Information evaluation method and device

Abstract

The invention provides an information evaluation method and device, and the method comprises the steps: determining a first graph feature based on a pre-constructed target graph network; wherein nodes of the target graph network comprise a plurality of user nodes and device nodes associated with the user nodes, and the first graph feature is used for representing a relationship between a node associated with a first user node and a node associated with a second user node in the target graph network; and based on the first graph feature, determining the intimacy between the first user node and the second user node. Through the information evaluation method provided by the invention, the intimacy evaluation efficiency can be improved, and the accuracy of an intimacy evaluation result is ensured.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to an information evaluation method and device. Background Art

[0002] With the continuous development of mobile network technology, users are in contact with each other more and more frequently. In order to ensure smooth communication, applicants are often required to fill in emergency contacts when handling financial services, handling certificates, signing contracts, etc. However, in related technologies, emergency contacts are often contacted manually by phone to verify the intimacy between the applicant and the emergency contact. This method not only consumes a lot of labor costs, but is also inefficient. Summary of the invention

[0003] The embodiments of the present application provide an information evaluation method and device to improve the efficiency of user intimacy evaluation.

[0004] In order to solve the above technical problems, this application is implemented as follows:

[0005] In a first aspect, an embodiment of the present application provides an information evaluation method. The method comprises:

[0006] Based on a pre-constructed target graph network, a first graph feature is determined; wherein the nodes of the target graph network include multiple user nodes and device nodes associated with the user nodes, and the first graph feature is used to characterize the relationship between the nodes associated with the first user node and the nodes associated with the second user node in the target graph network, the nodes associated with the first user node include at least one of a user node that has a call relationship with the first user node and a device node associated with the first user node, and the nodes associated with the second user node include at least one of a user node that has a call relationship with the second user node and a device node associated with the second user node;

[0007] Based on the first graph feature, determine the intimacy between the first user node and the second user node.

[0008] In a second aspect, an embodiment of the present application further provides an information evaluation device. The information evaluation device comprises:

[0009] A first determination module is used to determine a first graph feature based on a pre-constructed target graph network; wherein the nodes of the target graph network include multiple user nodes and device nodes associated with the user nodes, the first graph feature is used to characterize the relationship between the nodes associated with the first user node and the nodes associated with the second user node in the target graph network, the nodes associated with the first user node include at least one of a user node that has a call relationship with the first user node and a device node associated with the first user node, and the nodes associated with the second user node include at least one of a user node that has a call relationship with the second user node and a device node associated with the second user node;

[0010] The second determination module is used to determine the intimacy between the first user node and the second user node based on the first graph feature.

[0011] In a third aspect, an embodiment of the present application further provides an information evaluation device, comprising a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the above-mentioned information evaluation method when executed by the processor.

[0012] In a fourth aspect, an embodiment of the present application further provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above-mentioned information evaluation method are implemented.

[0013] In a fifth aspect, an embodiment of the present application further provides a computer program product, which is stored in a storage medium and implements the steps of the above-mentioned information evaluation method when the computer program product is executed by at least one processor.

[0014] In an embodiment of the present application, based on a pre-constructed target graph network, a first graph feature is determined, and based on the first graph feature, the intimacy between the first user node and the second user node is determined. This intimacy evaluation method is not only highly efficient, but also can save a lot of labor costs. In addition, since the first graph feature is used to characterize the relationship between the nodes associated with the first user node and the nodes associated with the second user node in the target graph network, the nodes associated with the first user node include at least one of a user node that has a call relationship with the first user node and a device node associated with the first user node, and the nodes associated with the second user node include at least one of a user node that has a call relationship with the second user node and a device node associated with the second user node. That is, in an embodiment of the present application, the intimacy between the first user node and the second user node is determined based on the call situation between the first user node and the second user node and the associated device nodes. Since the call situation between the user nodes can more directly reflect the intimacy between the user nodes, the device nodes associated with the user nodes can reflect whether the user nodes are associated with the same device node, and thus can also reflect the intimacy between the user nodes. Therefore, based on the call situation between the first user node and the second user node and the associated device nodes, the intimacy between the first user node and the second user node can be more accurately evaluated. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments of the present application 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 paying any creative labor.

[0016] Figure 1 is a flow chart of the information evaluation method provided in an embodiment of the present application;

[0017] Figure 2 It is one of the schematic diagrams of the target graph network provided in the embodiment of the present application;

[0018] Figure 3 This is the second schematic diagram of the target graph network provided in the embodiment of the present application;

[0019] Figure 4 is a structural diagram of an information evaluation device provided in an embodiment of the present application;

[0020] Figure 5 It is a structural diagram of an information evaluation device provided in yet another embodiment of the present application. DETAILED DESCRIPTION

[0021] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are 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 are within the scope of protection of this application.

[0022] For ease of understanding, some contents involved in the embodiments of the present application are described below:

[0023] Graph computing: A graph is an abstract data structure used to represent the relationship between objects, which is described using nodes (Vertex) and edges (Edge). Nodes represent objects, and edges represent the relationship between objects. Data that can be abstracted into a graph is called graph data. Graph computing is the process of using graphs as data models to express and solve problems.

[0024] Shortest path algorithm: used to calculate the shortest path between a pair of graph nodes. The commonly used one is Dijkstra's shortest path algorithm. Dijkstra's shortest path algorithm first finds the minimum weight relationship from the starting node to the nodes directly connected to it. The algorithm tracks these weights and moves to the nearest node, then performs the same calculation for the current node, but now the weight is the cumulative value calculated from the starting node. The algorithm continues to execute, calculating a cumulative weight "curve", always choosing the path with the smallest cumulative weight to move forward until the target node is reached.

[0025] Triangle Count: Counts the number of triangles passing through a node.

[0026] Connected component algorithm: Regardless of the direction of the relationship in the graph, if all nodes in a set can reach other nodes in the set, then the set belongs to a connected component.

[0027] Label Propagation Algorithm: A graph clustering algorithm where each node is initialized with a unique label (an identifier) ​​and an initial propagation seed node can be selected; during the algorithm iteration, node labels are propagated through the network; in each propagation iteration, each node updates its label to match the label with the maximum weight, which is calculated based on the weights of neighboring nodes and their relationships, and connections are disconnected uniformly and randomly; when each node has the labels of most of its neighbors, the label propagation algorithm converges and nodes with the same label eventually belong to the same cluster.

