A business object classification method, device, equipment and storage medium

Abstract

The present application discloses a classification method, apparatus, device and storage medium for business objects, the method comprising: converting a business object into a graph neural network, the business object having multiple feature data, the nodes in the graph neural network representing multiple feature vectors of the business object; calculating the degree of association between nodes in the graph neural network as the node degree of the node; executing the graph neural network to output a first probability that the business object belongs to a preset category; using a feature vector to identify a second probability that the business object belongs to a preset category; for the same category, merging the first probability and the second probability into a third probability according to the node degree; determining the category to which the business object belongs according to the third probability. This embodiment uses the node degree to organically combine the classification of community node predictions by the graph neural network and the classification of single node predictions based on feature vectors to finally determine the classification, thereby improving the accuracy of the classification.

Description

Technical Field

[0001] The present application relates to the technical field of computer processing, and in particular to a method, device, equipment and storage medium for classifying business objects. Background Art

[0002] In business scenarios such as community mining and anomaly detection, business objects such as users, videos, and audios are classified. Different classification algorithms have different advantages and disadvantages. Therefore, multiple classification algorithms are often used for classification, and the results of multiple classifications are integrated into a final classification result through an integration mechanism.

[0003] At present, the integration mechanism mostly uses the average method, that is, multiple classification algorithms output multiple score vectors for a business object, calculate the average of all score vectors as the final score vector, and take the category corresponding to the largest score vector as the final classification result.

[0004] The average method has a single processing method for each score vector, and uses a unified processing method for different business objects, resulting in low classification accuracy. Summary of the invention

[0005] The present application provides a method, apparatus, device and storage medium for classifying business objects to solve the problem of how to improve the accuracy of classification through an integrated mechanism.

[0006] According to one aspect of the present application, a method for classifying business objects is provided, comprising:

[0007] Converting a business object into a graph neural network, wherein the business object has a plurality of feature vectors, and nodes in the graph neural network represent the plurality of feature vectors of the business object;

[0008] Calculating the degree of association between the nodes in the graph neural network as the node degree of the node;

[0009] Executing the graph neural network to output a first probability that the business object belongs to a preset category;

[0010] Executing a preset classification model to use the feature vector to identify a second probability that the business object belongs to a preset category;

[0011] For the same category, the first probability and the second probability are merged into a third probability according to the node degree;

[0012] The category to which the business object belongs is determined according to the third probability.

[0013] According to another aspect of the present application, a device for classifying business objects is provided, comprising:

[0014] A graph neural network conversion module, used to convert a business object into a graph neural network, wherein the business object has multiple feature vectors, and the nodes in the graph neural network represent the multiple feature vectors of the business object;

[0015] A node degree calculation module, used to calculate the degree of association between the nodes in the graph neural network as the node degree of the node;

[0016] A first classification module, used to execute the graph neural network to output a first probability that the business object belongs to a preset category;

[0017] A second classification module, used for executing a preset classification model to use the feature vector to identify a second probability that the business object belongs to a preset category;

[0018] A probability fusion module, used for fusing the first probability and the second probability into a third probability according to the node degree for the same category;

[0019] A category determination module is used to determine the category to which the business object belongs according to the third probability.

[0020] According to another aspect of the present application, a business object classification device is provided, and the business object classification device includes:

[0021] at least one processor; and

[0022] a memory communicatively connected to the at least one processor; wherein,

[0023] The memory stores a computer program that can be executed by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the business object classification method described in any embodiment of the present application.

[0024] According to another aspect of the present application, a computer-readable storage medium is provided, wherein the computer-readable storage medium stores a computer program, and the computer program is used to enable a processor to implement the business object classification method described in any embodiment of the present application when executed.

