WO2021027317A1 - 基于关系网络的属性信息处理方法、装置、计算机设备和存储介质 - Google Patents
基于关系网络的属性信息处理方法、装置、计算机设备和存储介质 Download PDFInfo
- Publication number
- WO2021027317A1 WO2021027317A1 PCT/CN2020/086397 CN2020086397W WO2021027317A1 WO 2021027317 A1 WO2021027317 A1 WO 2021027317A1 CN 2020086397 W CN2020086397 W CN 2020086397W WO 2021027317 A1 WO2021027317 A1 WO 2021027317A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- node
- nodes
- attribute information
- attribute
- target
- Prior art date
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
Definitions
- This application relates to the field of computer technology, and in particular to a method, device, computer equipment, and storage medium for processing attribute information based on a relational network.
- Risk prediction refers to the process of using multiple information channels and analysis methods to determine identification indicators based on the enterprise's risk strategy and risk preference, and using these identification indicators as the starting point to identify potential risks in a timely manner.
- the inventor realizes that the traditional risk prediction method is only a simple index comparison. When the index does not meet the preset conditions, it is generally determined that there is a risk, which reduces the accuracy of risk prediction.
- a method for processing attribute information based on a relational network comprising: obtaining a relational network graph; the relational network graph including a plurality of object nodes and attribute nodes connected to each object node; identifying whether each object node corresponds to attribute information Missing; according to the recognition result, the multiple object nodes in the relationship network graph are divided into determined object nodes and pending object nodes; in the relationship network graph, a fully connected sub-network graph of each of the pending object nodes is drawn; In the fully connected sub-network graph, the determined target node that has a target association relationship with the pending target node is marked as a reference target node; calculates the comprehensive correlation degree between each reference target node and the pending target node; obtains the corresponding reference target node And determine the reference weight of the attribute information value according to the comprehensive association degree; determine the attribute information value corresponding to the target object according to multiple attribute information values and respective reference weights, so as to control the terminal according to The attribute information value is subjected to data processing.
- An attribute information processing device based on a relationship network.
- the device includes: an attribute information processing module for obtaining a relationship network graph; the relationship network graph includes a plurality of object nodes and an attribute node connected to each object node; Whether the corresponding attribute information of each object node is missing; according to the recognition result, the multiple object nodes in the relationship network graph are divided into determined object nodes and undetermined object nodes; in the relationship network graph, the information of each undetermined object node is drawn A fully connected sub-network graph; marking the determined object node in the fully connected sub-network graph that has a target association relationship with the undetermined object node as a reference object node; calculating the comprehensive degree of association between each reference object node and the undetermined object node; Obtain the attribute information value corresponding to the reference object node, and determine the reference weight of the attribute information value according to the comprehensive association degree; the attribute information value calculation module is used to calculate the attribute information value according to multiple attribute information values and corresponding reference weights, The attribute information value corresponding to the
- a computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the method for processing attribute information based on a relational network provided in any embodiment of the present application when the computer program is executed.
- a computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method for processing attribute information based on a relational network provided in any embodiment of the present application are realized.
- the above-mentioned attribute information processing methods, devices, computer equipment and storage media based on the relationship network collectively display the association relationships covering a large number of people in a relationship network graph, which is convenient for users to grasp the association relationships between customers from a global perspective.
- draw the fully connected sub-network graph corresponding to the node of the pending object from the relational network graph so that users can understand one of the customers more specifically;
- you can Combined with the consideration of the identity attribute information of the reference object node that has a strong correlation with the target object node you can Combined with the consideration of the identity attribute information of the reference object node that has a strong correlation with the target object node, the numerical attributes of the target node are supplemented, and the attribute information of the target object can be predicted by integrating multiple dimensional factors, which can not only improve the information supplement Full efficiency can also improve the accuracy of the completion information.
- Figure 1 is an application scenario diagram of an attribute information processing method based on a relational network in an embodiment
- FIG. 2 is a schematic flowchart of an attribute information processing method based on a relational network in an embodiment
- Fig. 3 is a schematic diagram of a relationship network graph used in an attribute information processing process in an embodiment
- FIG. 4 is a schematic flowchart of the steps of risk tracking based on attribute information values in an embodiment
- FIG. 5 is a structural block diagram of an attribute information processing device based on a relational network in an embodiment
- Fig. 6 is an internal structure diagram of a computer device in an embodiment.
- the method for processing attribute information based on the relational network provided in this application can be applied to the application environment as shown in FIG. 1.
- the terminal 102 and the server 104 communicate through the network.
- the terminal 102 may be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices.
- the server 104 may be implemented as an independent server or a server cluster composed of multiple servers.
- the server 104 obtains the identity attribute information of the corresponding target object from the data processing system according to the attribute information processing request triggered by the user based on the terminal 102, and according to the object identifier carried in the attribute information processing request.
- the server 104 identifies whether the attribute information in the identity attribute information is missing.
- the server 104 queries the information prediction model and the correlation coefficient weight corresponding to the data processing system, and inputs the identity attribute information into the information prediction model to obtain the target correlation coefficient.
- the server 104 calculates the first attribute reference value according to the target correlation coefficient and the correlation coefficient weight.
- the server 104 obtains the fully connected subnet graph of the target object.
- the fully connected sub-network graph includes a pending object node corresponding to the target object and at least one reference object node associated with the pending object node.
- the server 104 obtains the attribute information value corresponding to the reference object node.
- the server 104 calculates the comprehensive correlation degree between each reference object node and the undetermined object node, and determines the reference weight of the attribute information value according to the comprehensive correlation degree.
- the server 104 calculates the second attribute reference value according to multiple attribute information values and corresponding reference weights.
- the server 104 calculates the attribute information value corresponding to the target object according to the first attribute reference value and the second attribute reference value.
- the above-mentioned attribute information processing process in addition to the identity attribute information of the target object itself, also considers the identity attribute information of the reference object node that has a strong correlation with the target object node, and integrates multiple dimensions of the attribute information of the target object Making predictions can not only improve the efficiency of information completion, but also improve the accuracy of the information.
- a method for processing attribute information based on a relational network is provided. Taking the method applied to the server in FIG. 1 as an example, the method includes the following steps:
- Step 202 Obtain a relational network graph; the relational network graph includes multiple object nodes and attribute nodes connected to each object node.
- the target audience can be lost customers, existing customers, or potential customers. Among them, potential customers can be identified based on customer data of lost customers or existing customers. For example, guarantors or emergency contacts reserved by existing customers in different data processing systems can be used as potential customers.
- the identity attribute information includes object identification.
- the object identifier can be an ID number, mobile phone number, or email address. It is easy to understand that if the target object is an enterprise, the object identifier can also be an organization code, etc.
- Identity attribute information also includes name, gender, age, education, contact information, employer, insurance policy, bank card account, terminal device information, social network account, interest, wealth level, or risk level, etc.
- the identity attribute information of each target object can be obtained from different data processing systems, but the identity attribute information of some target objects may be incomplete and have missing attributes.
- the potential customer may also be obtained by the server monitoring the product review records left by the user browsing related products on the target website, which is not limited.
- Step 204 Identify whether the corresponding attribute information of each object node is missing.
- the server identifies whether the risk attribute in the identity attribute information is missing.
- Identity attribute information includes a variety of target attributes, such as basic attributes, interest attributes, asset attributes, risk attributes, etc.
