furthermore, the process of constructing the decoder is: perform node feature dot product on the latent variable matrix Z output by the encoder, and realize node relationship probability prediction through the activation function σ. Specifically, the decoder formula is as follows:

[0101] y′=δ(Z T *Z)

[0102] Among them, Z is the output of the encoder, and y' represents the calculated probability value.

[0103] S2: The processing device adopts a preset cross entropy loss function and Gaussian distribution function, and calculates the loss value of the prediction model according to the difference in element values ​​between the prediction chain prediction matrix and the adjacency matrix samples in accordance with the Gaussian distribution, and adjusts the parameters of the generated business interaction prediction matrix based on the loss value.

[0104] In the embodiment of the present disclosure, the definition of the loss value is as follows:

[0105] loss=CrossEntropy+KL[q(Z|X,A)||p(Z)]

[0106] Among them: KL divergence represents the difference between the distribution of Z and the Gaussian distribution; p(Z) represents a standard Gaussian distribution; q(Z|X,A) represents the target distribution; Cross Entropy represents the difference between A and A', A' represents the adjacency matrix reconstructed by the prediction model, where the adjacency matrix represents the constructed business interaction network structure. The specific calculation formula is as follows:

[0107]

[0108] Wherein, y represents the value (0 or 1) of an element in the adjacency matrix A, and y′ represents the probability value of the corresponding element in the reconstructed adjacency matrix A′.

[0109] Furthermore, after calculating the model loss value, the parameters of the business interaction prediction matrix generated in the prediction model are adjusted based on the loss value, and the training is determined to be completed when it is determined that the preset convergence condition is met. The convergence condition may be that the number of training times of the prediction model reaches the set number of training times, or the number of times the loss value is continuously lower than the set value reaches the preset threshold value.

[0110] Step 104: The processing device sets the values ​​of the elements in the business interaction prediction matrix that are not lower than the set threshold value to the values ​​in the adjacency matrix that represent the existence of business relationships, and sets the values ​​of the elements that are lower than the set threshold value to the values ​​in the adjacency matrix that represent the non-existence of business relationships, and then generates a business interaction network matrix, and determines the predicted business interaction network based on the business interaction network matrix.

[0111] Specifically, after the processing device obtains the business interaction prediction matrix, the elements whose element values ​​in the business interaction prediction matrix are not lower than the set threshold value are set to the values ​​in the adjacency matrix that represent the existence of business relationships, and the elements whose element values ​​are lower than the set threshold value are set to the values ​​in the adjacency matrix that represent the absence of business relationships, and then a business interaction network matrix is ​​generated, so that the elements in the business interaction network matrix and the elements in the adjacency matrix have the same meaning. When the adjacency matrix is ​​generated based on the interaction network, the business interaction matrix is ​​obtained by the reverse generation prediction corresponding to the business interaction network matrix.

[0112] It should be noted that, for the business interaction prediction matrix, the processing device can determine, for any row vector of the business interaction prediction matrix, the probability that a business relationship exists between an interaction object represented by the row vector and the respective interaction objects; and screen out the elements in the row vector whose element values ​​reach a set threshold, and set the business relationship determined by the elements as a strongly associated business relationship.

[0113] Specifically, after obtaining the business interaction prediction matrix, the processing device can, according to actual needs, filter the row vector that represents the probability that an interaction object, as a service provider, has a business relationship with other interaction objects according to a set threshold, and determine which other interaction objects, which are business requesters of the one interaction object, have a strongly associated business relationship with the one interaction object.

[0114] In the scenario of predicting the industrial chain network, after obtaining the business interaction prediction matrix, the elements of the business interaction matrix are filtered according to the set threshold. The set threshold can be configured according to actual needs, such as taking a value of 0.5. Further, the selling enterprise corresponding to each row element is determined, and each element in each row element with a probability value higher than 0.5 is determined, and the buying enterprise corresponding to each determined element is determined as a buying enterprise with a strong correlation with the selling enterprise. The strong correlation represents the possibility of transactions between different enterprises.