[0025] In this embodiment, the business object is converted into a graph neural network, the business object has multiple feature vectors, and the nodes in the graph neural network represent multiple feature vectors of the business object; the degree of association between the nodes is calculated in the graph neural network as the node degree of the node; the graph neural network is executed to output a first probability that the business object belongs to a preset category; the preset classification model is executed to use the feature vector to identify the second probability that the business object belongs to the preset category; for the same category, the first probability and the second probability are merged into a third probability according to the node degree; the category to which the business object belongs is determined according to the third probability. The graph neural network has an advantage in classifying and predicting business objects with a strong degree of association, but has a disadvantage in classifying and predicting business objects with a weak degree of association, while the classification model has an advantage in classifying and predicting business objects with a weak degree of association, but has a disadvantage in classifying and predicting business objects with a strong degree of association. The two have a complementary relationship, and the two can ensure the comprehensiveness of the classification effect. Considering the influence of the degree of association between business objects represented by the node degree on the graph neural network and the classification model, the results of the graph neural network classification and the classification model classification can be fused with the degree of association between business objects represented by the node degree. The node degree is different in different business scenarios, and the results of the fusion of the graph neural network classification and the classification model classification can be flexibly adjusted, which greatly improves the accuracy of classification.

[0026] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present application, nor is it intended to limit the scope of the present application. Other features of the present application will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0028] Figure 1 This is a flow chart of a method for classifying business objects provided according to Embodiment 1 of the present application;

[0029] Figure 2 This is an example diagram of a graph neural network provided according to Embodiment 1 of the present application;

[0030] Figure 3 It is a structural diagram of a device for classifying business objects provided according to Embodiment 2 of the present application;

[0031] Figure 4 It is a structural schematic diagram of a business object classification device that implements the business object classification method of an embodiment of the present application. DETAILED DESCRIPTION

[0032] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0033] It should be noted that the terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0034] Embodiment 1

[0035] Figure 1 This is a flow chart of a business object classification method provided in the first embodiment of the present application. This embodiment is applicable to the case where the results of multiple classifications are integrated into the final classification result through an integration mechanism based on node degree. The method can be executed by a business object classification device, which can be implemented in the form of hardware and / or software, and can be configured in a business object classification device. Figure 1 As shown, the method includes:

[0036] Step 101: Convert the business object into a graph neural network.

[0037] In different business scenarios, there are different business objects, which are collections of data with business domain characteristics.

[0038] For example, for user-oriented service areas, the business object can be users; for the news media field, the business object can be news data; for the mobile communication field, the business object can be mobile communication data; for the e-commerce (EC) field, the business object can be advertising data; for the autonomous driving field, the business object can be point cloud, and so on.

[0039] Although business objects carry business characteristics of different business fields, their essence is still data, such as text information, image data, audio data, video data, and so on.

[0040] For these business objects, you can call the method configured for the business in advance according to the specific business needs in the business scenario to extract features from the business objects, express them in the form of vectors, and form multiple feature vectors. That is, when classifying business objects, the business objects have multiple feature vectors, which are multi-dimensional vectors.

[0041] For example, if the business is to identify whether the user (client) is a web crawler, the user's (client's) behavior of accessing URI (Uniform Resource Identifier) ​​can be extracted as a feature vector, such as the number of visits to various URIs, the time of accessing various URIs, the density of accessing various URIs, and so on.

[0042] In this embodiment, the business object is converted into a graph neural network (GNN), which is a generalized neural network based on a graph structure and a connection model that captures the dependencies of a graph through message passing between nodes of the graph. Unlike a standard neural network, a graph neural network retains a state that can represent information of arbitrary depth from its neighborhood.

[0043] Among them, a graph is a data structure that models a set of objects (nodes) and their relationships (edges).

[0044] Generally speaking, graph neural networks can be divided into five categories: Graph Convolution Networks (GCN), Graph Attention Networks, Graph Autoencoders, Graph Generative Networks, and Graph Spatial-temporal Networks.

[0045] The input of the graph neural network includes the graph G = (V, E) and the feature vector Feat. In the graph G, the set of nodes (Vertex, also known as vertices) V = {v1, v2, ..., v p}, the set of edges E = {e1,e2,…,e q}, any edge e i =(v j ,v k) means e i Connect node v j and node v k , each node v i Corresponding to an n-dimensional feature vector Feat i ={f1,f2,…,f n}.