- the server identifies whether the target attribute in the identity attribute information is missing, and supplements the missing target attribute based on a preset relationship network graph.
- Step 206 According to the recognition result, the multiple object nodes in the relationship network graph are classified into determined object nodes and pending object nodes.
- Step 208 draw a fully connected sub-network graph of each pending object node in the relational network graph.
- the fully connected subnet graph includes pending object nodes, multiple other object nodes, and directed edges between nodes.
- the fully connected subnet graph can be obtained from the relational network graph.
- the relationship network graph can be pre-built by the server based on the identity attribute information of multiple target objects and social network information.
- the relational network graph includes multiple object nodes, an attribute node corresponding to each object node, and a directed edge for connecting the object node and the attribute node.
- the server recognizes whether different object nodes are connected with the same attribute node. If so, the server merges the same attribute node, and marks the merged attribute node as an associated node corresponding to multiple object nodes.
- Step 210 Mark the determined target node that has a target association relationship with the pending target node in the fully connected sub-network graph as a reference target node.
- the server draws a fully connected sub-network graph of each pending object node in the relationship network graph. Specifically, the server obtains the social network information of the undetermined object node, and calculates the comprehensive association degree between the undetermined object node and each object node once associated based on the social network information in the foregoing manner.
- One degree of association refers to the direct connection with two object nodes through a directed edge.
- the server compares whether the comprehensive relevance degree reaches the threshold, and retains the object nodes whose comprehensive relevance degree reaches the threshold (denoted as the once-relevant object node), and deletes the object nodes whose comprehensive relevance degree is less than the threshold.
- the server identifies whether at least one of the retained once-associated object nodes is a certain object node. In other words, the server determines whether at least one of the retained once-associated object nodes is an object node that contains complete identity attribute information. If there is a certain object node in the reserved once-associated object node, the server draws the pending object node, the reserved once-associated object node, and the directed edge connecting the pending object node and the once-associated object node in the relationship network graph to obtain the pending object The fully connected sub-network graph corresponding to the node.
- the server further filters one or more object nodes that are second-associated with the pending object nodes (denoted as second-degree associated object nodes) in the above manner.
- Two-degree association refers to the connection with two object nodes through two directed edges. It is easy to understand that a second degree associated object node is an object node that is directly connected to a first degree associated object node.
- the server further identifies whether at least one of the second-degree associated object nodes is a certain object node. If not, the third-degree associated object node of the pending object node is further screened according to the above method, and this is repeated until at least one certain object node is obtained by screening.
- the server draws the undecided object nodes, the filtered once-associated object nodes, second-degree associated object nodes, etc., in the relational network graph in the above manner, and obtains the fully connected sub-network graph corresponding to the undecided object nodes.
- the server supplements the missing attributes of the corresponding pending object node according to the identity attribute information of the reference object node in the fully connected sub-network graph.
- the server can supplement the missing attributes of multiple attribute types. Attribute types include basic attributes, interest attributes, asset attributes, risk attributes, etc. According to the difference of missing attributes, the fully connected sub-network graphs of the same pending object node can be different.
- the server presets the screening threshold of the associated object node, if the screening threshold is reached If there is still no certain object node in the associated object nodes of the level, then stop screening the associated object nodes, and generate the corresponding missing attribute supplementation failure prompt message.
- the association threshold is 2, and if there is still no definite object node in the second-degree associated object node, the prompt message of "the missing attribute supplement failed" is returned.
- the server presets a variety of attribute types and corresponding target association relationships. According to the attribute types of the missing attributes, the associated object nodes that have a target association relationship with the pending object nodes are selected as reference object nodes in the corresponding fully connected sub-network graph.
- the attribute type is a basic attribute
- the related objects that have a classmate relationship, a colleague relationship, and a friend relationship with the pending object can be determined as the reference object
- the attribute type is an interest attribute
- the associated object of the person relationship is determined as the reference object
- the attribute type is an asset attribute
- the associated object that has a kinship or friend relationship with the target object is determined as the reference object
- the attribute type is a risk attribute
- the associated object node whose node has other object nodes connected in common is determined as the reference object node. It is easy to understand that all associated objects in the fully connected sub-network graph can also be determined as reference objects, but different reference weights are preset for different reference objects according to different missing attributes, which is not limited.
- Figure 3 exemplarily shows a relationship network map.
- V1 to V8 are 8 object nodes
- Mij and Mijij are the attribute node "holding unit” corresponding to each object node
- Nij and Nijij are the attribute node "transfer ID” corresponding to each object node
- Oij and Oijij are the attribute node "insurance policy” corresponding to each object node
- Pij and Pijij are the attribute node "wireless network identity” corresponding to each object node
- Qij and Qijij are the attribute node "bank” corresponding to each object node card number”.
- 1 ⁇ i ⁇ 8; 1 ⁇ j. Directed edges can point from object nodes to attribute nodes.
- Mijij, Nijij, Oijij, P ijij, and Qijij are associated nodes of multiple target nodes.
- the node ID of the associated node can be generated based on the node IDs of multiple attribute nodes to be merged.
- the node identifier of the associated node obtained by merging the attribute node Q51 and the attribute node Q83 may be Q5183.
- the server also generates a node label corresponding to each object node according to the identity attribute information, such as a basic information label, a consumer interest label, a wealth level label, or a risk rating label.
- the server identifies whether each object node lacks a certain node label to determine whether the identity attribute information of the corresponding object node has a missing attribute.
- the server marks the object node without missing attributes as a definite object node, and marks the object node table with missing attributes as a pending object node.
- the server obtains the identity attribute information of the pending object and identifies the missing target attribute.
- the server determines the attribute type corresponding to the missing attribute of the pending object node, identifies the target association relationship corresponding to the attribute type, and marks one or more object nodes in the fully connected sub-network graph as reference object nodes.
- Step 212 Calculate the comprehensive correlation degree between each reference object node and the pending object node.
- the server calculates the comprehensive correlation degree between each reference object node and the pending object node. Two object nodes connected with one or more associated nodes are related. The server calculates the unilateral correlation degree of the two associated object nodes based on each associated node according to the social network information. By superimposing the unilateral association degrees of multiple associated nodes between two object nodes, the comprehensive association degree of the two object nodes can be obtained.
- Step 214 Obtain the attribute information value corresponding to the reference object node.
- the server obtains the identity attribute information corresponding to each reference object node, and extracts the attribute value of the attribute information corresponding to the missing attribute from the obtained identity attribute information (denoted as the attribute information value). It is easy to understand that the attribute value of the attribute information in the identity attribute information corresponding to some reference object nodes may also be missing. However, the fully connected sub-network graph includes at least one determined object node, so at least one attribute information value can be extracted.
- Step 216 Determine the reference weight of the attribute information value according to the comprehensive association degree.
- Step 218 Determine the attribute information value corresponding to the target object according to the multiple attribute information values and the respective reference weights, so as to control the terminal to perform data processing according to the attribute information value.
- the server performs a preset logical operation on the attribute information value to obtain the attribute information value corresponding to the target object.
- the preset logic operation may be a superposition operation based on the reference weight.
- the reference weight may be determined according to the comprehensive correlation between the corresponding reference object node and the pending object node.
- the preset logical operation may also be to take the median or average of multiple attribute information values.
- risk control can be performed on the corresponding target object. For example, when the attribute information value is higher than the threshold value, relevant business services are not provided, or additional business liability clauses are further generated based on high-risk conditions, so as to avoid risks and ensure the effect of risk control.