[0115] In this way, by combining deep learning and graph computing technology into the prediction of business interaction networks, it is possible to extract the features of business relationships between interaction objects, and realize feature fusion between interaction objects, so as to achieve deep mining of business interaction networks, greatly simplify the complexity of establishing traditional business interaction networks, greatly reduce manual intervention, and realize the prediction of potential business relationships.

[0116] The following is combined with Figure 6 , the tax authorities will explain the process of predicting the industrial chain network based on the tax data obtained within a certain period of time:

[0117] Step 601: Obtain tax data.

[0118] Specifically, in this scenario, the processing equipment of the tax authority obtains tax data generated within a period of time. The tax data may specifically be invoice data, wherein the invoice data includes unique identification information of both transaction companies, that is, the taxpayer identification number, and transaction information of both transaction companies.

[0119] Step 602: Establish an enterprise transaction network based on tax data.

[0120] The processing device establishes an enterprise transaction network based on the transaction relationship between the selling enterprise and the buying enterprise recorded in the invoice data.

[0121] Furthermore, the established enterprise transaction network is simplified by merging transaction data of the same type and deleting some transaction data.

[0122] For example, there are multiple transactions between the seller company A and the buyer company B for commodity X, that is, there are corresponding multiple invoice data. Therefore, in the enterprise transaction network, there is only one network line representing the transaction relationship between the seller company A and the buyer company B for commodity X.

[0123] For another example, assuming that enterprise A is the selling enterprise, it has transaction relationships with enterprises B, C, D, E, F, G, H, I, J, and K, and the corresponding transaction amounts are RMB 100,000, RMB 50,000, RMB 30,000, RMB 60,000, RMB 250,000, RMB 2,000, RMB 1,500, RMB 130,000, RMB 200,000, and RMB 360,000, respectively. Assuming that the set ratio value for quantity is 80%, if there are 10 enterprises in total that have transaction relationships with enterprise A, only transaction relationships with 8 enterprises are retained in the enterprise transaction network, that is, the 8 enterprises with the largest transaction amounts are selected, and the indication of transaction relationships is retained in the enterprise transaction network. Enterprise B: 100,000, enterprise C: 50,000, enterprise D: 30,000, enterprise E: 60,000, enterprise F: 250,000, enterprise G: 2,000, enterprise H: 1,500, enterprise I: 130,000, enterprise J: 200,000, enterprise K: 360,000. After sorting by transaction amount from large to small: Enterprise K: 360,000, enterprise F: 250,000, enterprise J: 200,000, enterprise I: 130,000, enterprise B: 100,000, enterprise E: 60,000, enterprise C: 50,000, enterprise D: 30,000, enterprise G: 2,000, enterprise H: 1,500. Therefore, the network lines corresponding to the transaction relationship between enterprises K, F, J, I, B, E, C, D and enterprise A are retained in the transaction network.

[0124] Step 603: Determine the adjacency matrix and feature matrix corresponding to the enterprise transaction network.

[0125] The processing device establishes a corresponding adjacency matrix and a characteristic matrix according to the established enterprise transaction network, wherein the establishment process of the adjacency matrix and the characteristic matrix has been described in Figure 1 The process is described in detail and will not be repeated here.

[0126] Step 604: input the adjacency matrix and the feature matrix into the trained prediction model to obtain the transaction prediction matrix output by the prediction model.

[0127] The processing device inputs the adjacency matrix and the feature matrix into the trained prediction model to obtain the transaction prediction matrix output by the prediction model. Each element in the transaction prediction matrix represents the probability of a transaction relationship between the corresponding enterprises. The specific generation process has been described in Figure 1 The companies corresponding to the row elements and column elements in the transaction prediction matrix are the same as the companies corresponding to the row elements and column elements in the adjacency matrix.

[0128] Step 605: Process the transaction prediction matrix to obtain an industrial chain network matrix, and generate an industrial chain network according to the industrial chain network matrix.