[0028] Louvain clustering algorithm: Louvain clustering algorithm is a community discovery algorithm based on modularity. The algorithm performs well in efficiency and effectiveness, and can discover hierarchical community structures. Its optimization goal is to maximize the modularity of the entire community network.

[0029] Logistic Regression (LR) model: It belongs to the generalized linear regression model. The result of linear regression is input into the sigmoid function. The output of the sigmoid function is the LR model result, which is between 0 and 1. If 0.5 is used as the threshold, the LR model can be used as a binary classification model. In other words, the LR model is a model built based on the sigmoid function. The sigmoid function is as follows:

[0030]

[0031] Accordingly, the structure of the LR model can be shown as follows: _pred Represents the prediction results, x1 to x n Characteristic variables representing the input:

[0032]

[0033] Information Value (IV) value: In the binary classification problem of machine learning, the IV value is mainly used to encode the input variables and evaluate the predictive ability. The size of the IV value of the feature variable indicates the strength of the variable's predictive ability. The value range of the IV value is [0, +∞). If the current group only contains responding customers or non-responding customers, IV = positive infinity. Exemplarily, the meaning of the quantitative indicators can be as follows: IV value <0.02 means it is useless for prediction; IV value is between 0.02-0.1, indicating weak predictive ability; IV value is between 0.1-0.3, indicating medium predictive ability; IV value is between 0.3-0.5, indicating strong predictive ability; IV value>0.5, indicating super strong predictive ability.

[0034] At present, applicants are often required to fill in emergency contacts in scenarios such as handling financial business, handling certificates, and signing contracts. However, in related technologies, emergency contacts are often contacted by manual phone calls to verify the intimacy relationship between the applicant and the emergency contact. This method not only requires a lot of labor costs, but also has low efficiency. Based on this, an embodiment of the present application provides an information evaluation method, which constructs a target graph network based on the association relationship between users (for example, call relationship) and the association relationship between users and devices, and evaluates the intimacy between users based on the graph features of the constructed target graph network. The information evaluation method can be executed by electronic devices, such as terminals, servers and other devices.

[0035] See also Figure 1 , Figure 1 is a flow chart of the information evaluation method provided in the embodiment of the present application, such as Figure 1 As shown, the following steps are included:

[0036] Step 101. Determine a first graph feature based on a pre-constructed target graph network; wherein the nodes of the target graph network include multiple user nodes and device nodes associated with the user nodes, and the first graph feature is used to characterize the relationship between nodes associated with the first user node and nodes associated with the second user node in the target graph network, the nodes associated with the first user node include at least one of a user node that has a call relationship with the first user node and a device node associated with the first user node, and the nodes associated with the second user node include at least one of a user node that has a call relationship with the second user node and a device node associated with the second user node.

[0037] In this embodiment, a user node can be used to represent a user or a contact person or a contact number or a user terminal, etc. The above contact number may include but is not limited to a mobile phone number, a social account, etc. A device node can be used to represent a device. The types of device nodes in the above target graph network may include but are not limited to at least one of the types of mobile device nodes, base station nodes, router nodes, near-field communication device nodes, etc. It can be understood that a mobile device node is used to represent a mobile device, such as a mobile phone, a wearable device (Wearable Device), etc. A base station node is used to represent a base station. A router node is used to represent a router. A near-field communication device node is used to represent an incoming communication device.

[0038] The device nodes associated with the above-mentioned user node, for example, the mobile device nodes associated with the above-mentioned user node may include the mobile device nodes used by the user node for calls; the base station nodes associated with the above-mentioned user node may include the base station nodes used by the user node for calls; the router nodes associated with the above-mentioned user node may include the router nodes used by the user node to connect to the network; the near-field communication device nodes associated with the above-mentioned user node may include the near-field communication device nodes that have performed process communications with the user node.

[0039] The edges of the target graph network may represent the association between user nodes and between user nodes and between user nodes and device nodes. In practical applications, a large amount of user data may be obtained. For example, the user data may include a contact number, call data of the contact number, device data associated with the contact number, etc., and then the target graph network may be constructed based on the obtained user data.

[0040] The first user node and the second user node may be any two user nodes in the target graph network.

[0041] The user node having a communication relationship with the first user node may include a user node having a direct communication relationship with the first user node. The user node having a direct communication relationship with the first user node may be understood as a user node having a communication relationship with the first user node or having conducted a communication.

[0042] In some optional embodiments, the user node having a call relationship with the first user node may also include a user node having an indirect call relationship with the first user node. The user node having an indirect call relationship with the first user node may be understood as a user node having no call with the first user node, but a user node having a call with the user node and a user node having a call with the first user node are associated with each other, for example, the user node having a call with the user node and the user node having a call with the first user node have the same user node.

[0043] Among them, the above-mentioned direct call relationship can also be called a first-level call relationship. The above-mentioned indirect call relationship can also be called a P-level call relationship, and P can include any integer greater than 1. Exemplarily, the existence of a first-level call relationship between two user nodes means that there is a call between the two user nodes. The existence of a second-level call relationship between two user nodes means that there is a call between one of the two user nodes and a user node that has a first-level call relationship with another user node; the existence of a third-level call relationship between two user nodes means that there is a call between one of the two user nodes and a user node that has a second-level call relationship with another user node, and so on. The existence of a P-level call relationship between two user nodes means that there is a call between one of the two user nodes and a user node that has a P-1-level call relationship with another user node.

[0044] The user node having a call relationship with the second user node may include a user node having a direct call relationship with the second user node. In some optional embodiments, the user node having a call relationship with the second user node may also include a user node having an indirect call relationship with the second user node. Among them, the user node having a direct call relationship with the second user node and the user node having an indirect call relationship with the second user node can refer to the above-mentioned user node having a direct call relationship with the first user node and the user node having an indirect call relationship with the first user node, respectively, and will not be repeated here.