- the association relationships covering large-scale groups of people are collectively displayed in a relationship network graph, which is convenient for users to grasp the association relationships between customers from a global perspective.
- the identity attribute information of the target object is obtained from the data processing system; the information prediction model and correlation coefficient weight corresponding to the data processing system are inquired; the identity attribute information is input into the information prediction model to obtain the target correlation coefficient; The coefficient and the weight of the correlation coefficient are calculated to obtain the first attribute reference value; the second attribute reference value is calculated according to the multiple attribute information values and the corresponding reference weights; the second attribute reference value is calculated according to the first attribute reference value and the second attribute reference value The attribute information value corresponding to the target object.
- the server has preset corresponding information prediction models for data processing systems that implement different types of business.
- the information prediction model may be constructed based on the identity attribute information of the blacklisted objects determined in the data processing system, and used for risk assessment of other target objects in the data processing system.
- the identity attribute information of the same target object may be extracted from multiple data processing systems.
- the server In order to conduct a comprehensive risk analysis of the target object, the server also presets the corresponding correlation coefficient weight for each data processing system.
- the correlation coefficient weight is used to comprehensively process the risk assessment results output by the information prediction model corresponding to each data processing system, and divide different weight ratios for the different importance of each data processing system to obtain a weighted comprehensive evaluation result.
- the different importance of each data processing system for risk assessment can be characterized based on the weight of the correlation coefficient.
- the weight of the correlation coefficient for various insurance data processing systems is higher, while for other Data processing systems that are less relevant to the insurance business can be set with lower correlation coefficient weights to highlight the importance of each insurance data processing system in comprehensive risk control, and finally get a comprehensive that meets the comprehensive risk control requirements of the insurance data processing system evaluation result.
- the correlation coefficient weight can be customized according to the actual needs of comprehensive risk control to meet different risk control needs.
- the information prediction model can be, but is not limited to, machine learning models, such as NN (Neural Networks), linear classifiers, SVM (Support Vector Machine, support vector machine models), naive Bayes models, and K-nearest neighbors Common learning models such as algorithm models. Risk control is carried out through the preset information prediction model, which avoids the influence of human factors, strengthens the pertinence of the data processing system in risk control, and improves the accuracy of risk control.
- NN Neurological Networks
- SVM Small Vector Machine, support vector machine models
- K-nearest neighbors Common learning models such as algorithm models.
- the server calculates the first attribute reference value corresponding to the target object when the data processing system performs comprehensive risk control according to the target correlation coefficient and the query correlation coefficient weight.
- the first attribute reference value reflects the comprehensive risk assessment results of the target object in each data processing system. By adjusting the weight of the correlation coefficient according to the comprehensive risk control requirement, the first attribute reference value meeting different risk control requirements can be obtained.
- the server performs a superposition operation based on a preset weight on the first attribute reference value and the second attribute reference value to obtain the attribute information value corresponding to the object to be determined.
- the preset weight may be a fixed value.
- the server pre-trains the corresponding missing attribute completion model.
- the server may give the calculated confidence of the second attribute reference value according to the accuracy of the model, so the preset weight may also be dynamically determined according to the confidence of the second attribute reference value.
- the first attribute reference value can be calculated based on the identity attribute information of the target object obtained in the data processing system, and the preset information prediction model and correlation coefficient weight corresponding to the data processing system;
- the connected subnet graph can determine the second attribute reference value of the attribute information of the target object; according to the first attribute reference value and the second attribute reference value, the attribute information value corresponding to the target object can be obtained.
- the identity attribute information of the target object itself, it also considers the identity attribute information of the reference object node that has a strong correlation with the target object node, and predicts the attribute information of the target object by integrating multiple dimensional factors, which can not only improve the information
- the completion efficiency can also improve the accuracy of the completion information.
- before querying the information prediction model and correlation coefficient weight corresponding to the data processing system it further includes: obtaining the identity attribute information of the sample object and the corresponding labeled target correlation coefficient; constructing the input layer according to the identity attribute information of the sample object , Construct the output layer according to the marked target correlation coefficient, obtain the preset mapping parameters, construct the middle layer according to the mapping parameters; construct the prediction model for the information to be trained according to the input layer, middle layer and output layer; obtain blacklist samples from the data processing system Set, divide the blacklist sample set into mutually exclusive training sample set and test sample set; input the training sample set into the information prediction model to be trained, adjust the mapping parameters according to the output result of the information prediction model to be trained, and obtain the information prediction model after training ; Evaluate and test the trained information prediction model through the test sample set. When the test result reaches the threshold, use the trained information prediction model as the information prediction model.
- the server can preprocess the identity attribute information of multiple sample objects such as encoding and sorting to obtain the input vector, and construct the input layer of the information prediction model according to the input vector.
- the server preprocesses the labeled target correlation coefficients in the above manner to obtain the output vector, and constructs the output layer of the information prediction model according to the output vector.
- the server sets the mapping parameters between the input vector and the output vector, and builds the middle layer of the information prediction model according to the mapping parameters.
- the mapping parameters can be weight ratios, functional formulas, and so on.
- the mapping parameters of the middle layer can be RBF ((Radial Basis Function, Radial Basis Function) kernel function, linear kernel function, polynomial kernel function , Sigmoid kernel function, etc. If the RBF kernel function is used, the intermediate layer can be adjusted by adjusting the penalty factor C and the kernel parameter ⁇ in the RBF kernel function, and finally a suitable information prediction model can be obtained.
- RBF Random Basis Function, Radial Basis Function
- the intermediate layer can be adjusted by adjusting the penalty factor C and the kernel parameter ⁇ in the RBF kernel function, and finally a suitable information prediction model can be obtained.
- the server constructs the information prediction model to be trained based on the input layer, the middle layer and the output layer.
- the server trains the information prediction model to be trained through the known blacklisted customer data. Specifically, the server obtains blacklisted customer data from the data processing system.
- the blacklisted customer data includes customer characteristic information corresponding to the blacklisted customer, and the customer characteristic information can be used to effectively train the training information prediction model.
- the training sample set is input into the information prediction model to be trained, the mapping parameters are adjusted according to the output of the information prediction model to be trained and the verification data in the training sample set, and the adjusted mapping parameters are adjusted After the middle layer, finally build a trained information prediction model based on the input layer, adjusted middle layer and output layer.
- the combination of the penalty factor and the kernel parameter (C, ⁇ ) can be continuously adjusted during the training process, and finally a trained SVM model that meets the needs can be obtained.
- the method before obtaining the relationship network graph, further includes: obtaining the identity attribute information and social network information of a plurality of target objects; generating the object node and attribute node corresponding to each target object according to the identity attribute information, using directed Connect the attribute node to the corresponding object node; identify whether different object nodes are connected with the same attribute node; if so, merge the same attribute node, and mark the same attribute node as an associated node corresponding to multiple object nodes; according to The social network information calculation object node is based on the unilateral relevance of different associated nodes, and the unilateral relevance is added to the directed edges connected to the corresponding associated nodes to obtain the relationship network graph.
- the server obtains the identity attribute information and social network information of multiple target objects.
- the target audience can be lost customers, existing customers, or potential customers.
- potential customers can be identified based on customer data of lost customers or existing customers. For example, guarantors or emergency contacts reserved by existing customers can be used as potential customers.
- Potential customers can also be obtained by monitoring the product review records left by users browsing related products on the target website, and there is no restriction on this.