[0046] In this embodiment, multiple feature vectors of business objects are input into a graph neural network, and multiple nodes and multiple edges are output, wherein the nodes in the graph neural network represent multiple feature vectors of business objects, and the edges in the graph neural network represent the relationship between business objects, that is, there is a certain correlation between business objects with edges.

[0047] Step 102: Calculate the degree of association between nodes in the graph neural network as the node degree of the node.

[0048] In a graph neural network, some nodes have certain associations (also called correlations) with other nodes. These associations are expressed in the form of the number of edges, weights, etc. In this embodiment, for each node in the graph neural network, the degree of association between the node and other nodes can be calculated as the node degree.

[0049] The node degree of a node can be used to distinguish the type of the node, which includes community nodes and isolated nodes. Community nodes, also known as non-isolated nodes, are nodes with a high degree of association with other nodes, as shown by the node degree of the node being greater than or equal to a preset association threshold. Isolated nodes are nodes with a low degree of association with other nodes or even zero, as shown by the node degree of the node being greater than or equal to a preset association threshold, that is, the node degree of a community node is greater than the node degree of an isolated node.

[0050] Taking the number of edges as an example, for each node in the graph neural network, the number of edges connecting the nodes (also called associations) can be counted as the node degree of the node, where an edge connects (also called associations) two nodes.

[0051] In this example, when the number of edges connected to a node is equal to zero, it means that the node has no associated edges with other nodes. At this time, the node is an isolated node.

[0052] When the number of edges connected to a node is greater than or equal to one, it means that the node has any edges associated with other nodes. In this case, the node is a community node.

[0053] Then, the association threshold may be set to one, an isolated node may be a node whose node degree is less than one (the association threshold), and a community node may be a node whose node degree is greater than or equal to one (the association threshold).

[0054] In a statistical method, in a graph neural network, let the set of nodes V = {v1,v2,…,v p}, the edge set E = {e1,e2,…,e q}, for each node v in the graph neural network i , v i ∈V, for node v i Initialize node degree to zero

[0055] Query edge e in graph neural network i =(v j , v k), e i ∈E, indicating e i Connect node v j and node v k , v j 、v k ∈V.

[0056] For each edge in the graph neural network, the node degrees of the two nodes connected by the edge are accumulated by one, that is,

[0057] In such Figure 2 There are 9 nodes in the graph neural network shown. The node degree of node v1 is 5, the node degree of node v2 is 2, the node degree of node v3 is 2, the node degree of node v4 is 0, the node degree of node v5 is 2, the node degree of node v6 is 5, the node degree of node v7 is 4, the node degree of node v8 is 3, and the node degree of node v9 is 1. Nodes v1, v2, v3, v5, v6, v7, v8, and v9 are all community nodes, and node v4 is an isolated node.

[0058] Of course, the above-mentioned method of calculating the node degree is only used as an example. When implementing this embodiment, other methods of calculating the node degree can be set according to actual conditions, for example, the weights of all edges associated with the node degree are summed as the node degree of the node, etc., and this embodiment of the application does not limit this. In addition, in addition to the above-mentioned method of calculating the node degree, those skilled in the art can also adopt other methods of calculating the node degree according to actual needs, and this embodiment does not limit this.

[0059] Step 103: execute the graph neural network to output a first probability that the business object belongs to a preset category.

[0060] In this embodiment, the complete graph neural network can be trained in advance in an end-to-end manner according to specific business requirements in the business scenario.

[0061] When classifying business objects, a graph neural network is executed. The graph neural network uses the underlying graph as a computational graph and learns neural network primitives to generate an embedding vector (Embedding, i.e., feature vector) of a single node by transferring, transforming, and aggregating the features of the nodes on the entire graph. The generated embedding vector can be used as the input of the differentiable prediction layer for node classification.

[0062] In this embodiment, for each node v i , the graph neural network outputs an m-dimensional vector S i =(s1,s2,…,s m ), where s t Indicates that the node v i Belongs to the default category C t The probability (also known as the score) is recorded as the first probability, where t∈m and m is a positive integer.