[0045] The first graph feature is a graph feature associated with the first user node and the second user in the target graph network. For example, the first graph feature may include but is not limited to the shortest path between the first user node and the second user node, the number of common neighboring nodes, the number of triangles passing through the first user node and the second user node, and other graph features, which are used to characterize the relationship between the node associated with the first user node and the node associated with the second user node in the target graph network. Exemplarily, the relationship between the node associated with the first user node and the node associated with the second user node may include at least one of the following:

[0046] The relationship between the device nodes associated with the first user node and the device nodes associated with the second user node, for example, whether the device nodes associated with the first user node and the device nodes associated with the second user node are identical, and the number of identical device nodes between the device nodes associated with the first user node and the device nodes associated with the second user node;

[0047] The relationship between a user node having a call relationship with a first user node and a user node having a call relationship with a second user node, for example, whether there are the same user nodes between a user node having a call relationship with the first user node and a user node having a call relationship with the second user node, and the number of the same user nodes between a user node having a call relationship with the first user node and a user node having a call relationship with the second user node.

[0048] Step 102: Based on the first graph feature, determine the intimacy between the first user node and the second user node.

[0049] Exemplarily, corresponding evaluation values ​​can be calculated based on each graph feature of the first graph feature, and the intimacy between the first user node and the second user node can be determined based on the evaluation values ​​corresponding to each graph feature. For example, the result obtained by weighting the evaluation values ​​corresponding to each graph feature is used as the intimacy between the first user node and the second user node; or, the first graph feature can be input into a pre-trained intimacy evaluation model to determine the intimacy between the first user node and the second user node.

[0050] The information evaluation method of the embodiment of the present application determines the first graph feature based on a pre-constructed target graph network, and determines the intimacy between the first user node and the second user node based on the first graph feature. This intimacy evaluation method is not only highly efficient, but also can save a lot of labor costs. In addition, since the first graph feature is used to characterize the relationship between the nodes associated with the first user node and the nodes associated with the second user node in the target graph network, the nodes associated with the first user node include at least one of the user node that has a call relationship with the first user node and the device node associated with the first user node, and the nodes associated with the second user node include at least one of the user node that has a call relationship with the second user node and the device node associated with the second user node. That is, the embodiment of the present application determines the intimacy between the first user node and the second user node based on the call situation between the first user node and the second user node and the associated device nodes. Since the call situation between the user nodes can more directly reflect the intimacy between the user nodes, the device nodes associated with the user nodes can reflect whether the user nodes are associated with the same device node, and thus can also reflect the intimacy between the user nodes. Therefore, based on the call situation between the first user node and the second user node and the associated device nodes, the intimacy between the first user node and the second user node can be more accurately evaluated.

[0051] Optionally, the edge between two user nodes in the target graph network indicates that there is a call between the two user nodes, and the weight of the edge between the two user nodes is determined according to the number of calls between the two user nodes. The edge between the user node and the device node in the target graph network indicates that there is an association between the user node and the device node, and the weight of the edge between the user node and the device node is determined according to the number of associations between the user node and the device node.

[0052] There is an association between the above-mentioned user node and the device node. For example, if there is an association between the user node and the base station node, it can be understood that the user node has made a call through the base station node; if there is an association between the user node and the mobile device node, it can be understood that the user node has made a call through the mobile device node; if there is an association between the user node and the router node, it can be understood that the user node has connected to the network through the router node; if there is an association between the user node and the near-field communication device node, it can be understood that the user node has conducted near-field communication with the near-field communication device node. It can be understood that, in the case where a user node is used to represent a user or a contact person or a contact number, the interaction between the above-mentioned user node and the device node can be understood as the interaction between the above-mentioned user node and the device node through its associated user device.

[0053] For example, an example of a target graph network can be Figure 2 or Figure 3 As shown, in Figure 2 or Figure 3 In the figure, “John”, “Tom”, “Kobe”, “Lily”, “Kid”, “Bob” and “Jamez” all represent user nodes, “5G-Station: 2345” represents base station node, “WLAN: 234” and “WLAN: 2” all represent router nodes, “QRscan: 234433” represents near field communication device node, and “IMEI: 12” represents mobile device node. The weight of the edge between user nodes (i.e., the number on the edge) represents the number of calls, and the weight of the edge between user nodes and device nodes represents the number of associations between user nodes and device nodes.

[0054] In the target network of this embodiment, the weight of the edge between two user nodes is determined according to the number of calls between the two user nodes, and the weight of the edge between the user node and the device node is determined according to the number of associations between the user node and the device node. Since the number of calls between user nodes and the number of associations between user nodes and device nodes can both reflect the degree of intimacy between the user nodes, for example, the more calls between two user nodes, the more times the two user nodes are associated with the same device node, indicating that the two user nodes are closer. Therefore, in some optional embodiments, the above-mentioned first graph feature can also be used to reflect the number of calls between the first user node and the second user node, the number of times the first user node and the second user node are associated with the same device node, etc., thereby improving the accuracy of the intimacy between the first user node and the second user node determined based on the first graph feature.

[0055] Optionally, the first graph feature includes at least one of the following:

[0056] the shortest path between the first user node and the second user node in the target graph network;

[0057] the number of common neighbor nodes between the first user node and the second user node in the target graph network;

[0058] The number of nodes in the target graph network after deduplication of the union of the neighboring nodes of the first user node and the neighboring nodes of the second user node;

[0059] the number of target triangles in the target graph network, the target triangles being triangles passing through the first user node and the second user node;

[0060] The perimeter of a target triangle in the target graph network, the perimeter of the target triangle being the sum of the weights of three sides of the target triangle;

[0061] In the target graph network, whether the first user node and the second user node are located in the same connected component;

[0062] In the target graph network, whether the first user node and the second user node are located in the same cluster.

[0063] For the above-mentioned shortest path, illustratively, the shortest path between the first user node and the second user node in the target graph network can be calculated by Dijkstra's shortest path algorithm or Floyd's shortest path algorithm. Among them, the calculation of the shortest path in this embodiment needs to consider the weight of the edge in the target graph network. It should be noted that the shortest path can measure the difficulty of two nodes to establish a connection. If the distance is long, it means that more nodes and relationships must be passed through before they can be connected to each other. If the distance is short, it means that fewer nodes or no other nodes can be passed through to directly establish a connection, so it has a great impact on the intimacy of the two nodes.