- the identity attribute information includes object identification.
- the object identifier can be an ID number, mobile phone number, or email address. It is easy to understand that if the target object is an enterprise, the object identifier can also be an organization code, etc.
- Identity attribute information also includes name, gender, age, education, contact information, employer, insurance policy, bank card account, terminal device information, social network account, interest, wealth level, or risk level, etc.
- Social network information includes wifi connection information, location sharing information, instant messaging information, electronic transfer information or remote call information, etc.
- the server generates an object node corresponding to the target object according to the object identifier, and generates one or more attribute nodes corresponding to the target object according to other identity attribute information.
- an attribute node can be generated with the job unit as the identifier, or an attribute node can be generated with the transfer ID as the identifier.
- Each attribute node is associated with a corresponding node description.
- the directed edge points from the object node to the attribute node.
- An object node can be connected to multiple attribute nodes.
- the server recognizes whether different object nodes are connected with the same attribute node. If so, the server merges the same attribute node, and marks the merged attribute node as an associated node corresponding to multiple object nodes.
- the same object node can be connected to multiple types of attribute nodes, such as the type of job unit, bank card account type, and common network type. By merging the same attribute nodes, multiple object nodes can be associated. In other words, the association relationship between multiple target objects can be identified according to the identity attribute information.
- having the same "employment unit” attribute node indicates that two target objects may have a colleague relationship; having the same "educational background” attribute node indicates that two target objects may have a classmate relationship; having the same "bank card account number” or
- the "common network type” attribute node indicates the possible kinship of two target objects, etc., and each type of association relationship is analyzed in this way.
- Two object nodes connected with one or more associated nodes are related.
- the server calculates the two related object nodes based on the social network information based on the unilateral association degree of each associated node, and adds the unilateral association degree to the directed edges connected to the corresponding associated nodes to obtain the relationship network graph.
- the comprehensive association degree of the two object nodes can be obtained.
- the implicit association relationship can be automatically extracted, which can greatly improve the efficiency of obtaining the association relationship compared with the traditional manual analysis method. Not only identify the association relationship between multiple target objects based on the identity attribute information, but also based on the association strength of the associated target objects based on social network information mining, which can expand the dimension of information mining and increase the depth of information mining, thereby improving the accuracy of relationship network mining .
- object nodes can be added at any time based on attribute nodes, which facilitates the extension of the associated network, and can gradually increase the scale of the population covered by the relationship network graph.
- calculating the comprehensive association degree between each reference object node and the undetermined object node includes: obtaining social network information corresponding to the undetermined object node; and calculating the undetermined object node based on different associated nodes and different reference object nodes according to the social network information The unilateral correlation degree between each reference object node and the undetermined object node is superimposed on the unilateral correlation degree of multiple associated nodes to obtain the comprehensive relevance degree of the reference object node and the undetermined object node.
- the relationship type can be family relationship, classmate relationship, colleague relationship, friend relationship, transfer relationship, location proximity relationship, etc.
- the server can identify the type of relationship between the two associated object nodes according to the identity attribute information and the social network information. For example, by connecting the same home wifi, corporate wifi, or public wifi with the target object A and the target objects B, C, and D, it can be recognized that they may have related relationships such as relatives, friends, colleagues, or nearby people.
- Different basic correlation coefficients can be preset for different relationship types. There may be multiple associations between two target objects. For example, target objects A and B can be classmates, colleagues, and friends. In this case, the server may also preset different basic correlation coefficients for different combination of relationship types.
- mapping relationships between different relationship types and basic correlation coefficients, or mapping relationships between different relationship type combinations and basic correlation coefficients can be preset. For example, when the relationship network map is used to mine customer interest attributes, the immediate family relationship is set to 1, and the colleague relationship is set to 0.5, etc.; when the relationship network map is used to review the customer's risk attributes, the friend relationship is set to 1, and the kinship relationship is set to 1. Set to 0.4 and so on.
- the basic correlation coefficients corresponding to multiple purposes are preset to realize a variety of unilateral correlation calculation methods, and the value meaning of each relationship type for evaluating the correlation degree can be fully considered, thereby improving the relationship network graph The accuracy of supplementing different missing attributes.
- the server determines the shortest social distance between two associated object nodes based on the relationship network graph.
- the shortest social distance refers to the number of associated nodes that must pass at least from one object node to another object node. For example, in Figure 3 of the above example, the shortest social distance between the object nodes V3 and V5 is 1, and the shortest social distance between the object nodes V4 and V8 is 2.
- the server counts the event type and frequency of the associated event that occurred in the statistical period of the two object nodes associated with the social network information.
- Related events can be interactive operations such as connecting to the same local area network, sending social information based on an instant messaging platform, or bank card transfer.
- the server is preset with multiple event types, multiple occurrence frequency intervals corresponding to each event type, and first adjustment coefficients corresponding to each occurrence frequency interval.
- the server also presets a variety of second adjustment coefficients corresponding to the shortest social distance. According to the first adjustment coefficient and the second adjustment coefficient, the basic correlation coefficient is increased or reduced to obtain the target correlation coefficient.
- the server marks the target correlation coefficient as the unilateral correlation degree of the corresponding object node based on the corresponding associated node.
- the directed edge connected to the corresponding object node shows the corresponding unilateral association degree.
- the unilateral association degree between the associated node Q5183 and the two connected object nodes V5 and V8 is 3.21, that is The unilateral correlation degree of the target node V5 and the target node V8 based on the associated node Q5183 is 3.21.
- the unilateral correlation degree between the associated node M5482 and the two connected object nodes V5 and V8 is 0.89.
- the server superimposes the unilateral association degrees of multiple associated nodes between the two object nodes to obtain the comprehensive association degrees of the two object nodes.
- the server superimposes the comprehensive association degree as the reference weight and the attribute information value of the corresponding reference object node to obtain the second attribute reference value.
- adjusting the basic correlation coefficient can improve the calculation accuracy of the unilateral correlation degree, and thus the accuracy of the interest attribute value.
- the method further includes the step of risk tracking based on the attribute information value:
- Step 402 Compare whether the attribute information value corresponding to the target object exceeds a threshold.
- Step 404 if yes, determine the industry type to which the target object belongs.
- Step 406 Obtain risk data; the risk data includes the risk data of the target object, the risk data of the industry type, and the risk data of the product resource corresponding to the target object.
- Step 408 Extract a risk label from the risk data.
- Step 410 Calculate the similarity between the extracted risk label and the risk labels of a plurality of pre-stored blacklist objects, and mark the blacklist objects whose similarity exceeds a preset value as similar objects.
- the server compares whether the value of the attribute information exceeds the threshold. If so, it means that the target object has a higher risk of default, and the server marks the target object as a risk object.
- the server determines the similar object corresponding to the risk object according to the identity attribute information. Specifically, the server pre-stores a variety of risk portraits of blacklisted objects (denoted as bad sample portraits).
- the bad sample portrait includes multiple risk labels.
- the risk label is used to characterize the subjects in which the blacklisted objects have problems. As time changes, risk measures may also change. In order to improve the matching accuracy of similar objects, the corresponding bad sample images can also be dynamically updated.
- the server Based on the identity attribute information of the target object, the server generates risk labels corresponding to various preset risk indicators, and uses the multiple risk labels to generate a risk profile of the target object (denoted as a portrait to be matched).