[0129] The processing device processes the transaction prediction matrix. Specifically, according to the element value of each element in the transaction prediction matrix, the element value of the element whose value is not less than the set threshold value is set to the value of the element in the adjacency matrix that represents the existence of the transaction relationship, and the element value whose value is less than the set threshold value is set to the element value of the element in the adjacency matrix that represents the absence of the transaction relationship, thereby establishing an industrial chain network matrix, and reversely establishing an industrial chain network based on the industrial chain network matrix.

[0130] Based on the same inventive concept, refer to Figure 7 As shown, in an embodiment of the present disclosure, a prediction device for a service interaction network is proposed, comprising:

[0131] The acquisition unit 701 is used to acquire each collected business data, and establish a corresponding interaction network according to each interaction object associated with each business data and the identification information of each interaction object, wherein the interaction network is used to represent the business relationship and business value between each interaction object;

[0132] An establishing unit 702 is used to establish a corresponding adjacency matrix according to the business relationship between each interactive object in the interactive network, and to establish a corresponding feature matrix according to the business value of each interactive object for different businesses in the interactive network, wherein the element in the adjacency matrix representing the existence of a business relationship takes a first set value, and the element representing the absence of a business relationship takes a second set value, and the number of column elements in the feature matrix is ​​positively correlated with the total number of business types;

[0133] An input unit 703 is used to input the adjacency matrix and the feature matrix into the trained prediction model to obtain a business interaction prediction matrix output by the prediction model, wherein the elements in the business interaction prediction matrix represent the probability of the existence of a business relationship between each interaction object in the predicted business interaction network;

[0134] The processing unit 704 is used to set the elements in the business interaction prediction matrix whose element values ​​are not lower than the set threshold value to the first set value, and set the elements whose element values ​​are lower than the set threshold value to the second set value, to generate a business interaction network matrix, and determine the predicted business interaction network based on the business interaction network matrix.

[0135] Optionally, after the corresponding interaction network is established and before the corresponding adjacency matrix is ​​established according to the business relationship between the interaction objects in the interaction network, the establishing unit 702 is further configured to:

[0136] Merge the business data of the same business type between any two interaction objects in the interaction network, and for any one of the interaction objects included in the interaction network, obtain the first-category business values ​​associated with the interaction object when it is a service requester, and obtain the second-category business values ​​associated with the interaction object when it is a service provider; and, in the interaction network, retain the first-category business value and the second-category business value with the largest value when the quantity accounts for a set ratio value.

[0137] Optionally, when establishing a corresponding adjacency matrix according to the business relationship between each interaction object in the interaction network, the establishing unit 702 is used to:

[0138] Obtain each interactive object existing in the interactive network, sort the each interactive object in a random order, and establish an adjacency matrix representing the business relationship between the each interactive object based on the sorted each interactive object, wherein the elements representing two interactive objects having a business relationship in the adjacency matrix have the same value, and any row vector in the adjacency matrix represents the business relationship between an interactive object and each interactive object.

[0139] Optionally, in the adjacency matrix and the feature matrix, the interaction objects corresponding to the row vectors with the same relative position are the same.

[0140] Optionally, before inputting the adjacency matrix and the feature matrix into the trained industrial chain prediction model, the step further includes training the industrial chain prediction model, and the input unit 703 is used to:

[0141] A training sample set is obtained, wherein a training sample includes an adjacency matrix sample and a feature matrix sample.

[0142] Read the training samples and perform the following operations respectively until the industrial chain prediction model meets the preset convergence conditions:

[0143] Inputting the adjacency matrix samples and the feature matrix samples into the constructed industrial chain prediction model to obtain the industrial chain prediction matrix output by the industrial chain prediction model, wherein the industrial chain prediction model is constructed based on the GVAE technology;

[0144] Using the preset cross entropy loss function and Gaussian distribution function, in accordance with the Gaussian distribution, the loss value of the industrial chain prediction model is calculated according to the difference in element values ​​between the prediction chain prediction matrix and the adjacency matrix samples, and based on the loss value, the parameters for generating the industrial chain prediction matrix are adjusted.