[0064] The number of common neighbor nodes mentioned above can represent the number of common neighbor nodes that are first associated between two user nodes in the target graph network. For example, see Figure 2 , the number of common neighbor nodes of the two nodes "John" and "James" is 2, namely "5G-Station: 2345" and "WLAN: 234". It should be noted that in the target graph network, the more common neighbor nodes two user nodes have, the more user nodes or device nodes the two user nodes are commonly associated with, and the higher the intimacy between the two user nodes.

[0065] The number of nodes after deduplication in the union of the neighbor nodes of the first user node and the neighbor nodes of the second user node can be used to reflect the social capabilities of the first user node and the second user node. Since the neighbor nodes of a user node can measure the social capabilities of the user node, the more neighbor nodes there are, the stronger the social capabilities are. If the social capabilities of two user nodes are both strong, the two user nodes are more likely to have an intersection. Therefore, the number of nodes after deduplication in the union of the neighbor nodes of the first user node and the neighbor nodes of the second user node also has a certain impact on the intimacy of the two user nodes.

[0066] The target triangle is a triangle passing through the first user node and the second user node, that is, the first user node and the second user node are two vertices of the target triangle.

[0067] It should be noted that triangle counting is used to determine the number of triangles that can be directly formed between each node in the graph and its neighboring nodes. A triangle is a set of three nodes, and each of the three nodes is directly connected to the other two nodes. Triangle counting can reflect the ability to form a loop between any three points in the graph, and is also a basic concept when calculating the global clustering coefficient. By calculating the number of target triangles in the above target graph network, the connectivity and connection density between two user nodes can be measured, which has a great impact on the intimacy assessment. By calculating the perimeter of the above target triangle, the difficulty of the close connection between two user nodes can be measured. The smaller the average perimeter, the easier it is to connect closely.

[0068] For connected components, regardless of direction, if each node in a set can reach other nodes in the set, then the set belongs to a connected component. For example, the connected components of the target graph network can be calculated by algorithms such as breadth-first search or depth-first search. It should be noted that if two nodes are not in a connected graph, it means that the two nodes will not be connected in the graph network and have extremely low intimacy, so this graph feature has a good predictive ability for determining negative examples.

[0069] For clustering, exemplarily, a label propagation algorithm (LPA) can be used to cluster the nodes of the target graph network, and / or a Louvain clustering algorithm can be used to cluster the nodes of the target graph network. Accordingly, whether the first user node and the second user node are located in the same cluster may include whether the first user node and the second user node are located in the same first cluster, and / or whether the first user node and the second user node are located in the same second cluster, wherein the first cluster is a cluster based on the LPA algorithm, and the second cluster is a cluster based on the Louvain clustering algorithm.

[0070] It should be noted that the biggest advantage of the above LPA algorithm is that the algorithm logic is very simple, and the process speed is faster than the modularity optimization algorithm, but the clustering results are unstable and highly random. The clustering results of the above Louvain clustering algorithm are relatively stable, but it requires many rounds of iterative calculations and is relatively time-consuming. It should also be noted that the above clusters can also be called societies or communities.

[0071] This embodiment performs intimacy evaluation between user nodes based on the above graph features, which is conducive to ensuring the accuracy of the intimacy evaluation results.

[0072] In some optional embodiments, the IV value of each of the multiple graph features may be calculated respectively, and according to the IV value of each graph feature, the graph features used for evaluating the intimacy of the user nodes may be screened out from the multiple graph features, and then the intimacy between the user nodes may be determined based on the screened graph features. That is, the first graph feature includes the screened graph feature.

[0073] The above-mentioned multiple graph features may include at least two of the following:

[0074] The shortest path between two user nodes;

[0075] The number of common neighbor nodes between the above two user nodes;

[0076] The number of nodes after deduplication in the union of the neighboring nodes of the above two user nodes;

[0077] The number of triangles passing through the above two user nodes;

[0078] The perimeter of the triangle passing through the two user nodes;

[0079] Whether the above two user nodes are located in the same connected component;

[0080] Whether the above two user nodes are located in the same first sub-cluster;

[0081] Whether the above two user nodes are located in the same second cluster.

[0082] Exemplarily, the IV value calculation process of each graph feature may include the following steps:

[0083] Step a1: bin multiple sample data based on graph features, and calculate the weight of evidence (WOE) of each bin. For discrete graph features, make one bin for each category. For numerical graph features, binning can be performed using equal frequency binning, equal height binning, and other binning methods.

[0084] Among them, for the WOE of the i-th bin, its calculation formula can be as follows:

[0085]

[0086] Among them, Yi represents the number of sample data with intimacy label of 1 in the i-th bin, and Ys represents the total number of sample data with intimacy label of 1; Ni represents the number of sample data with intimacy label of 0 in the i-th bin, and Ns represents the total number of sample data with intimacy label of 0. i is an integer less than or equal to n, and n is the number of bins.

[0087] Step a2: Calculate the IV value of each bin.

[0088] Among them, for the IV value of the i-th bin, its calculation formula can be as follows:

[0089]

[0090] Step a3: Calculate the IV value of the graph feature.

[0091] Among them, the calculation formula for the IV value of the graph feature can be as follows:

[0092]

[0093] Exemplarily, when the IV value of each graph feature is obtained, graph features with an IV value greater than or equal to 0.1 may be selected for intimacy evaluation.

[0094] This embodiment screens graph features for intimacy assessment based on IV values ​​of graph features, which helps to improve the efficiency of intimacy assessment while ensuring the accuracy of intimacy assessment and reduce the computing power required for intimacy assessment.

[0095] Optionally, in the case where the device node includes a base station node, the number of associations between the user node and the base station node is the number of times the user node makes calls through the base station node;

[0096] and / or,

[0097] In the case where the device node includes a router node, the number of associations between the user node and the router node is the number of times the user node connects to the network through the router node;

[0098] and / or,

[0099] In the case where the device node includes a mobile device node, the number of associations between the user node and the mobile device node is the number of times the user node makes calls through the mobile device node;

[0100] and / or,

[0101] In the case where the device node includes a near field communication device node, the number of associations between the user node and the near field communication device node is the number of near field communications performed between the user node and the near field communication device node.