- the server invokes a preset risk cue analysis model to calculate the cosine similarity between the image to be matched and the image of the bad sample to obtain the similarity. If the similarity exceeds the specified value, the server marks the corresponding blacklist object as a similar object of the risk object.
- Step 412 Obtain risk indicators of similar objects at multiple time nodes, determine the risk points of the target object according to the risk indicators, and add the risk points to the identity attribute information.
- the server identifies multiple risk points of risk objects based on similar objects. Specifically, each bad sample profile is associated with risk indicators of multiple time nodes. The server predicts the risk clues of the risk object based on the risk indicators of multiple time nodes associated with the matching bad sample portrait. At different time nodes, the risk patterns of different blacklisted objects may be similar. In other words, similar objects of current risk objects may change over time.
- the server generates risk clues from two perspectives: "same risk label as similar objects" and "time sequence of same risk label”. Specifically, it can be judged whether the risk target has the same risk index as the similar target and whether the time sequence of the same risk index is consistent with similar cases. If there is the same risk indicator as the similar object and the appearance sequence of the same risk indicator is consistent with the similar object, the server marks the same risk indicator (recorded as a symptom indicator) at the last time node as a risk point. For example, if the bad sample A has 6 abnormal indicators, and the target object B has 5 abnormal indicators, it is predicted that the sixth abnormal indicator may appear, so that the sixth abnormal indicator can be marked as a risk point of the target object B .
- the server connects multiple risk points in series to generate risk clues corresponding to the risk object. Specifically, determine the monitoring cycle of risk objects.
- the monitoring period can be dynamically determined based on the risk score or the industry type of the risk object, or it can be a preset fixed value, which is not limited.
- the server determines the risk points of the risk object in each monitoring period according to the above method, and connects multiple risk points in series in chronological order to obtain risk clues corresponding to the risk object.
- the server generates a corresponding risk analysis report based on the attribute information value, similar objects, and risk clues, and associates the risk analysis report with the corresponding pending object node in the relationship network graph.
- the similar objects of the target object are further determined, and based on the similar objects, the risk points that the target object may appear at multiple time nodes in the future are predicted.
- the identity attribute information supplemented based on the above information can be convenient for users Comprehensively and quickly understand the risk situation of the target object and improve the accuracy of risk analysis.
- an attribute information processing device based on a relational network
- an attribute information processing module and an attribute information value calculation module 506, wherein the attribute information processing module includes a first attribute information processing The module 502 and the second attribute information processing module 504, wherein:
- the second attribute information processing module 504 is used to obtain a relationship network graph; the relationship network graph includes multiple object nodes and attribute nodes connected to each object node; identify whether the corresponding attribute information of each object node is missing;
- the multiple object nodes in the relationship network graph are divided into determined object nodes and pending object nodes; the fully connected sub-network graph of each pending object node is drawn in the relationship network graph; the fully connected sub-network
- the determined object node that has a target association relationship with the undetermined object node is marked as a reference object node; calculate the comprehensive association degree between each reference object node and the undetermined object node; obtain the attribute information value corresponding to the reference object node, and
- the reference weight of the attribute information value is determined according to the comprehensive association degree; the second attribute reference value is calculated according to multiple attribute information values and corresponding reference weights.
- the attribute information value calculation module 506 is configured to calculate a second attribute reference value according to a plurality of attribute information values and corresponding reference weights, so as to control the terminal to perform data processing according to the second attribute information value.
- the first attribute information processing module 502 is used to obtain the identity attribute information of the target object from the data processing system; query the information prediction model and correlation coefficient weight corresponding to the data processing system; and convert the identity attribute information
- the information prediction model is input to obtain the target correlation coefficient; according to the target correlation coefficient and the correlation coefficient weight, the first attribute reference value is calculated.
- the second attribute information processing module 504 is further configured to calculate a second attribute reference value according to multiple attribute information values and corresponding reference weights.
- the attribute information value calculation module 506 is further configured to calculate the attribute information value of the target object according to the first attribute reference value and the second attribute reference value.
- the device further includes a network graph construction module 508, which is used to obtain the identity attribute information and social network information of multiple target objects; generate the object node and attribute node corresponding to each target object according to the identity attribute information, and use Directed edges connect attribute nodes to corresponding object nodes; identify whether different object nodes are connected to the same attribute node; if so, merge the same attribute nodes, and mark the same attribute nodes as associated nodes corresponding to multiple object nodes Calculate the unilateral relevance of the object node based on different associated nodes according to the social network information, and add the unilateral relevance to the directed edges connected to the corresponding associated nodes to obtain the relationship network graph.
- a network graph construction module 508 which is used to obtain the identity attribute information and social network information of multiple target objects; generate the object node and attribute node corresponding to each target object according to the identity attribute information, and use Directed edges connect attribute nodes to corresponding object nodes; identify whether different object nodes are connected to the same attribute node; if so, merge the same attribute nodes,
- the second attribute information processing module 504 is also used to obtain social network information corresponding to the undetermined object node; calculate the undetermined object node based on the social network information based on the unilateral association degree between different associated nodes and different reference object nodes ; Superimpose the unilateral association degrees of multiple associated nodes between each reference object node and the undetermined object node to obtain the comprehensive association degree of the reference object node and the undetermined object node.
- the device further includes an attribute information tracking module 510, which is used to compare whether the attribute information value corresponding to the target object exceeds a threshold; if so, determine the type of industry to which the target object belongs; obtain risk data; Risk data, industry type risk data, and target object corresponding product resource risk data; extract risk labels from the risk data; calculate the similarity between the extracted risk labels and the risk labels of multiple blacklist objects stored in advance, which will be similar Blacklist objects whose degrees exceed the preset value are marked as similar objects; the risk indicators of similar objects at multiple time nodes are obtained, the risk points of the target object are determined according to the risk indicators, and the risk points are added to the identity attribute information.
- an attribute information tracking module 510 which is used to compare whether the attribute information value corresponding to the target object exceeds a threshold; if so, determine the type of industry to which the target object belongs; obtain risk data; Risk data, industry type risk data, and target object corresponding product resource risk data; extract risk labels from the risk data; calculate the similarity between the extracted risk labels and the risk labels
- Each module in the above-mentioned relational network-based attribute information processing device may be implemented in whole or in part by software, hardware, and a combination thereof.
- the foregoing modules may be embedded in the form of hardware or independent of the processor in the computer device, or may be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the foregoing modules.
- a computer device is provided.
- the computer device may be a server, and its internal structure diagram may be as shown in FIG. 6.
- the computer equipment includes a processor, a memory, a network interface and a database connected through a system bus. Among them, the processor of the computer device is used to provide calculation and control capabilities.
- the memory of the computer device includes a non-volatile storage medium and an internal memory.
- the non-volatile storage medium stores an operating system, a computer program, and a database.
- the internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium.
- the database of the computer equipment is used to store information prediction models and relational network graphs.
- the network interface of the computer device is used to communicate with an external terminal through a network connection.
- FIG. 6 is only a block diagram of part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied.
- the specific computer device may Including more or less parts than shown in the figure, or combining some parts, or having a different part arrangement.
- a computer-readable storage medium has a computer program stored thereon, and when the computer program is executed by a processor, the steps of the method for processing attribute information based on a relational network provided in any embodiment of the present application are realized.
- Non-volatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory.
- Volatile memory may include random access memory (RAM) or external cache memory.
- RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Channel (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.
