WO2020253358A1 - 业务数据的风控分析处理方法、装置和计算机设备 - Google Patents

业务数据的风控分析处理方法、装置和计算机设备 Download PDF

Info

Publication number
WO2020253358A1
WO2020253358A1 PCT/CN2020/086088 CN2020086088W WO2020253358A1 WO 2020253358 A1 WO2020253358 A1 WO 2020253358A1 CN 2020086088 W CN2020086088 W CN 2020086088W WO 2020253358 A1 WO2020253358 A1 WO 2020253358A1
Authority
WO
WIPO (PCT)
Prior art keywords
data
relationship
business
risk control
risk
Prior art date
Application number
PCT/CN2020/086088
Other languages
English (en)
French (fr)
Inventor
李国安
Original Assignee
深圳壹账通智能科技有限公司
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by 深圳壹账通智能科技有限公司 filed Critical 深圳壹账通智能科技有限公司
Publication of WO2020253358A1 publication Critical patent/WO2020253358A1/zh

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F16/00Information retrieval; Database structures therefor; File system structures therefor
    • G06F16/30Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
    • G06F16/36Creation of semantic tools, e.g. ontology or thesauri
    • G06F16/367Ontology
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F40/00Handling natural language data
    • G06F40/20Natural language analysis
    • G06F40/279Recognition of textual entities
    • G06F40/289Phrasal analysis, e.g. finite state techniques or chunking
    • G06F40/295Named entity recognition
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q10/00Administration; Management
    • G06Q10/06Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063Operations research, analysis or management
    • G06Q10/0635Risk analysis of enterprise or organisation activities
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q10/00Administration; Management
    • G06Q10/06Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
    • G06Q10/063Operations research, analysis or management
    • G06Q10/0639Performance analysis of employees; Performance analysis of enterprise or organisation operations
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06QINFORMATION 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
    • G06Q40/00Finance; Insurance; Tax strategies; Processing of corporate or income taxes
    • G06Q40/03Credit; Loans; Processing thereof

Definitions

  • This application relates to the field of computer technology, in particular to a method, device and computer equipment for risk control analysis and processing of business data based on relational network analysis.
  • a risk control analysis and processing method for business data that can effectively improve the accuracy of risk control analysis.
  • the changed business data is acquired, and the target user identification is determined based on the changed business data And business type identification; the information in the risk control database is used to describe the corresponding relationship between business data, user identification and relationship map data; the corresponding relationship map data is obtained according to the target user identification; wherein, the relationship map
  • the data is the business data and historical association data generated by the business server processing the business request through the preset relationship analysis model, the multiple entity data and the association relationship between the multiple entity data are determined, and the association relationship is established according to the association relationship
  • the relationship map obtain a preset risk control analysis model according to the business type identification, use the risk control analysis model to determine the risk factors of the changed business data corresponding to the relationship map data, and according to the risk factors and the relationship map
  • the data calculates the risk index value, generates an analysis result according to the risk index value; generates a corresponding risk control report according to a preset method according
  • a risk control analysis and processing device for business data comprising: a data monitoring module for acquiring changed business data when it is detected that the business data in the risk control database is changed, and determining target users based on the changed business data Identification and business type identification; the information in the risk control database is used to describe the correspondence between business data, user identification and relationship graph data; a data acquisition module is used to obtain information from the risk control database according to the target user identification Corresponding relationship map data is obtained in the database; wherein, the relationship map data is business data and historical associated data generated by processing business requests by the business server through a preset relationship analysis model, and multiple entity data and multiple entities are determined The relationship between data and the relationship map established according to the relationship; the risk control analysis module is used to obtain a preset risk control analysis model according to the business type identification, and use the risk control analysis model to determine the Change business data corresponding to the risk factor of the relationship map data, calculate the risk index value according to the risk factor and the relationship map data, and generate the analysis result according to the risk index value; the risk control report generation
  • a computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the steps of the business data risk control analysis and processing method 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 risk control analysis and processing method for business data provided in any embodiment of the present application are realized.
  • This application uses the relationship graph and the risk control analysis model to perform risk control analysis on the changed business data in real time, which can more comprehensively analyze the risks of users, thereby accurately and effectively monitoring the risks in the financial business, and effectively Improve the accuracy and efficiency of risk control.