[0063] Step 104: Execute a preset classification model to use the feature vector to identify a second probability that the business object belongs to a preset category.

[0064] In this embodiment, one or more classification models can be trained in advance according to specific business requirements in the business scenario. The classification model is a model independent of the graph neural network and is a model based on feature vectors for classification. That is, the input is a business object containing only feature vectors, and the output is the probability (also called score) that the business object belongs to a preset category.

[0065] Furthermore, the classification model can apply machine learning algorithms, such as SVM (support vector machines), LightGBM (Light Gradient Boosting Machine), etc. The classification model can apply deep learning algorithms, such as CNN (Convolutional Neural Network), RNN (Recurrent Neural Network), etc.

[0066] Among them, for the deep learning algorithm, the structure of the classification model is not limited to artificially designed neural networks, but can also be a neural network optimized by a model quantization method, a neural network searched for specific business needs in a business scenario by a NAS (Neural Architecture Search) method, and so on. This embodiment does not impose any restrictions on this.

[0067] In this embodiment, for each node v i , the input is the n-dimensional feature vector Feat of the business objecti ={f1,f2,…,f n}, each classification model outputs an m-dimensional vector S i =(s1,s2,…,s m ), where s t Indicates that the node v i Belongs to the default category C t The probability (also known as the score) is recorded as the second probability, where t∈m and m is a positive integer.

[0068] Step 105: For the same category, the first probability and the second probability are merged into a third probability according to the node degree.

[0069] For the same business object, the graph neural network predicts a first probability that the business object belongs to multiple (two or more) categories. At the same time, each classification model predicts a second probability that the business object belongs to multiple (two or more) categories. The categories divided by the graph neural network for the business object are the same as the categories divided by the classification network. Then, for the same category, the same business object has at least two probabilities of belonging to the category (i.e., a first probability and one or more second probabilities).

[0070] Graph neural networks have advantages in classifying and predicting community nodes. The features of business objects are propagated between nodes through the edges between nodes. The more edges associated with nodes, the richer the feature propagation, thereby optimizing the results of community mining. However, for isolated nodes, the classification effect of graph neural networks is often poor because there are no associated edges or the associated edges are sparse.

[0071] The classification model takes the feature vector of a single business data as input and does not consider the correlation between business objects. It has a good classification effect on isolated nodes with no correlation or sparse correlation. However, for business objects with rich correlation, the classification effect is poor due to ignoring the correlation.

[0072] For the same category, considering the impact of the degree of association between business objects represented by node degree on the graph neural network and classification model, the degree of association between business objects represented by node degree can be used as a parameter to adjust the first probability and the second probability. The first probability and the second probability can be linearly or nonlinearly fused into a third probability through the node degree. The third probability is a comprehensive measure of the results of graph neural network classification and the results of classification model classification.

[0073] In one embodiment of the present application, step 105 may include the following steps:

[0074] Step 1051: Calculate the first weight of the graph neural network for classification and the second weight of the classification model for classification according to the node degree.

[0075] Taking into account the influence of node degree representation and the degree of association between business objects on graph neural networks and classification models, we can configure the weights suitable for classification in the graph neural network with reference to the node degree representation and the degree of association between business objects, which are recorded as the first weight. The first weight can reflect the advantages of the graph neural network in classifying community nodes and its disadvantages in classifying isolated nodes. Similarly, we can configure the weights suitable for classification in the classification model with reference to the node degree representation and the degree of association between business objects, which are recorded as the second weight. The second weight can reflect the disadvantages of the classification model in classifying community nodes and its advantages in classifying isolated nodes.

[0076] In a specific implementation, the first mapping function can be configured for the graph neural network in advance according to the specific business requirements in the business scenario, and the second mapping function can be configured for the classification model.

[0077] Then, when configuring the first weight for the category classified by the neural network and configuring the second weight for the category classified by the classification model, the node degree can be substituted into the first mapping function configured for the graph neural network to generate the first weight of the graph neural network for classification, and the node degree can be substituted into the second mapping function configured for the classification model to generate the second weight of the second mapping function for classification.