- SRAM static RAM
- DRAM dynamic RAM
- SDRAM synchronous DRAM
- DDRSDRAM double data rate SDRAM
- ESDRAM enhanced SDRAM
- SLDRAM synchronous chain Channel
- memory bus Radbus direct RAM
- RDRAM direct memory bus dynamic RAM
- RDRAM memory bus dynamic RAM
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Marketing (AREA)
- Game Theory and Decision Science (AREA)
- General Business, Economics & Management (AREA)
- Educational Administration (AREA)
- Operations Research (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Description
Claims (19)
- 一种基于关系网络的属性信息处理方法,所述方法包括:获取关系网络图谱;所述关系网络图谱包括多个对象节点及每个对象节点连接的属性节点;识别每个对象节点对应属性信息是否缺失;根据识别结果将所述关系网络图谱中多个对象节点区分为确定对象节点和待定对象节点;在所述关系网络图谱中划取每个所述待定对象节点的全连通子网络图;将所述全连通子网络图中与待定对象节点存在目标关联关系的确定对象节点标记为参考对象节点;计算每个参考对象节点与所述待定对象节点的综合关联度;根据所述综合关联度确定所述属性信息值的参考权重;根据多个属性信息值及分别对应的参考权重,确定所述目标对象对应的属性信息值,以控制终端根据所述属性信息值进行数据处理。
- 根据权利要求1所述的方法,其中,所述获取关系网络图谱之前,所述方法还包括:获取多个目标对象的身份属性信息和社交网络信息;根据所述身份属性信息生成每个目标对象对应的对象节点及属性节点,采用有向边将所述属性节点连接至相应对象节点;识别不同所述对象节点是否连接有相同的属性节点;若是,对所述相同的属性节点进行合并,并将所述相同的属性节点标记为对应多个对象节点的关联节点;根据所述社交网络信息计算所述对象节点基于不同关联节点的单边关联度,将所述单边关联度添加至相应关联节点相连的有向边,得到关系网络图谱。
- 根据权利要求2所述的方法,其中,所述计算每个参考对象节点与所述待定对象节点的综合关联度,包括:获取所述待定对象节点对应的社交网络信息;根据所述社交网络信息计算所述待定对象节点基于不同关联节点与不同参考对象节点之间的单边关联度;对每个参考对象节点与所述待定对象节点之间多个关联节点的单边关联度进行叠加,得到所述参考对象节点与所述待定对象节点的综合关联度。
- 根据权利要求1所述的方法,其中,所述根据多个属性信息值及分别对应的参考权重,计算得到所述目标对象对应的属性信息值包括:从数据处理系统获取目标对象的身份属性信息;查询与所述数据处理系统对应的信息预测模型和关联系数权重;将所述身份属性信息输入所述信息预测模型,得到目标关联系数;根据所述目标关联系数及所述关联系数权重,计算得到第一属性参考值;根据多个属性信息值及分别对应的参考权重,计算得到第二属性参考值;根据所述第一属性参考值及所述第二属性参考值,计算得到所述目标对象的属性信息值。
- 根据权利要求1所述的方法,其中,所述方法还包括:比较所述目标对象对应的属性信息值是否超过阈值;若是,确定所述目标对象所属的行业类型;获取风险数据;所述风险数据包括所述目标对象的风险数据、所述行业类型的风险数据以及所述目标对象对应产品资源的风险数据;在所述风险数据中提取风险标签;计算提取得到的风险标签与预存储的多个黑名单对象的风险标签的相似度,将所述相似度超过预设值的黑名单对象标记为相似对象;获取所述相似对象在多个时间节点的风险指标,根据所述风险指标确定所述目标对象的风险点,将所述风险点补入所述身份属性信息。
- 一种基于关系网络的属性信息处理装置,所述装置包括:属性信息处理模块,用于获取关系网络图谱;所述关系网络图谱包括多个对象节点及每个对象节点连接的属性节点;识别每个对象节点对应属性信息是否缺失;根据识别结果将所述关系网络图谱中多个对象节点区分为确定对象节点和待定对象节点;在所述关系网络图谱中划取每个所述待定对象节点的全连通子网络图;将所述全连通子网络图中与待定对象节点存在目标关联关系的确定对象节点标记为参考对象节点;计算每个参考对象节点与所述待定对象节点的综合关联度;获取所述参考对象节点对应的属性信息值,并根据所述综合关联度确定所述属性信息值的参考权重;根据多个属性信息值及分别对应的参考权重,计算得到第二属性参考值;属性信息值计算模块,用于根据所述第一属性参考值与所述第二属性参考值,确定所述目标对象对应的属性信息值,以控制终端根据所述属性信息值进行数据处理。
- 根据权利要求6所述的装置,其中,所述装置还包括网络图谱构建模块,用于获取多个目标对象的身份属性信息和社交网络信息;根据所述身份属性信息生成每个目标对象对应的对象节点及属性节点,采用有向边将所述属性节点连接至相应对象节点;识别不同所述对象节点是否连接有相同的属性节点;若是,对所述相同的属性节点进行合并,并将所述相同的属性节点标记为对应多个对象节点的关联节点;根据所述社交网络信息计算所述对象节点基于不同关联节点的单边关联度,将所述单边关联度添加至相应关联节点相连的有向边,得到关系网络图谱。
- 根据要求要求7所述的装置,其中,所述属性信息处理模块还用于获取所述待定对象节点对应的社交网络信息;根据所述社交网络信息计算所述待定对象节点基于不同关联节点与不同参考对象节点之间的单边关联度;对每个参考对象节点与所述待定对象节点之间多个关联节点的单边关联度进行叠加,得到所述参考对象节点与所述待定对象节点的综合关联度。
- 根据权利要求6所述的基于关系网络的属性信息处理装置,其中,所述装置还包括属性信息跟踪模块用于比较所述目标对象对应的属性信息值是否超过阈值;若是,确定所述目标对象所属的行业类型;获取风险数据;所述风险数据包括所述目标对象的风险数据、所述 行业类型的风险数据以及所述目标对象对应产品资源的风险数据;在所述风险数据中提取风险标签;计算提取得到的风险标签与预存储的多个黑名单对象的风险标签的相似度,将所述相似度超过预设值的黑名单对象标记为相似对象;获取所述相似对象在多个时间节点的风险指标,根据所述风险指标确定所述目标对象的风险点,将所述风险点补入所述身份属性信息。
- 一种计算机设备,包括存储器和处理器,所述存储器存储有计算机程序,其中,所述处理器执行所述计算机程序时实现如下步骤:获取关系网络图谱;所述关系网络图谱包括多个对象节点及每个对象节点连接的属性节点;识别每个对象节点对应属性信息是否缺失;根据识别结果将所述关系网络图谱中多个对象节点区分为确定对象节点和待定对象节点;在所述关系网络图谱中划取每个所述待定对象节点的全连通子网络图;将所述全连通子网络图中与待定对象节点存在目标关联关系的确定对象节点标记为参考对象节点;计算每个参考对象节点与所述待定对象节点的综合关联度;根据所述综合关联度确定所述属性信息值的参考权重;根据多个属性信息值及分别对应的参考权重,确定所述目标对象对应的属性信息值,以控制终端根据所述属性信息值进行数据处理。
- 根据权利要求10所述的计算机设备,其中,所述获取关系网络图谱之前,所述处理器还执行所述计算机程序实现:获取多个目标对象的身份属性信息和社交网络信息;根据所述身份属性信息生成每个目标对象对应的对象节点及属性节点,采用有向边将所述属性节点连接至相应对象节点;识别不同所述对象节点是否连接有相同的属性节点;若是,对所述相同的属性节点进行合并,并将所述相同的属性节点标记为对应多个对象节点的关联节点;根据所述社交网络信息计算所述对象节点基于不同关联节点的单边关联度,将所述单边关联度添加至相应关联节点相连的有向边,得到关系网络图谱。
- 根据权利要求11所述的计算机设备,其中,所述处理器执行所述计算机程序时实现所述计算每个参考对象节点与所述待定对象节点的综合关联度,包括:获取所述待定对象节点对应的社交网络信息;根据所述社交网络信息计算所述待定对象节点基于不同关联节点与不同参考对象节点之间的单边关联度;对每个参考对象节点与所述待定对象节点之间多个关联节点的单边关联度进行叠加,得到所述参考对象节点与所述待定对象节点的综合关联度。