[0102] It should be noted that, in the case where a user node represents a user or a contact person or a contact number, the above-mentioned user node uses a base station node to make a call, which can be understood as the above-mentioned user node accessing the base station node through its associated user equipment to make a call, for example, the user accesses the base station through its associated mobile phone to make a call; the above-mentioned user node connects to the network through a router node, which can be understood as the above-mentioned user node accessing the network provided by the router node through the associated user equipment, for example, the user accesses the wireless network provided by the router through its associated mobile phone; the above-mentioned user node makes a call through a mobile device node, for example, the contact number is associated with the mobile device for a call, that is, the SIM card corresponding to the contact number is inserted into the mobile device for a call; the above-mentioned user node performs near-field communication with a near-field communication device node, which can be understood as the above-mentioned user node performs near-field communication with a near-field communication device node through its associated user equipment, for example, the user performs near-field communication with the near-field communication device through its associated mobile phone.

[0103] It is understandable that the association between user nodes and device nodes can be used to analyze the intimacy between user nodes. Taking the mobile phone number identified by the above user node as an example, if two mobile phone numbers make calls through the same 5G base station, and the signal coverage radius of the 5G base station is about 300 meters, this means that the users of these two mobile phone numbers are located closer; if the two mobile phone numbers are connected to the network through the same wireless router, it means that the users of these two mobile phone numbers are located closer; if the two mobile phone numbers make calls through the same mobile device, it means that the relationship between the users of these two mobile phone numbers is very close.

[0104] In this embodiment, since the base station nodes, router nodes, mobile device nodes and near-field communication device nodes associated with the user nodes can all reflect the close relationship between the user nodes, this embodiment calculates at least one of the number of associations between the user node and the base station node, the number of associations between the user node and the router node, the number of associations between the user node and the mobile device node, and the number of associations between the user node and the near-field communication device node for intimacy assessment between user nodes, which is beneficial to improving the accuracy of intimacy assessment.

[0105] Optionally, the weight of an edge between two user nodes in the target graph network is determined according to the number of calls between the two user nodes and the weights corresponding to the user nodes;

[0106] The weight of the edge between the user node and the device node in the target graph network is determined according to the number of associations between the user node and the device node and the weight corresponding to the device node; wherein different types of device nodes have different corresponding weights.

[0107] Since different types of device nodes have different degrees of influence on the intimacy between user nodes, for example, if two mobile phone numbers make a call through the same mobile device, the impact on intimacy is significantly greater than the impact on intimacy if the two mobile phone numbers connect to the network through the same wireless router. Therefore, this embodiment sets different weights for different types of device nodes, which is conducive to improving the accuracy of intimacy assessment.

[0108] Optionally, before determining the first graph feature based on the pre-constructed target graph network, the method further includes:

[0109] Obtaining target parameters corresponding to each user node pair in a plurality of user node pairs included in the initial graph network, the target parameters including: a total number of calls on a target path, and a total number of associations corresponding to each type of device node on the target path; the target path is the shortest path between two user nodes of the user node pair in the initial graph network;

[0110] Based on the target parameters corresponding to each user node pair and the intimacy labels corresponding to each user node pair, determining the weight corresponding to the user node and the weight corresponding to each type of device node;

[0111] According to the weight corresponding to the user node and the weight corresponding to each type of device node, the weight of the edge of the initial graph network is updated to obtain the target graph network.

[0112] In this embodiment, the weight of the edge between two user nodes in the above-mentioned initial graph network is the number of calls between the two user nodes, and the weight of the edge between the user node and the device node in the initial graph network is the number of associations between the user node and the device node.

[0113] Each of the above-mentioned user node pairs includes two user nodes in the initial graph network. For example, an applicant node and an emergency contact node form a user node pair. Each user node pair has a corresponding intimacy label, which is used to indicate whether the two user nodes of the user node pair are close. For example, the intimacy label is 1, indicating that the two user nodes of the user node pair are close; the intimacy label is 0, indicating that the two user nodes of the user node pair are not close.

[0114] The above-mentioned initial graph network includes multiple user nodes and device nodes associated with the user nodes. The edge between two user nodes in the above-mentioned initial graph network indicates that there is a call between the two user nodes. The weight of the edge between the two user nodes is the number of calls between the two user nodes. The edge between the user node and the device node in the above-mentioned initial graph network indicates that there is an association between the user node and the device node. The weight of the edge between the user node and the device node is the number of associations between the user node and the device node. Exemplarily, the above-mentioned initial graph network can be as follows Figure 3 shown.

[0115] The types of the above-mentioned device nodes may include at least one of a base station node, a mobile device node, a router node, and a near field communication device node.

[0116] For each user node pair, the corresponding target path is the shortest path between the two user nodes of the user node pair in the initial graph network. The above-mentioned total number of calls on the target path can be understood as the sum of all call times on the target path. The above-mentioned total number of associations corresponding to each type of device node on the target path can be understood as the sum of the number of associations between each type of device node on the target path and all user nodes of the user node pair. For example, if the user node pair includes a third user node and a fourth user node, the above-mentioned total number of associations corresponding to each type of device node on the target path can be the sum of the number of associations between each type of device node on the target path and the third user node and the number of associations between each type of device node on the target path and the fourth user node.

[0117] In some optional embodiments, the target parameters corresponding to each user node pair and the intimacy label corresponding to each user node pair can be input into a logistic regression (LR) model to obtain the weights corresponding to the user nodes and the weights corresponding to each type of device nodes. The intimacy label corresponding to the user node pair is used to indicate whether the two user nodes of the user node pair are close.

[0118] The characteristic variables of the above LR model include the total number of calls on the target path and the total number of associations corresponding to each type of device node on the target path. The weight corresponding to the above user node is the coefficient corresponding to the total number of calls in the above LR model, and the weight corresponding to each type of device node is the coefficient corresponding to the total number of associations corresponding to each type of device node in the above LR model. The output of the above LR model is an intimacy label.

[0119] The following uses the types of the above-mentioned device nodes including base station nodes, mobile device nodes, router nodes and near field communication device nodes as examples for explanation:

[0120] Exemplarily, an applicant node and an emergency contact node form a user node pair, and the target parameters of multiple user node pairs can be as shown in Table 1. In Table 1, the 5G-Station column represents the total number of associations corresponding to the base station node on the target path, the WLAN column represents the total number of associations corresponding to the router node on the target path, the QR-scan column represents the total number of associations corresponding to the near field communication device node on the target path, and the IMEI column represents the total number of associations corresponding to the mobile device node on the target path.