[0145] Optionally, the processing unit 704 is further configured to:

[0146] For each row vector of the industrial chain prediction matrix, determine the probability that a business relationship exists between an interaction object represented by the row vector and each interaction object; and filter out each element in the row vector whose element value reaches a set threshold, and set the business relationship determined by each element as a strongly associated business relationship.

[0147] Based on the same inventive concept, refer to Figure 8 As shown, the prediction device 800 of the service interaction network can be a server or a terminal device with processing function. Figure 8 , the apparatus 800 includes a processing component 822, which further includes one or more processors, and a memory resource represented by a memory 832 for storing instructions, such as an application, that can be executed by the processing component 822. The application stored in the memory 832 may include one or more modules, each corresponding to a set of instructions. In addition, the processing component 822 is configured to execute the instructions to perform the above method.

[0148] The device 800 may also include a power supply component 826 configured to perform power management of the device 800, a wired or wireless network interface 850 configured to connect the device 800 to a network, and an input / output (I / O) interface 858. The device 800 may operate based on an operating system stored in the memory 832, such as Windows Server™, Mac OS X™, Unix™, Linux™, FreeBSD™, or the like.

[0149] Based on the same inventive concept, a storage medium is provided in an embodiment of the present disclosure based on business interaction network prediction. When instructions in the storage medium are executed by an electronic device, the electronic device can execute any of the above methods.

[0150] In summary, in the embodiments of the present disclosure, each collected business data is obtained, the business provider and the business requester included in the business data are determined as interaction objects, and a corresponding interaction network is established according to the identification information of each interaction object, wherein the interaction network represents the business relationship and business value between the interaction objects, and then for any one of the interaction objects in the interaction network, the corresponding business requesters when the interaction object is a business provider are determined, and the business values ​​of each business type provided by the interaction object are respectively determined; an adjacency matrix representing the business relationship between the business provider and the corresponding business requester is established, and the business values ​​of different business types provided by each interaction object are respectively determined. , establish a corresponding feature matrix, then input the adjacency matrix and the feature matrix into the trained prediction model to obtain the business interaction prediction matrix output by the prediction model, wherein the elements in the business interaction prediction matrix represent the probability of the existence of a business relationship between the various interaction objects in the predicted business interaction network, and then set the elements in the business interaction prediction matrix whose element values ​​are not lower than the set threshold value to the values ​​in the adjacency matrix representing the existence of a business relationship, and set the elements whose element values ​​are lower than the set threshold value to the values ​​in the adjacency matrix representing the absence of a business relationship, generate a business interaction network matrix, and determine the predicted business interaction network based on the business interaction network matrix. In this way, an interaction network is gradually constructed based on each interaction object included in the business data, and the business interaction network is predicted based on the adjacency matrix and feature matrix extracted from the interaction network. The interaction objects are used as the basis for building the interaction network, and then the business interaction network is predicted based on the interaction network. This avoids the disadvantage that when establishing a predicted interaction network, it is necessary to retrieve and determine all interaction objects of the same business type, which makes the established business interaction network incomplete. In addition, with the help of the prediction model, the characteristics of the adjacency matrix and the feature matrix can be integrated to generate a business interaction prediction matrix, and the probability of a business relationship between two interaction objects can be predicted respectively. Then, after processing based on the business interaction prediction matrix, a business interaction network matrix that can generate a corresponding business interaction network can be obtained. With the help of the prediction of the prediction model, interaction objects with potential interactions can be predicted.

[0151] Those skilled in the art will appreciate that the embodiments of the present disclosure may be provided as methods, systems, or computer program products. Therefore, the present disclosure may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present disclosure may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0152] The present disclosure is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present disclosure. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0153] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0154] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0155] Although the preferred embodiments of the present disclosure have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present disclosure.

[0156] Obviously, those skilled in the art can make various changes and modifications to the embodiments of the present disclosure without departing from the spirit and scope of the embodiments of the present disclosure. Thus, if these modifications and variations of the embodiments of the present disclosure fall within the scope of the claims of the present disclosure and their equivalents, the present disclosure is also intended to include these modifications and variations.