  • FIG. 1 is an application scenario diagram of a method for risk control analysis and processing of business data in an embodiment
  • FIG. 2 is a schematic flowchart of a method for analyzing and processing business data risk control in an embodiment
  • FIG. 3 is a schematic flowchart of the data relationship analysis step in an embodiment
  • Figure 4 is a schematic flow chart of the data risk control analysis step in an embodiment
  • FIG. 5 is a structural block diagram of a risk control analysis and processing device for business data in an embodiment
  • Fig. 6 is an internal structure diagram of a computer device in an embodiment.
  • the risk control analysis and processing method for business data provided in this application can be applied to the application environment shown in FIG. 1.
  • the terminal 102 communicates with the risk control server 104 through the network, and the risk control server 104 communicates with the business server 106 through the network.
  • the terminal 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablets, and portable wearable devices.
  • the business server 106 can be a server corresponding to the business system.
  • the risk control server 104 and the business server 106 can be independent A server or a server cluster composed of multiple servers.
  • a method for analyzing and processing business data is provided for risk control. Taking the method applied to the risk control server in Fig. 1 as an example, the method includes the following steps:
  • Step 202 When it is detected that the business data in the risk control database is changed, obtain the changed business data, and determine the target user identification and business type identification according to the changed business data; the information in the risk control database is used to describe the business data, user identification and Correspondence between the relationship map data.
  • the risk control server can monitor the business data of multiple individual users or enterprise users in real time to monitor the risks of individuals or enterprises in real time.
  • business data can include financial data, user data, and business process data;
  • business data can include data generated after multiple business servers process business requests sent by users, and can also include risk control servers using web crawling technology
  • Business data and historical behavior data related to users captured on the three-party platform are correspondingly stored in a preset risk control database.
  • the risk control database stores business data, relationship map data, and the correspondence between user identification and relationship map data.
  • the risk control server can obtain the business data and related data of multiple users in real time, such as transaction data and credit data, etc., and use the distributed file system to store the real-time obtained data in a preset risk control database.
  • the risk control server can be pre-configured with data of multiple indicator types, and the risk control server detects the business data in the database in real time according to the indicator type. When the data corresponding to the preset indicator type in the risk control database changes, it indicates the business data Substantial changes have taken place. Therefore, important data that has changed can be detected effectively in time.
  • Step 204 according to the target user identification from the corresponding relationship map data from the risk control database; wherein, the relationship map data is the business data and historical correlation data generated by the business server processing the business request based on the preset relationship analysis model. Multiple entity data and the relationship between multiple entity data, and a relationship map established based on the relationship.
  • the relationship graph can be a knowledge graph.
  • the knowledge graph refers to combining the theories and methods of applied mathematics, graphics, information visualization technology, information science and other disciplines with metrological citation analysis, co-occurrence analysis and other methods to combine complex knowledge
  • the field is displayed through data mining, information processing, knowledge measurement and graph drawing to reveal the dynamic development law of the knowledge field.
  • the relationship graph data may be business data and historical associated data generated by the risk control server through a preset relationship analysis model based on the business server processing business requests, and determine multiple entity data and the association relationship between multiple entity data. And based on the relationship map established by the association relationship. For example, it may include relationship maps corresponding to multiple types of financial services.
  • the risk control server can obtain business data and historical associated data, and further obtain the preset relationship analysis model, analyze the entity data contained in the business data and historical associated data through the relationship analysis model, and analyze the entity data and corresponding data attributes and data.
  • Keyword tags analyze the relationship between multiple entity data according to business type, data attributes and keyword tags, and establish a relationship map based on the relationship between multiple entity data, and store the relationship map and the corresponding relationship map data To the risk control database.
  • Step 206 Obtain a preset risk control analysis model according to the business type identification, use the risk control analysis model to determine the risk factor of the map data corresponding to the changed business data, calculate the risk index value according to the risk factor and the relationship map data, and generate it according to the risk index value Analyze the results.
  • Step 208 Generate a corresponding risk control report in a preset manner according to the analysis result, and send the risk control report to the monitoring terminal.
  • the risk control server When the risk control server detects that there is a change in the business data in the risk control database, it obtains the changed business data and the corresponding relationship graph data according to the user identification, and further obtains the preset risk control analysis model, through the preset in the risk control analysis model
  • the indicator analyzes the change data and the relationship graph data.