[0121] Table 1

[0122]

[0123] Exemplarily, the above LR model may be:

[0124]

[0125] Among them, y label1 represents the intimacy label, w0 represents a constant, x1 represents the total number of calls on the target path, w1 represents the weight corresponding to the user node, x2 represents the total number of associations corresponding to the base station node on the target path, w2 represents the weight corresponding to the base station node, x3 represents the total number of associations corresponding to the router node on the target path, w3 represents the weight corresponding to the router node, x4 represents the total number of associations corresponding to the near field communication device node on the target path, w4 represents the weight corresponding to the near field communication device node, x5 represents the total number of associations corresponding to the mobile device node on the target path, w5 represents the weight corresponding to the mobile device node.

[0126] Specifically, the target parameters and intimacy labels corresponding to multiple user node pairs can be input into the above-mentioned LR model to train the LR model. After the model training is completed, the various coefficients of the LR model can be obtained, that is, the weights corresponding to the above-mentioned user nodes and the weights corresponding to each type of device nodes, for example, the above-mentioned w1 to w5.

[0127] After obtaining the weights corresponding to the above-mentioned user nodes and the weights corresponding to each type of device nodes, the weights of the edges between two user nodes in the initial graph network can be updated according to the weights corresponding to the user nodes. For example, the weights of the edges between the two user nodes can be updated to the product of the number of calls between the two user nodes and the weights corresponding to the user nodes; and the weights of the edges between the user nodes and each type of device nodes in the initial graph network can be updated respectively according to the weights corresponding to each type of device nodes. The weights of the edges between the user nodes and each type of device nodes can be updated to the product of the number of associations between the user nodes and each type of device nodes and the weights corresponding to the device nodes of that type. The initial graph network after updating the edge weights is the above-mentioned target graph network.

[0128] This embodiment determines the weights corresponding to user nodes and the weights corresponding to each type of device nodes based on the total number of calls on the shortest path between two user nodes of a user node pair in the initial graph network, the total number of associations corresponding to each type of device node, and the intimacy label corresponding to each user node pair. This ensures that the weights corresponding to each type of node obtained can more accurately reflect the influence of each type of node on the intimacy assessment, which is conducive to improving the accuracy of the intimacy assessment.

[0129] Optionally, determining the intimacy between the first user node and the second user node based on the first graph feature includes:

[0130] The first graph feature is input into a target intimacy evaluation model to obtain the intimacy between the first user node and the second user node.

[0131] Among them, the target intimacy assessment model is trained based on the second graph features and intimacy labels corresponding to multiple sample data, the second graph features are graph features associated with two user nodes included in the sample data determined based on the target graph network, and the intimacy label is used to indicate whether the two user nodes included in the sample data are intimate.

[0132] It is understandable that the graph feature items included in the second graph feature are the same as the graph feature items included in the first graph feature. For example, if the second graph feature includes the shortest path between two user nodes, the number of common neighboring nodes between the two user nodes, and the number of triangles passing through the two user nodes, then the first graph feature also includes the shortest path between the two user nodes, the number of common neighboring nodes between the two user nodes, and the number of triangles passing through the two user nodes.

[0133] The intimacy label of the positive sample data indicates that the two user nodes included in the positive sample data are intimacy, for example, the intimacy label of the positive sample data may be 1. The intimacy label of the negative sample data indicates that the two user nodes included in the negative sample data are not intimacy, for example, the intimacy label of the negative sample data may be 0.

[0134] The number of the positive sample data and the number of the negative sample data can be reasonably set according to actual needs. In some optional embodiments, the ratio between the number of the positive sample data and the number of the negative sample data is 1:1.

[0135] The above-mentioned initial intimacy evaluation model may include but is not limited to an LR model, a decision tree model, a neural network model, etc. Exemplarily, when the above-mentioned initial intimacy evaluation model is an LR model, the initial intimacy evaluation model may be as follows:

[0136]

[0137] Among them, y represents the intimacy label, z i represents the i-th graph feature in the second graph feature, k i represents the coefficient of the i-th graph feature, n represents the number of graph features, and c is a constant.

[0138] It is understandable that the target intimacy evaluation model is the initial intimacy evaluation model after training. After the training is completed, the above y represents the intimacy.

[0139] This embodiment obtains the intimacy evaluation value between the first user node and the second user node by inputting the first graph feature into the target intimacy evaluation model. Since the target intimacy evaluation model has a stronger feature learning ability, this is conducive to further improving the accuracy of intimacy evaluation.

[0140] Optionally, after inputting the first graph feature into a target intimacy evaluation model to obtain the intimacy between the first user node and the second user node, the method further includes:

[0141] Determining an intimacy level between the first user node and the second user node according to the intimacy between the first user node and the second user node, a first threshold and a second threshold;

[0142] Among them, the first threshold is the maximum value of the intimacy corresponding to the negative sample data among multiple sample data, the second threshold is the minimum value of the intimacy corresponding to the positive sample data among the multiple sample data, and the multiple sample data are sample data used for training the target intimacy evaluation model.

[0143] The intimacy corresponding to the negative sample data is: the intimacy between the two user nodes included in the negative sample data is obtained by inputting the second graph feature corresponding to the negative sample data into the target intimacy assessment model; the intimacy corresponding to the positive sample data is: the intimacy between the two user nodes included in the positive sample data is obtained by inputting the second graph feature corresponding to the positive sample data into the target intimacy assessment model. In practical applications, after training the target intimacy assessment model based on the above multiple sample data, the second graph feature corresponding to each sample data can be input into the intimacy assessment module to obtain the intimacy between the two user nodes included in each sample data.

[0144] Exemplarily, the maximum value in the intimacy of the negative sample data can be determined as a low intimacy threshold, i.e., the first threshold, and the minimum value in the intimacy of the positive sample data can be determined as a high intimacy threshold, i.e., the second threshold, and multiple intimacy levels can be divided according to the first threshold and the second threshold. For example, when the intimacy is less than or equal to the first threshold, it is determined to be a low intimacy level; when the intimacy is greater than the first threshold and less than or equal to the second threshold, it is determined to be a medium intimacy level; when the intimacy is greater than the second threshold, it is determined to be a high intimacy level, which facilitates more intuitive knowledge of the intimacy relationship between two user nodes.