Claims

1. A method for predicting a business interaction network, characterized in that: include: Acquire each collected business data, and establish a corresponding interaction network according to each interaction object associated with each business data and the identification information of each interaction object, wherein the interaction network is used to characterize the business relationship and business value between each interaction object; According to the business relationship between each interactive object in the interactive network, a corresponding adjacency matrix is ​​established, and according to the business value of each interactive object for different businesses in the interactive network, a corresponding feature matrix is ​​established, wherein the element representing the existence of the business relationship in the adjacency matrix takes a first set value, and the element representing the absence of the business relationship takes a second set value, and the number of column elements in the feature matrix is ​​positively correlated with the total number of business types; Input the adjacency matrix and the feature matrix into the trained prediction model to obtain a business interaction prediction matrix output by the prediction model, wherein the elements in the business interaction prediction matrix represent the probability of the existence of a business relationship between each interaction object in the predicted business interaction network; a training sample used to train the prediction model includes: an adjacency matrix sample and a feature matrix sample; The elements whose element values ​​in the business interaction prediction matrix are not lower than the set threshold value are set to the first set value, and the elements whose element values ​​are lower than the set threshold value are set to the second set value, to generate a business interaction network matrix, and to determine the predicted business interaction network based on the business interaction network matrix.

2. The method according to claim 1, characterized in that After the corresponding interactive network is established, and before the corresponding adjacency matrix is ​​established according to the business relationship between the interactive objects in the interactive network, the method further includes: Merge the business data of the same business type between any two interaction objects in the interaction network, and for any one of the interaction objects included in the interaction network, obtain the first-category business values ​​associated with the interaction object when it is a service requester, and obtain the second-category business values ​​associated with the interaction object when it is a service provider; and, in the interaction network, retain the first-category business value and the second-category business value with the largest value when the quantity accounts for a set ratio value.

3. The method according to claim 1, characterized in that The step of establishing a corresponding adjacency matrix according to the business relationship between the various interaction objects in the interaction network includes: Obtain each interactive object existing in the interactive network, sort the each interactive object in a random order, and establish an adjacency matrix representing the business relationship between the each interactive object based on the sorted each interactive object, wherein the elements representing two interactive objects having a business relationship in the adjacency matrix have the same value, and any row vector in the adjacency matrix represents the business relationship between an interactive object and each interactive object.

4. The method according to any one of claims 1 to 3, characterized in that In the adjacency matrix and the feature matrix, each row vector having the same relative position corresponds to the same interaction object.

5. The method according to any one of claims 1 to 3, characterized in that: Before inputting the adjacency matrix and the feature matrix into the trained industrial chain prediction model, the method further includes training the industrial chain prediction model, including: Get a training sample set; Read the training samples and perform the following operations respectively until the industrial chain prediction model meets the preset convergence conditions: Inputting the adjacency matrix samples and the feature matrix samples into the constructed industrial chain prediction model to obtain the industrial chain prediction matrix output by the industrial chain prediction model, wherein the industrial chain prediction model is constructed based on the GVAE technology; The preset cross entropy loss function and Gaussian distribution function are used. Under the condition of conforming to the Gaussian distribution, the loss value of the industrial chain prediction model is calculated according to the difference in element values ​​between the industrial chain prediction matrix and the adjacency matrix samples. Based on the loss value, the parameters for generating the industrial chain prediction matrix are adjusted.

6. The method according to claim 5, characterized in that Further including: For each row vector of the industry chain prediction matrix, determine the probability that a business relationship exists between an interaction object represented by the row vector and each interaction object; And each element in the row vector whose element value reaches a set threshold is screened out, and the business relationship determined by the each element is set as a strongly associated business relationship.