  • the risk control analysis model may be a model pre-built by the risk control server based on analyzing and training a large amount of business data and relational graph data.
  • the risk control server inputs business data and relationship map data into the risk control analysis model, performs cluster analysis and feature extraction on the business data and relationship map data, extracts feature vectors that meet the preset threshold, and extracts the relationship The characteristics of the map elements in the map data.
  • Use the risk control analysis model to determine the risk factors of the map data corresponding to the changed business data, calculate the risk index value according to the risk factor and the relationship map data, generate the analysis result according to the risk index value, and add to the analysis result according to multiple risk control index values The corresponding risk label.
  • the risk control server further generates a corresponding risk control report in a preset manner according to the analysis result, and sends the risk control report to the monitoring terminal.
  • the relationship graph and the risk control analysis model to perform risk control analysis on the changed business data in real time, it is possible to accurately and effectively monitor the risks of multiple users, thereby effectively improving the efficiency of risk control.
  • the risk control server detects the data in the risk control database in real time. When it detects that the business data in the risk control database changes, it determines the target user identification and business type identification based on the changed business data.
  • the information in the control database is used to describe the correspondence between business data, user IDs, and relationship map data; the risk control server obtains the corresponding relationship map data from the risk control database according to the target user ID; the relationship map data is preset
  • the relationship analysis model is based on the business data and historical association data generated by the business server processing the business request, determines the multiple entity data and the relationship between multiple entity data, and establishes a relationship graph based on the relationship, which can Effectively construct a relationship graph associated with user identification.
  • the risk control server then obtains the preset risk control analysis model according to the business type identification, uses the risk control analysis model to determine the risk factor of the map data corresponding to the changed business data, calculates the risk index value according to the risk factor and the relationship map data, and calculates the risk index value according to the risk index value Generate analysis results, generate corresponding risk control reports in a preset manner according to the analysis results, and send the risk control reports to the monitoring terminal, so that the monitoring terminal performs risk decisions and other processing for the user based on the risk control report.
  • the step of establishing the relationship map data specifically includes the following content:
  • Step 302 Obtain the business data generated by the business server processing the business request, and store the business data and the user identifier carried in the business data in the risk control database.
  • the risk control server can monitor the business data of multiple individual users or enterprise users in real time to monitor the risks of individuals or enterprises in real time.
  • business data can include financial data, user data, and business process data;
  • business data can include data generated after multiple business servers process business requests sent by users, and can also include risk control servers using web crawling technology
  • Business data and historical behavior data related to users captured on the three-party platform are correspondingly stored in a preset risk control database.
  • Step 304 Obtain corresponding historical association data according to the user identification, and determine multiple entity data and association relationships between multiple entity data based on the business data and historical association data through a preset relationship analysis model; business data and history The associated data all carry the corresponding service type identifier.
  • the risk control server may further obtain corresponding business data and historical associated data according to the user identification, and both the business data and historical associated data carry the corresponding business type identification.
  • the business type identifier is used to describe the business type corresponding to the business data to distinguish the business data generated by different types of business.
  • historical related data may include the user's historical behavior data, business related data, and other related data, etc.
  • the source of the historical related data may include historical business data processed by multiple service servers on the business request sent by the user, or may include wind
  • the control server uses web crawler technology to grab business data and historical behavior data associated with users from third-party platforms.
  • the entity can be a specific person, thing, or thing, or it can be an abstract concept or connection, which means things that exist objectively and can be distinguished from each other; entity data means data in the database used to describe objects or concepts in the real world.
  • the relationship graph can be a knowledge graph.
  • the knowledge graph refers to combining the theories and methods of applied mathematics, graphics, information visualization technology, information science and other disciplines with metrological citation analysis, co-occurrence analysis and other methods to combine complex knowledge
  • the field is displayed through data mining, information processing, knowledge measurement and graph drawing to reveal the dynamic development law of the knowledge field.
  • the risk control server can build multiple relationship maps in advance. For example, it may include relationship maps corresponding to multiple types of financial services.
  • the relationship analysis model can be that the risk control server performs feature extraction on a large amount of historical business data and related data in advance, analyzes the data features and relationship features between these data, and performs data features and relationship features between the analyzed historical data. Continue to learn and train, analyze the connections between data features, and train to obtain a relationship analysis model.