[0145] It should be noted that this embodiment determines the maximum value of the intimacy of the negative sample data as the low intimacy threshold, i.e., the first threshold, and determines the minimum value of the intimacy of the positive sample data as the high intimacy threshold, i.e., the second threshold. This helps to reduce the inaccuracy of intimacy level division caused by the intimacy assessment error of the target intimacy assessment model.

[0146] In some optional embodiments, the method further includes:

[0147] At least one of the intimacy between the first user node and the second user node and the intimacy level between the first user node and the second user node is pushed to the risk control system.

[0148] Specifically, when the risk control system receives at least one of the intimacy between the first user node and the second user node and the intimacy level between the first user node and the second user node, it can perform risk control based on at least one of the intimacy between the first user node and the second user node and the intimacy level between the first user node and the second user node. For example, for a business application that requires approval, if the intimacy between the first user node and the second user node indicates that the intimacy between the first user node and the second user node is low or the intimacy level is low, the application may be rejected, that is, the user's business application may be rejected.

[0149] See also Figure 4 , Figure 4 is a structural diagram of the information evaluation device provided in the embodiment of the present application. Figure 4 As shown, the information evaluation device 400 includes:

[0150] A first determination module 401 is used to determine a first graph feature based on a pre-constructed target graph network; wherein the nodes of the target graph network include multiple user nodes and device nodes associated with the user nodes, the first graph feature is used to characterize the relationship between the nodes associated with the first user node and the nodes associated with the second user node in the target graph network, the nodes associated with the first user node include at least one of a user node that has a call relationship with the first user node and a device node associated with the first user node, and the nodes associated with the second user node include at least one of a user node that has a call relationship with the second user node and a device node associated with the second user node;

[0151] The second determination module 402 is used to determine the intimacy between the first user node and the second user node based on the first graph feature.

[0152] Optionally, the edge between two user nodes in the target graph network indicates that there is a call between the two user nodes, and the weight of the edge between the two user nodes is determined according to the number of calls between the two user nodes. The edge between the user node and the device node in the target graph network indicates that there is an association between the user node and the device node, and the weight of the edge between the user node and the device node is determined according to the number of associations between the user node and the device node.

[0153] Optionally, the first graph feature includes at least one of the following:

[0154] the shortest path between the first user node and the second user node in the target graph network;

[0155] the number of common neighbor nodes between the first user node and the second user node in the target graph network;

[0156] The number of nodes in the target graph network after deduplication of the union of the neighboring nodes of the first user node and the neighboring nodes of the second user node;

[0157] the number of target triangles in the target graph network, the target triangles being triangles passing through the first user node and the second user node;

[0158] The perimeter of a target triangle in the target graph network, the perimeter of the target triangle being the sum of the weights of three sides of the target triangle;

[0159] In the target graph network, whether the first user node and the second user node are located in the same connected component;

[0160] In the target graph network, whether the first user node and the second user node are located in the same cluster.

[0161] Optionally, in the case where the device node includes a base station node, the number of associations between the user node and the base station node is the number of times the user node makes calls through the base station node;

[0162] and / or,

[0163] In the case where the device node includes a router node, the number of associations between the user node and the router node is the number of times the user node connects to the network through the router node;

[0164] and / or,

[0165] In the case where the device node includes a mobile device node, the number of associations between the user node and the mobile device node is the number of times the user node makes calls through the mobile device node;

[0166] and / or,

[0167] In the case where the device node includes a near field communication device node, the number of associations between the user node and the near field communication device node is the number of near field communications performed between the user node and the near field communication device node.

[0168] Optionally, the weight of an edge between two user nodes in the target graph network is determined according to the number of calls between the two user nodes and the weights corresponding to the user nodes;

[0169] The weight of the edge between the user node and the device node in the target graph network is determined according to the number of associations between the user node and the device node and the weight corresponding to the device node; wherein different types of device nodes have different corresponding weights.

[0170] Optionally, the device further comprises:

[0171] A first acquisition module is used to acquire a target parameter corresponding to each user node pair in a plurality of user node pairs included in the initial graph network, wherein the target parameter includes: a total number of calls on a target path, and a total number of associations corresponding to each type of device node on the target path; the target path is the shortest path between two user nodes of the user node pair in the initial graph network;

[0172] A third determination module, configured to determine a weight corresponding to the user node and a weight corresponding to each type of device node based on a target parameter corresponding to each user node pair and an intimacy label corresponding to each user node pair;

[0173] An updating module is used to update the weights of the edges of the initial graph network according to the weights corresponding to the user nodes and the weights corresponding to each type of device nodes to obtain the target graph network.

[0174] Optionally, the second determining module is specifically configured to:

[0175] The first graph feature is input into a target intimacy evaluation model to obtain the intimacy between the first user node and the second user node.

[0176] Optionally, the device further comprises:

[0177] a fourth determination module, configured to input the first graph feature into a target intimacy evaluation model to obtain the intimacy between the first user node and the second user node, and then determine the intimacy level between the first user node and the second user node according to the intimacy between the first user node and the second user node, a first threshold and a second threshold;

[0178] Among them, the first threshold is the maximum value of the intimacy corresponding to the negative sample data among multiple sample data, the second threshold is the minimum value of the intimacy corresponding to the positive sample data among the multiple sample data, and the multiple sample data are sample data used for training the target intimacy evaluation model.

[0179] The information evaluation device 400 provided in the embodiment of the present application can implement each process in the above method embodiment, and will not be described again here to avoid repetition.

[0180] See also Figure 5 , Figure 5 is a structural diagram of an information evaluation device provided by another embodiment of the present application, such as Figure 5 As shown, the information evaluation device 500 includes: a processor 501, a memory 502, and a computer program stored in the memory 502 and executable on the processor. The various components in the information evaluation device 500 are coupled together via a bus interface 503. When the computer program is executed by the processor 501, the following steps are implemented:

[0181] Based on a pre-constructed target graph network, a first graph feature is determined; wherein the nodes of the target graph network include multiple user nodes and device nodes associated with the user nodes, and the first graph feature is used to characterize the relationship between the nodes associated with the first user node and the nodes associated with the second user node in the target graph network, the nodes associated with the first user node include at least one of a user node that has a call relationship with the first user node and a device node associated with the first user node, and the nodes associated with the second user node include at least one of a user node that has a call relationship with the second user node and a device node associated with the second user node;

[0182] Based on the first graph feature, determine the intimacy between the first user node and the second user node.