7. A prediction device for a business interaction network, characterized in that: include: An acquisition unit, used to acquire each collected business data, and establish a corresponding interaction network according to each interaction object associated with each business data and the identification information of each interaction object, wherein the interaction network is used to characterize the business relationship and business value between each interaction object; An establishing unit is used to establish a corresponding adjacency matrix according to the business relationship between each interactive object in the interactive network, and to establish a corresponding feature matrix according to the business value of each interactive object for different businesses in the interactive network, wherein the element in the adjacency matrix representing the existence of a business relationship takes a first set value, and the element representing the absence of a business relationship takes a second set value, and the number of column elements in the feature matrix is ​​positively correlated with the total number of business types; An input unit is used to input the adjacency matrix and the feature matrix into the trained prediction model to obtain a business interaction prediction matrix output by the prediction model, wherein the elements in the business interaction prediction matrix represent the probability of the existence of a business relationship between each interaction object in the predicted business interaction network; a training sample used to train the prediction model includes: an adjacency matrix sample and a feature matrix sample; The processing unit is used to set the elements in the business interaction prediction matrix whose element values ​​are not lower than the set threshold value to the first set value, and set the elements whose element values ​​are lower than the set threshold value to the second set value, generate a business interaction network matrix, and determine the predicted business interaction network based on the business interaction network matrix.

8. The device according to claim 7, characterized in that After the corresponding interaction network is established, and before the corresponding adjacency matrix is ​​established according to the business relationship between the interaction objects in the interaction network, the establishing unit is further used to: Merge the business data of the same business type between any two interaction objects in the interaction network, and for any one of the interaction objects included in the interaction network, obtain the first-category business values ​​associated with the interaction object when it is a service requester, and obtain the second-category business values ​​associated with the interaction object when it is a service provider; and, in the interaction network, retain the first-category business value and the second-category business value with the largest value when the quantity accounts for a set ratio value.

9. The device according to claim 7, characterized in that When establishing the corresponding adjacency matrix according to the business relationship between the various interactive objects in the interactive network, the establishing unit is used to: Obtain each interactive object existing in the interactive network, sort the each interactive object in a random order, and establish an adjacency matrix representing the business relationship between the each interactive object based on the sorted each interactive object, wherein the elements representing two interactive objects having a business relationship in the adjacency matrix have the same value, and any row vector in the adjacency matrix represents the business relationship between an interactive object and each interactive object.

10. The device according to any one of claims 7 to 9, characterized in that In the adjacency matrix and the feature matrix, each row vector having the same relative position corresponds to the same interaction object.

11. The device according to any one of claims 7 to 9, characterized in that Before inputting the adjacency matrix and the feature matrix into the trained industrial chain prediction model, the step further includes training the industrial chain prediction model, wherein the input unit is used to: Get a training sample set; Read the training samples and perform the following operations respectively until the industrial chain prediction model meets the preset convergence conditions: Inputting the adjacency matrix samples and the feature matrix samples into the constructed industrial chain prediction model to obtain the industrial chain prediction matrix output by the industrial chain prediction model, wherein the industrial chain prediction model is constructed based on the GVAE technology; The preset cross entropy loss function and Gaussian distribution function are used. Under the condition of conforming to the Gaussian distribution, the loss value of the industrial chain prediction model is calculated according to the difference in element values ​​between the industrial chain prediction matrix and the adjacency matrix samples. Based on the loss value, the parameters for generating the industrial chain prediction matrix are adjusted.

12. The device according to claim 11, characterized in that The processing unit is further configured to: For each row vector of the industry chain prediction matrix, determine the probability that a business relationship exists between an interaction object represented by the row vector and each interaction object; And each element in the row vector whose element value reaches a set threshold is screened out, and the business relationship determined by the each element is set as a strongly associated business relationship.

13. An electronic device, characterized in that: include: A memory for storing executable instructions; A processor is used to read and execute executable instructions stored in a memory to implement the method for predicting a business interaction network according to any one of claims 1 to 6.

14. A computer-readable storage medium, characterized in that: When the instructions in the storage medium are executed by an electronic device, the electronic device is enabled to execute the method for predicting a business interaction network as claimed in any one of claims 1 to 6.

Images