  • the relationship analysis model can be a neural network model based on deep learning. Analyze the entity data contained in the business data and historical associated data through the relationship analysis model, and analyze the entity data and the corresponding data attributes and keyword tags.
  • the risk control server may first perform feature extraction on business data and historical associated data.
  • a clustering algorithm may be used to perform cluster analysis on business data and historical associated data, and extract feature variables that meet preset thresholds, and pass
  • the entity data in the business data and historical related data is analyzed, and the entity data and the corresponding data attributes and keyword tags are analyzed, and according to the business type, data attributes and keyword tags Analyze the relationship between multiple entity data.
  • the relationship analysis model deploys related data feature analysis model algorithms, which can effectively analyze multiple entity data and multiple entity data. The relationship between.
  • Step 306 Establish a relationship map according to the association relationship, and store the relationship map and the corresponding relationship map data in the risk control database.
  • the risk control server generates corresponding multiple data nodes based on multiple entity data; generates description information of multiple data nodes based on the business type, data attributes and keyword tags of the business data and historical associated data; based on the description information and multiple
  • the relationship characteristics between entity data determine the mapping relationship and relationship type of multiple data nodes, and link multiple data nodes according to the mapping relationship and relationship type; generate corresponding relationship graphs based on multiple data nodes and association relationships, and
  • the constructed relationship map and the corresponding relationship map data are stored in the risk control database by means of a library. This can effectively establish a relationship graph between one or more users.
  • the step of determining the association relationship between multiple entity data specifically includes the following content: input business data and historical association data into the relationship analysis model, and perform feature extraction on the business data and historical association data, Obtain feature vectors of multiple entity data.
  • the feature vector is analyzed through the relationship analysis model, and the data attributes and keyword tags of the entity data are obtained. Analyze the relationship characteristics between multiple entity data according to the business type, data attributes and keyword tags of the entity data, and determine the association relationship between multiple entity data according to the relationship characteristics.
  • the risk control server can build multiple relationship maps in advance. For example, it may include relationship maps corresponding to multiple types of financial services. After the risk control server obtains the business data and historical associated data, it further obtains a preset relationship analysis model.
  • the relationship analysis model may be a neural network model based on deep learning. Analyze the entity data contained in the business data and historical associated data through the relationship analysis model, and analyze the entity data and the corresponding data attributes and keyword tags.
  • the risk control server may first perform feature extraction on business data and historical associated data.
  • a clustering algorithm may be used to perform cluster analysis on business data and historical associated data, and extract feature variables that meet preset thresholds, and pass
  • the entity data in the business data and historical related data is analyzed, and the entity data and the corresponding data attributes and keyword tags are analyzed, and the relationship extraction technology is used according to the business type and data attributes
  • the keyword tag extracts the relationship between entities from the business data and historical association data, extracts the relationship characteristics between multiple entity data, and then determines the multiple entity data based on the relationship characteristics between multiple entity data The relationship between.
  • the risk control server establishes a relationship map according to the association relationship between multiple entity data, and uses a library to store the constructed relationship map and the corresponding relationship map data in the risk control database.
  • the core subject of credit is the loan applicant, and the loan applicant may be an individual or a company.
  • the loan risk is evaluated based on the applicant’s basic information, behavioral information, business status, and social relations.
  • the credit-related entities can be: people, enterprises, bank accounts, banks, mortgages, application events, litigation events, etc., as well as basic information entities such as phone calls, emails, and addresses.
  • the step of establishing a relationship graph based on the association relationship includes: generating multiple corresponding data nodes based on multiple entity data; generating multiple data nodes based on the business type, data attributes, and keyword tags of the business data and historical associated data.
  • the description information of each data node determine the mapping relationship and relationship type of multiple data nodes according to the relationship characteristics between the description information and multiple entity data, and link multiple data nodes according to the mapping relationship and relationship type;
  • the data node and the association relationship generate a corresponding relationship graph.
  • the risk control server After the risk control server obtains the business data and historical associated data, it obtains the preset relationship analysis model, analyzes the entity data contained in the business data and historical associated data through the relationship analysis model, and analyzes the entity data and the corresponding data attributes and keys Word tags, and then analyze the relationship between multiple entity data based on business types, data attributes and keyword tags.