[0078] The first mapping function and the second mapping function are usually monotonically increasing functions. The first weight is positively correlated with the node degree, that is, the larger the node degree, the larger the first weight, and vice versa, the smaller the node degree, the smaller the first weight. The second weight is positively correlated with the node degree, that is, the larger the node degree, the larger the second weight, and vice versa, the smaller the node degree, the smaller the second weight.

[0079] The first mapping function and the second mapping function are designed in pairs. The rate at which the first weight increases is not consistent with the rate at which the second weight increases. For the same node degree, there is a difference between the first weight and the second weight, which makes the importance between the graph neural network and the classification model different, adapting to different business scenarios.

[0080] If the node degree indicates that the node is a community node, the first weight is greater than or equal to the second weight to reflect the importance of the graph neural network to the community node.

[0081] If the node degree indicates that the node is an isolated node, the first weight is smaller than the second weight to reflect the importance of the classification model to the isolated node.

[0082] In one example, the first mapping function includes:

[0083]

[0084] Among them, H G(x) is the first weight, x is the node degree, δ is the lower limit of the weight, δ∈[0,1), α is a hyperparameter, such as α=1.

[0085] Accordingly, the second mapping function includes:

[0086]

[0087] Among them, H P (x) is the second weight, x is the node degree, δ is the lower limit of the weight, δ∈[0,1), α is a hyperparameter, such as α=1.

[0088] In this example, the weights (first weight, second weight) grow slowly and smoothly with the node degree to better adapt to the specific business needs of certain business scenarios.

[0089] In another example, the first mapping function includes:

[0090]

[0091] Among them, H G (x) is the first weight, x is the node degree, δ is the lower limit of the weight, δ∈[0,1), β is the hyperparameter, such as β=1.

[0092] Accordingly, the second mapping function includes:

[0093]

[0094] Among them, H P (x) is the second weight, x is the node degree, δ is the lower limit of the weight, δ∈[0,1), β and γ are both hyperparameters, such as β=1, γ=2.

[0095] In this example, there is a range for the node degree, and the weights (first weight, second weight) increase rapidly within this range to better adapt to the specific business needs of certain business scenarios.

[0096] In yet another example, the first mapping function includes:

[0097]

[0098] Among them, H G (x) is the first weight, x is the node degree, δ is the lower limit of the weight, δ∈[0,1).

[0099] Accordingly, the second mapping function includes:

[0100]

[0101] Among them, H P(x) is the second weight, x is the node degree, δ is the lower limit of the weight, μ and ε are hyperparameters, such as μ=2, ε=2.

[0102] In this example, there is a range for the node degree, and the weights (first weight, second weight) increase rapidly within this range to better adapt to the specific business needs of certain business scenarios.

[0103] Of course, the first mapping function and the second mapping function are only used as examples. When implementing this embodiment, other first mapping functions and second mapping functions can be set according to actual conditions, and this embodiment of the present application does not limit this. In addition, in addition to the first mapping function and the second mapping function, those skilled in the art can also use other first mapping functions and second mapping functions according to actual needs, and this embodiment does not limit this.

[0104] Step 1052: For the same category, the product of the first weight and the first probability is set as a first weight adjustment value, and the product of the second weight and the second probability is set as a second weight adjustment value.

[0105] Step 1053: Calculate the sum of the first weighted value and the second weighted value as the third probability.

[0106] In this embodiment, the first probability and the second probability are linearly fused into a third probability, that is, for the same category, for the graph neural network, the first weight is multiplied by the first probability, and the product is recorded as the first adjusted weight value; for the classification model, the second weight is multiplied by the second probability, and the product is recorded as the second adjusted weight value; the first adjusted weight value and the second adjusted weight value are added, and the sum is recorded as the third probability.

[0107] Assume that the graph neural network predicts node v i The first probability of belonging to each category is For node v i Configure the first weight

[0108] Classification model prediction node v i The second probability of belonging to each category is For node v i Configure the second weight

[0109] The integration node v i The third probability of belonging to each category is

[0110] Step 106: Determine the category to which the business object belongs according to the third probability.