[0183] It should be understood that in the embodiment of the present application, when the computer program is executed by the processor 501, it can implement each process in the above-mentioned information evaluation method embodiment, and in order to avoid repetition, it will not be repeated here.

[0184] An embodiment of the present application also provides an information evaluation device, including a processor, a memory, and a computer program stored in the memory and executable on the processor. When the computer program is executed by the processor, the various processes of the above-mentioned information evaluation method embodiment are implemented, and the same technical effect can be achieved. To avoid repetition, it will not be described here.

[0185] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, each process of the above-mentioned information evaluation method embodiment is implemented, and the same technical effect can be achieved. To avoid repetition, it is not repeated here. The computer-readable storage medium is, for example, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0186] An embodiment of the present application also provides a computer program product, which is stored in a storage medium. When the computer program product is executed by at least one processor, it implements the various processes of the above-mentioned information evaluation method embodiment and can achieve the same technical effect. To avoid repetition, it will not be repeated here.

[0187] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0188] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a magnetic disk, or an optical disk), and includes a number of instructions for a terminal (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods described in each embodiment of the present application.

[0189] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. An information evaluation method, characterized in that: include: Based on a pre-constructed target graph network, a first graph feature is determined; wherein the nodes of the target graph network include multiple user nodes and device nodes associated with the user nodes, and the first graph feature is used to characterize the relationship between the nodes associated with the first user node and the nodes associated with the second user node in the target graph network, the nodes associated with the first user node include at least one of a user node that has a call relationship with the first user node and a device node associated with the first user node, and the nodes associated with the second user node include at least one of a user node that has a call relationship with the second user node and a device node associated with the second user node; Based on the first graph feature, determine the intimacy between the first user node and the second user node.

2. The method according to claim 1, characterized in that The edge between two user nodes in the target graph network indicates that there is a call between the two user nodes, and the weight of the edge between the two user nodes is determined according to the number of calls between the two user nodes. The edge between the user node and the device node in the target graph network indicates that there is an association between the user node and the device node, and the weight of the edge between the user node and the device node is determined according to the number of associations between the user node and the device node.

3. The method according to claim 2, characterized in that The first graph feature includes at least one of the following: the shortest path between the first user node and the second user node in the target graph network; the number of common neighbor nodes between the first user node and the second user node in the target graph network; The number of nodes in the target graph network after deduplication of the union of the neighboring nodes of the first user node and the neighboring nodes of the second user node; the number of target triangles in the target graph network, the target triangles being triangles passing through the first user node and the second user node; The perimeter of a target triangle in the target graph network, the perimeter of the target triangle being the sum of the weights of three sides of the target triangle; In the target graph network, whether the first user node and the second user node are located in the same connected component; In the target graph network, whether the first user node and the second user node are located in the same cluster.

4. The method according to claim 2, characterized in that: In the case where the device node includes a base station node, the number of associations between the user node and the base station node is the number of times the user node makes calls through the base station node; and / or, In the case where the device node includes a router node, the number of associations between the user node and the router node is the number of times the user node connects to the network through the router node; and / or, In the case where the device node includes a mobile device node, the number of associations between the user node and the mobile device node is the number of times the user node makes calls through the mobile device node; and / or, In the case where the device node includes a near field communication device node, the number of associations between the user node and the near field communication device node is the number of near field communications performed between the user node and the near field communication device node.

5. The method according to claim 2, characterized in that: The weight of the edge between two user nodes in the target graph network is determined according to the number of calls between the two user nodes and the weights corresponding to the user nodes; The weight of the edge between the user node and the device node in the target graph network is determined according to the number of associations between the user node and the device node and the weight corresponding to the device node; wherein different types of device nodes have different corresponding weights.

6. The method according to claim 5, characterized in that The method further comprises: Obtaining target parameters corresponding to each user node pair in a plurality of user node pairs included in the initial graph network, the target parameters including: a total number of calls on a target path, and a total number of associations corresponding to each type of device node on the target path; the target path is the shortest path between two user nodes of the user node pair in the initial graph network; Based on the target parameters corresponding to each user node pair and the intimacy labels corresponding to each user node pair, determining the weight corresponding to the user node and the weight corresponding to each type of device node; According to the weight corresponding to the user node and the weight corresponding to each type of device node, the weight of the edge of the initial graph network is updated to obtain the target graph network.

7. The method according to any one of claims 1 to 6, characterized in that The determining, based on the first graph feature, the intimacy between the first user node and the second user node includes: The first graph feature is input into a target intimacy evaluation model to obtain the intimacy between the first user node and the second user node.

8. The method according to claim 7, characterized in that After inputting the first graph feature into a target intimacy evaluation model to obtain the intimacy between the first user node and the second user node, the method further includes: Determining an intimacy level between the first user node and the second user node according to the intimacy between the first user node and the second user node, a first threshold and a second threshold; Among them, the first threshold is the maximum value of the intimacy corresponding to the negative sample data among multiple sample data, the second threshold is the minimum value of the intimacy corresponding to the positive sample data among the multiple sample data, and the multiple sample data are sample data used for training the target intimacy evaluation model.

9. An information evaluation device, characterized in that: include: A first determination module is used to determine a first graph feature based on a pre-constructed target graph network; wherein the nodes of the target graph network include multiple user nodes and device nodes associated with the user nodes, the first graph feature is used to characterize the relationship between the nodes associated with the first user node and the nodes associated with the second user node in the target graph network, the nodes associated with the first user node include at least one of a user node that has a call relationship with the first user node and a device node associated with the first user node, and the nodes associated with the second user node include at least one of a user node that has a call relationship with the second user node and a device node associated with the second user node; The second determination module is used to determine the intimacy between the first user node and the second user node based on the first graph feature.

10. An information evaluation device, characterized in that The method comprises a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program implements the steps of the information evaluation method according to any one of claims 1 to 8 when executed by the processor.

11. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps of the information evaluation method according to any one of claims 1 to 8 are implemented.

12. A computer program product, characterized in that The computer program product is stored in a storage medium, and when the computer program product is executed by at least one processor, the steps of the information evaluation method according to any one of claims 1 to 8 are implemented.

Images