- 根据权利要求10所述的计算机设备,其中,所述处理器执行所述计算机程序时实现 所述根据多个属性信息值及分别对应的参考权重,计算得到所述目标对象对应的属性信息值包括:从数据处理系统获取目标对象的身份属性信息;查询与所述数据处理系统对应的信息预测模型和关联系数权重;将所述身份属性信息输入所述信息预测模型,得到目标关联系数;根据所述目标关联系数及所述关联系数权重,计算得到第一属性参考值;根据多个属性信息值及分别对应的参考权重,计算得到第二属性参考值;根据所述第一属性参考值及所述第二属性参考值,计算得到所述目标对象的属性信息值。
- 根据权利要求10所述的计算机设备,其中,所述处理器还执行所述计算机程序时实现:比较所述目标对象对应的属性信息值是否超过阈值;若是,确定所述目标对象所属的行业类型;获取风险数据;所述风险数据包括所述目标对象的风险数据、所述行业类型的风险数据以及所述目标对象对应产品资源的风险数据;在所述风险数据中提取风险标签;计算提取得到的风险标签与预存储的多个黑名单对象的风险标签的相似度,将所述相似度超过预设值的黑名单对象标记为相似对象;获取所述相似对象在多个时间节点的风险指标,根据所述风险指标确定所述目标对象的风险点,将所述风险点补入所述身份属性信息。
- 一种计算机可读存储介质,其上存储有计算机程序,其中,所述计算机程序被处理器执行时实现如下步骤:获取关系网络图谱;所述关系网络图谱包括多个对象节点及每个对象节点连接的属性节点;识别每个对象节点对应属性信息是否缺失;根据识别结果将所述关系网络图谱中多个对象节点区分为确定对象节点和待定对象节点;在所述关系网络图谱中划取每个所述待定对象节点的全连通子网络图;将所述全连通子网络图中与待定对象节点存在目标关联关系的确定对象节点标记为参考对象节点;计算每个参考对象节点与所述待定对象节点的综合关联度;根据所述综合关联度确定所述属性信息值的参考权重;根据多个属性信息值及分别对应的参考权重,确定所述目标对象对应的属性信息值,以控制终端根据所述属性信息值进行数据处理。
- 根据权利要求15所述的计算机可读存储介质,其中,所述获取关系网络图谱之前,所述计算机程序还被处理器执行实现:获取多个目标对象的身份属性信息和社交网络信息;根据所述身份属性信息生成每个目标对象对应的对象节点及属性节点,采用有向边将所述属性节点连接至相应对象节点;识别不同所述对象节点是否连接有相同的属性节点;若是,对所述相同的属性节点进行合并,并将所述相同的属性节点标记为对应多个对象节点的关联节点;根据所述社交网络信息计算所述对象节点基于不同关联节点的单边关联度,将所述单边关联度添加至相应关联节点相连的有向边,得到关系网络图谱。
- 根据权利要求16所述的计算机可读存储介质,其中,所述计算机程序被处理器执行实现所述计算每个参考对象节点与所述待定对象节点的综合关联度,包括:获取所述待定对象节点对应的社交网络信息;根据所述社交网络信息计算所述待定对象节点基于不同关联节点与不同参考对象节点之间的单边关联度;对每个参考对象节点与所述待定对象节点之间多个关联节点的单边关联度进行叠加,得到所述参考对象节点与所述待定对象节点的综合关联度。
- 根据权利要求15所述的计算机可读存储介质,其中,所述计算机程序被处理器执行实现所述根据多个属性信息值及分别对应的参考权重,计算得到所述目标对象对应的属性信息值包括:从数据处理系统获取目标对象的身份属性信息;查询与所述数据处理系统对应的信息预测模型和关联系数权重;将所述身份属性信息输入所述信息预测模型,得到目标关联系数;根据所述目标关联系数及所述关联系数权重,计算得到第一属性参考值;根据多个属性信息值及分别对应的参考权重,计算得到第二属性参考值;根据所述第一属性参考值及所述第二属性参考值,计算得到所述目标对象的属性信息值。
- 根据权利要求15所述的计算机可读存储介质,其中,所述计算机程序还被处理器执行实现:比较所述目标对象对应的属性信息值是否超过阈值;若是,确定所述目标对象所属的行业类型;获取风险数据;所述风险数据包括所述目标对象的风险数据、所述行业类型的风险数据以及所述目标对象对应产品资源的风险数据;在所述风险数据中提取风险标签;计算提取得到的风险标签与预存储的多个黑名单对象的风险标签的相似度,将所述相似度超过预设值的黑名单对象标记为相似对象;获取所述相似对象在多个时间节点的风险指标,根据所述风险指标确定所述目标对象的风险点,将所述风险点补入所述身份属性信息。
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910748766.0A CN110659799A (zh) | 2019-08-14 | 2019-08-14 | 基于关系网络的属性信息处理方法、装置、计算机设备和存储介质 |
CN201910748766.0 | 2019-08-14 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2021027317A1 true WO2021027317A1 (zh) | 2021-02-18 |
Family
ID=69037389
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2020/086397 WO2021027317A1 (zh) | 2019-08-14 | 2020-04-23 | 基于关系网络的属性信息处理方法、装置、计算机设备和存储介质 |
Country Status (2)
Country | Link |
---|---|
CN (1) | CN110659799A (zh) |
WO (1) | WO2021027317A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112926993A (zh) * | 2021-04-13 | 2021-06-08 | 郭栋 | 基于区块链安全大数据的信息生成方法及区块链服务系统 |
Families Citing this family (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110659799A (zh) * | 2019-08-14 | 2020-01-07 | 深圳壹账通智能科技有限公司 | 基于关系网络的属性信息处理方法、装置、计算机设备和存储介质 |
CN111210008B (zh) * | 2020-01-09 | 2022-05-24 | 支付宝(杭州)信息技术有限公司 | 利用lstm神经网络模型处理交互数据的方法及装置 |
CN111274495B (zh) * | 2020-01-20 | 2023-08-25 | 平安科技(深圳)有限公司 | 用户关系强度的数据处理方法、装置、计算机设备及存储介质 |
CN111325258B (zh) * | 2020-02-14 | 2023-10-24 | 腾讯科技(深圳)有限公司 | 特征信息获取方法、装置、设备及存储介质 |
CN111343161B (zh) * | 2020-02-14 | 2021-12-10 | 平安科技(深圳)有限公司 | 异常信息处理节点分析方法、装置、介质及电子设备 |
CN111340611B (zh) * | 2020-02-20 | 2024-03-08 | 中国建设银行股份有限公司 | 一种风险预警方法和装置 |
CN111597401B (zh) * | 2020-05-20 | 2021-08-17 | 腾讯科技(深圳)有限公司 | 基于图关系网络的数据处理方法、装置、设备及介质 |
CN113704363B (zh) * | 2020-05-22 | 2024-04-30 | 百度在线网络技术(北京)有限公司 | 一种权重确定方法、装置、设备及存储介质 |
CN111860655B (zh) * | 2020-07-23 | 2023-05-30 | 中国联合网络通信集团有限公司 | 用户的处理方法、装置和设备 |
CN112016836B (zh) * | 2020-08-31 | 2023-11-03 | 中国银联股份有限公司 | 一种对象间相似性的确定方法及装置 |
CN112270195A (zh) * | 2020-11-25 | 2021-01-26 | 上海水滴征信服务有限公司 | 确定企业关联关系、重名对象判定 |
CN112667696A (zh) * | 2020-12-30 | 2021-04-16 | 深圳市华傲数据技术有限公司 | 数据处理方法、设备及存储介质 |
CN113157704B (zh) * | 2021-05-06 | 2023-07-25 | 成都卫士通信息产业股份有限公司 | 层级关系分析方法、装置、设备及计算机可读存储介质 |
CN114866437B (zh) * | 2022-04-19 | 2023-11-21 | 北京博睿宏远数据科技股份有限公司 | 一种节点检测方法、装置、设备及介质 |