[0111] In this embodiment, classification rules may be designed in advance according to confidence. If the third probability of a category satisfies the rule, it means that the confidence that the business object belongs to the category is high, and it can be finally determined that the business object belongs to the category.

[0112] Exemplarily, the largest one is selected from the third probabilities of various categories as the target probability, and the category corresponding to the target probability is determined as the category to which the business object belongs.

[0113] Of course, the above method for determining the category is only used as an example. When implementing this embodiment, other methods for determining the category can be set according to actual conditions. For example, the largest one is selected from the third probability greater than the probability threshold as the target probability, and the category corresponding to the target probability is determined as the category to which the business object belongs, etc. The embodiment of the present application does not limit this. In addition, in addition to the above method for determining the category, those skilled in the art can also adopt other methods for determining the category according to actual needs, and this embodiment does not limit this.

[0114] In this embodiment, the business object is converted into a graph neural network, the business object has multiple feature vectors, and the nodes in the graph neural network represent multiple feature vectors of the business object; the degree of association between the nodes is calculated in the graph neural network as the node degree of the node; the graph neural network is executed to output a first probability that the business object belongs to a preset category; the preset classification model is executed to use the feature vector to identify the second probability that the business object belongs to the preset category; for the same category, the first probability and the second probability are merged into a third probability according to the node degree; the category to which the business object belongs is determined according to the third probability. The graph neural network has an advantage in classifying and predicting business objects with a strong degree of association, but has a disadvantage in classifying and predicting business objects with a weak degree of association, while the classification model has an advantage in classifying and predicting business objects with a weak degree of association, but has a disadvantage in classifying and predicting business objects with a strong degree of association. The two have a complementary relationship, and the two can ensure the comprehensiveness of the classification effect. Considering the influence of the degree of association between business objects represented by the node degree on the graph neural network and the classification model, the results of the graph neural network classification and the classification model classification can be fused with the degree of association between business objects represented by the node degree. The node degree is different in different business scenarios, and the results of the fusion of the graph neural network classification and the classification model classification can be flexibly adjusted, which greatly improves the accuracy of classification.

[0115] Embodiment 2

[0116] Figure 3 This is a schematic diagram of a business object classification device provided in Example 2 of the present application. Figure 3 As shown, the device comprises:

[0117] A graph neural network conversion module 301, used to convert a business object into a graph neural network, wherein the business object has multiple feature vectors, and the nodes in the graph neural network represent the multiple feature vectors of the business object;

[0118] A node degree calculation module 302 is used to calculate the degree of association between the nodes in the graph neural network as the node degree of the node;

[0119] A first classification module 303, configured to execute the graph neural network to output a first probability that the business object belongs to a preset category;

[0120] A second classification module 304, configured to execute a preset classification model to use the feature vector to identify a second probability that the business object belongs to a preset category;

[0121] A probability fusion module 305, configured to fuse the first probability and the second probability into a third probability according to the node degree for the same category;

[0122] The category determination module 306 is configured to determine the category to which the business object belongs according to the third probability.

[0123] In one embodiment of the present application, the node degree calculation module 302 includes:

[0124] An edge statistics module is used to count the number of edges connecting each node in the graph neural network as the node degree of the node, wherein one edge connects two nodes.

[0125] In one embodiment of the present application, the edge statistics module includes:

[0126] A node degree initialization module, used for initializing the node degree of each node in the graph neural network to zero;

[0127] An edge query module, used to query edges in the graph neural network;

[0128] A node degree accumulation module is used to accumulate the node degree of the two nodes connected by each edge in the graph neural network by one.

[0129] In one embodiment of the present application, the probability fusion module 305 includes:

[0130] A weight calculation module, used to calculate the first weight of the graph neural network for classification and the second weight of the classification model for classification according to the node degree;

[0131] a weight adjustment value calculation module, configured to set, for the same category, a product of the first weight and the first probability as a first weight adjustment value, and a product of the second weight and the second probability as a second weight adjustment value;

[0132] The weight adjustment value summing module is used to calculate the sum of the first weight adjustment value and the second weight adjustment value as a third probability.