CN116931889B (zh) * | 2023-09-18 | 2023-12-12 | 浙江工企信息技术股份有限公司 | 一种基于对象树的软件建模方法及系统 |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107239882A (zh) * | 2017-05-10 | 2017-10-10 | 平安科技(深圳)有限公司 | 风险评估方法、装置、计算机设备及存储介质 |
CN108132998A (zh) * | 2017-12-21 | 2018-06-08 | 浪潮软件集团有限公司 | 一种人员关系分析方法和系统 |
CN109272396A (zh) * | 2018-08-20 | 2019-01-25 | 平安科技(深圳)有限公司 | 客户风险预警方法、装置、计算机设备和介质 |
CN109345158A (zh) * | 2018-12-19 | 2019-02-15 | 重庆百行智能数据科技研究院有限公司 | 企业风险识别方法、装置和计算机可读存储介质 |
CN109829629A (zh) * | 2019-01-07 | 2019-05-31 | 平安科技(深圳)有限公司 | 风险分析报告的生成方法、装置、计算机设备和存储介质 |
CN110659799A (zh) * | 2019-08-14 | 2020-01-07 | 深圳壹账通智能科技有限公司 | 基于关系网络的属性信息处理方法、装置、计算机设备和存储介质 |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US9251470B2 (en) * | 2014-05-30 | 2016-02-02 | Linkedin Corporation | Inferred identity |
CN108287864B (zh) * | 2017-12-06 | 2020-07-10 | 深圳市腾讯计算机系统有限公司 | 一种兴趣群组划分方法、装置、介质及计算设备 |
-
2019
- 2019-08-14 CN CN201910748766.0A patent/CN110659799A/zh active Pending
-
2020
- 2020-04-23 WO PCT/CN2020/086397 patent/WO2021027317A1/zh active Application Filing
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107239882A (zh) * | 2017-05-10 | 2017-10-10 | 平安科技(深圳)有限公司 | 风险评估方法、装置、计算机设备及存储介质 |
CN108132998A (zh) * | 2017-12-21 | 2018-06-08 | 浪潮软件集团有限公司 | 一种人员关系分析方法和系统 |
CN109272396A (zh) * | 2018-08-20 | 2019-01-25 | 平安科技(深圳)有限公司 | 客户风险预警方法、装置、计算机设备和介质 |
CN109345158A (zh) * | 2018-12-19 | 2019-02-15 | 重庆百行智能数据科技研究院有限公司 | 企业风险识别方法、装置和计算机可读存储介质 |
CN109829629A (zh) * | 2019-01-07 | 2019-05-31 | 平安科技(深圳)有限公司 | 风险分析报告的生成方法、装置、计算机设备和存储介质 |
CN110659799A (zh) * | 2019-08-14 | 2020-01-07 | 深圳壹账通智能科技有限公司 | 基于关系网络的属性信息处理方法、装置、计算机设备和存储介质 |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112926993A (zh) * | 2021-04-13 | 2021-06-08 | 郭栋 | 基于区块链安全大数据的信息生成方法及区块链服务系统 |
Also Published As
Publication number | Publication date |
---|---|
CN110659799A (zh) | 2020-01-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021027317A1 (zh) | 基于关系网络的属性信息处理方法、装置、计算机设备和存储介质 | |
CN110659395B (zh) | 关系网络图谱构建方法、装置、计算机设备和存储介质 | |
CN109272396B (zh) | 客户风险预警方法、装置、计算机设备和介质 | |
US10936952B2 (en) | Detecting content items in violation of an online system policy using templates based on semantic vectors representing content items | |
WO2021012790A1 (zh) | 页面数据生成方法、装置、计算机设备及存储介质 | |
US10701010B2 (en) | Claiming conversations between users and non-users of a social networking system | |
CN108769026B (zh) | 用户账号检测系统和方法 | |
CN109740620A (zh) | 人群画像分类模型的建立方法、装置、设备和存储介质 | |
CN110659396B (zh) | 缺失属性信息补全方法、装置、计算机设备和存储介质 | |
US20190073592A1 (en) | Detecting content items in violation of an online system policy using semantic vectors | |
US10460174B2 (en) | System and methods for analysis of user-associated images to generate non-user generated labels and utilization of the generated labels | |
CN111192153B (zh) | 人群关系网络构建方法、装置、计算机设备和存储介质 | |
CN110555164B (zh) | 群体兴趣标签的生成方法、装置、计算机设备和存储介质 | |
CN110245714B (zh) | 图像识别方法、装置及电子设备 | |
CN109376237A (zh) | 客户稳定性的预测方法、装置、计算机设备和存储介质 | |
CN108334625B (zh) | 用户信息的处理方法、装置、计算机设备和存储介质 | |
CN109063984B (zh) | 风险旅客方法、装置、计算机设备和存储介质 | |
CN110135943B (zh) | 产品推荐方法、装置、计算机设备和存储介质 | |
CN112131277A (zh) | 基于大数据的医疗数据异常分析方法、装置和计算机设备 | |
CN110647676B (zh) | 基于大数据的兴趣属性挖掘方法、装置和计算机设备 | |
CN112417315A (zh) | 基于网站注册的用户画像生成方法、装置、设备和介质 | |
CN110162689B (zh) | 信息推送方法、装置、计算机设备和存储介质 | |
CN116805039B (zh) | 特征筛选方法、装置、计算机设备和数据扰动方法 | |
CN110598124B (zh) | 数值属性挖掘方法、装置、计算机设备和存储介质 | |
CN111316259A (zh) | 用于反馈和裁定的语义属性的动态合成和瞬时聚簇的系统和方法 |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20851709 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20851709 Country of ref document: EP Kind code of ref document: A1 |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 01/08/2022) |
|
122 | Ep: pct application non-entry in european phase |
Ref document number: 20851709 Country of ref document: EP Kind code of ref document: A1 |