[0133] In one embodiment of the present application, the weight calculation module includes:

[0134] A first mapping module, used for substituting the node degree into a first mapping function configured for the graph neural network to generate a first weight of the graph neural network for classification;

[0135] A second mapping module, used for substituting the node degree into a second mapping function configured for the classification model to generate a second weight of the second mapping function for classification;

[0136] Wherein, the first weight is positively correlated with the node degree, and the second weight is positively correlated with the node degree;

[0137] If the node degree indicates that the node is a community node, the first weight is greater than or equal to the second weight;

[0138] If the node degree indicates that the node is an isolated node, the first weight is less than the second weight.

[0139] In an example of this embodiment, the first mapping function includes:

[0140]

[0141] Among them, H G (x) is the first weight, x is the node degree, δ is the lower limit of the weight, and α is a hyperparameter;

[0142] The second mapping function comprises:

[0143]

[0144] Among them, H P (x) is the second weight, x is the node degree, δ is the lower limit of the weight, and α is a hyperparameter.

[0145] In another example of this embodiment, the first mapping function includes:

[0146]

[0147] Among them, H G(x) is the first weight, x is the node degree, δ is the lower limit of the weight, and β is a hyperparameter;

[0148] The second mapping function comprises:

[0149]

[0150] Among them, H P (x) is the second weight, x is the node degree, δ is the lower limit of the weight, and β and γ are both hyperparameters.

[0151] In yet another example of this embodiment, the first mapping function includes:

[0152]

[0153] Among them, H G (x) is the first weight, x is the node degree, and δ is the lower limit of the weight;

[0154] The second mapping function comprises:

[0155]

[0156] Among them, H P (x) is the second weight, x is the node degree, δ is the lower limit of the weight, and μ and ε are hyperparameters.

[0157] In one embodiment of the present application, the category determination module 306 includes:

[0158] A target probability selection module, used for selecting the largest one from the third probabilities of each category as the target probability;

[0159] The target probability determination module is used to determine that the category corresponding to the target probability is the category to which the business object belongs.

[0160] The business object classification device provided in the embodiment of the present application can execute the business object classification method provided in any embodiment of the present application, and has the corresponding functional modules and beneficial effects for executing the business object classification method.

[0161] Embodiment 3

[0162] Figure 4 A schematic structural diagram of a business object classification device 10 that can be used to implement an embodiment of the present application is shown.

[0163] like Figure 4As shown, the business object classification device 10 includes at least one processor 11, and a memory connected to the at least one processor 11, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., wherein the memory stores a computer program that can be executed by at least one processor, and the processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 to the random access memory (RAM) 13. In the RAM 13, various programs and data required for the operation of the business object classification device 10 can also be stored. The processor 11, the ROM 12, and the RAM 13 are connected to each other through a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0164] A number of components in the business object classification device 10 are connected to the I / O interface 15, including: an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the business object classification device 10 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0165] The processor 11 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any appropriate processor, controller, microcontroller, etc. The processor 11 executes the various methods and processes described above, such as a classification method for business objects.

[0166] In some embodiments, the classification method of business objects may be implemented as a computer program, which is tangibly contained in a computer-readable storage medium, such as a storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on the classification device 10 of business objects via the ROM 12 and / or the communication unit 19. When the computer program is loaded into the RAM 13 and executed by the processor 11, one or more steps of the classification method of business objects described above may be performed. Alternatively, in other embodiments, the processor 11 may be configured to execute the classification method of business objects in any other appropriate manner (e.g., by means of firmware).

Claims

1. A method for classifying business objects, characterized in that: include: Converting a business object into a graph neural network, wherein the business object has a plurality of feature vectors, and nodes in the graph neural network represent the plurality of feature vectors of the business object; Calculating the degree of association between the nodes in the graph neural network as the node degree of the node; Executing the graph neural network to output a first probability that the business object belongs to a preset category; Executing a preset classification model to use the feature vector to identify a second probability that the business object belongs to a preset category; For the same category, the first probability and the second probability are merged into a third probability according to the node degree; The category to which the business object belongs is determined according to the third probability.

2. The method according to claim 1, characterized in that The step of calculating the degree of association between the nodes in the graph neural network as the node degree of the node includes: For each of the nodes in the graph neural network, the number of edges connecting the node is counted as the node degree of the node, wherein one edge connects two nodes.

3. The method according to claim 2, characterized in that The step of counting the number of edges connected to each node in the graph neural network as the node degree of the node includes: For each of the nodes in the graph neural network, initializing the node degree of the node to zero; Querying edges in the graph neural network; For each of the edges in the graph neural network, the node degrees of the two nodes connected by the edge are accumulated by one.

4. The method according to claim 1, characterized in that: For the same category, fusing the first probability and the second probability into a third probability according to the node degree includes: According to the node degrees, respectively calculate a first weight of the graph neural network for classification and a second weight of the classification model for classification; For the same category, the product of the first weight and the first probability is set as a first weight adjustment value, and the product of the second weight and the second probability is set as a second weight adjustment value; A sum of the first weight adjustment value and the second weight adjustment value is calculated as a third probability.

5. The method according to claim 4, characterized in that The calculating, according to the node degrees, respectively the first weight of the graph neural network for classification and the second weight of the classification model for classification includes: Substituting the node degree into a first mapping function configured for the graph neural network to generate a first weight of the graph neural network for classification; Substituting the node degree into a second mapping function configured for the classification model to generate a second weight of the second mapping function for classification; Wherein, the first weight is positively correlated with the node degree, and the second weight is positively correlated with the node degree; If the node degree indicates that the node is a community node, the first weight is greater than or equal to the second weight; If the node degree indicates that the node is an isolated node, the first weight is less than the second weight.

6. The method according to claim 5, characterized in that The first mapping function comprises: Among them, H G (x) is the first weight, x is the node degree, δ is the lower limit of the weight, and α is a hyperparameter; The second mapping function comprises: Among them, H P (x) is the second weight, x is the node degree, δ is the lower limit of the weight, and α is a hyperparameter.

7. The method according to claim 5, characterized in that The first mapping function comprises: Among them, H G (x) is the first weight, x is the node degree, δ is the lower limit of the weight, and β is a hyperparameter; The second mapping function comprises: Among them, H P (x) is the second weight, x is the node degree, δ is the lower limit of the weight, and β and γ are both hyperparameters.

8. The method according to claim 5, characterized in that The first mapping function comprises: Among them, H G (x) is the first weight, x is the node degree, and δ is the lower limit of the weight; The second mapping function comprises: Among them, H P (x) is the second weight, x is the node degree, δ is the lower limit of the weight, and μ and ε are hyperparameters.

9. The method according to any one of claims 1 to 8, characterized in that The determining the category to which the business object belongs according to the third probability includes: Select the largest one from the third probabilities of the various categories as the target probability; The category corresponding to the target probability is determined to be the category to which the business object belongs.

10. A device for classifying business objects, characterized in that: include: A graph neural network conversion module, used to convert a business object into a graph neural network, wherein the business object has multiple feature vectors, and the nodes in the graph neural network represent the multiple feature vectors of the business object; A node degree calculation module, used to calculate the degree of association between the nodes in the graph neural network as the node degree of the node; A first classification module, used to execute the graph neural network to output a first probability that the business object belongs to a preset category; A second classification module, used for executing a preset classification model to use the feature vector to identify a second probability that the business object belongs to a preset category; A probability fusion module, used for fusing the first probability and the second probability into a third probability according to the node degree for the same category; A category determination module is used to determine the category to which the business object belongs according to the third probability.

11. A business object classification device, characterized in that: The business object classification device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, The memory stores a computer program executable by the at least one processor, and the computer program is executed by the at least one processor so that the at least one processor can execute the business object classification method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and the computer program is used to enable a processor to implement the method for classifying business objects according to any one of claims 1 to 9 when the computer program is executed.

Images