  • the risk control server analyzes the entity data contained in the business data and historical associated data through the relationship analysis model, it generates corresponding multiple data nodes based on the entity data, and based on the business type and data attributes of the business data and historical associated data And the keyword tag generates the description information of the data node.
  • the mapping relationship of multiple data nodes is generated according to the relationship characteristics and description information between multiple entity data, where the mapping relationship includes a corresponding relationship type.
  • the risk control server links multiple data nodes according to the mapping relationship between the multiple data nodes and the corresponding relationship type, and generates a corresponding relationship map in a preset manner according to the multiple data nodes and description information after the link.
  • an ontology model namely a relationship analysis model
  • the risk control server can use the preset relationship map to form a corresponding map with the relationship between nodes and entity data.
  • the risk control server can map and merge data from different sources with nodes as the main target, and describe different data attributes The node that the data corresponds to, and use the relationship to describe the relationship between the data of each node.
  • the risk control server can associate and store various types of data of nodes through the node link technology, and use the graph database to connect the originally unconnected data, integrate the discrete data, and form a network knowledge structure through the relevance of different knowledge , In order to effectively establish a relationship map to provide more valuable decision support.
  • the step of detecting that the business data in the risk control database has changed includes: obtaining updated data in the risk control database; obtaining the business type of the updated data, and comparing the business type of the updated data with the preset index type ; When there is update data of the preset index type, it means that the business data in the risk control database has changed.
  • the risk control server can monitor whether the business data and associated data in the risk control database have changed in real time.
  • the risk control server can obtain the updated data in the risk control database in real time, where the updated data can include new and modified business data, etc. And obtain the business type of the updated data, and detect the business type of the updated data and the preset indicator type. When the updated data of the preset indicator type is detected, it means that there is a substantial change in the business data in the risk control database.
  • the risk control server can obtain business data and associated data generated by multiple users in real time, such as transaction data and credit information data, etc., and use the distributed file system to store the data obtained in real time in a preset risk control database .
  • the risk control server can be pre-configured with data of multiple indicator types, and the risk control server detects the business data in the database in real time according to the indicator type.
  • the data corresponding to the preset indicator type in the risk control database changes, it indicates the business data Substantial changes have taken place. For example, you can use tools to monitor changes in data such as additions and modifications in the database, and synchronize them to the big data platform to see whether the calculated data of the preset indicator type has substantially changed.
  • the integration of kafka and spark can be used to obtain the changed data in the database in time, so that important data that has changed can be detected in a timely and effective manner.
  • the method before obtaining the preset risk control analysis model according to the business type identification, the method further includes: obtaining multiple business data, using the multiple business data to generate training set data and validation set data; Perform cluster analysis on the training set data, extract feature vectors that reach the preset threshold according to the clustering results; input the feature vectors into the preset neural network model for training, and obtain the initial risk control analysis model; input the validation set data into the initial Training and verification are performed in the risk control analysis model until the verification pass rate of the validation set data meets the preset threshold, then the training is stopped to obtain the required risk control analysis model.
  • the risk control server can obtain a large amount of business data and associated data in advance, and use multiple business data to generate training set data and verification set data.
  • the training set data may be a data set that has been manually labeled.
  • the risk control server further performs big data analysis on the training set data. For example, it can perform cluster analysis and feature extraction on the training set data through a clustering algorithm, extract the feature vector of the training set data according to the clustering result, and combine the feature vector Input to the preset neural network model for training, calculate the importance of each feature vector through the preset analysis model, and obtain the initial risk control analysis model.
  • the preset neural network model can be a neural network model based on a decision tree or a neural network model based on deep learning.
  • the verification set data is further used to verify the risk control analysis model.
  • the verification pass rate in the verification set data meets the preset threshold, the corresponding category probability is in the preset range
  • the number inside reaches the preset data it means that the risk control analysis model has been successfully trained, and the training is stopped to obtain the required risk control analysis model.
  • the risk control analysis model is trained using a large amount of data, it analyzes the inherent inevitable connections of these data and the characteristics between the data, and then trains the required risk control analysis based on the inherent inevitable connections of these data and the characteristics between the data Model, through the training of the risk control analysis model to analyze the business data and relationship map data, so as to effectively obtain the corresponding analysis result data.
  • the risk management server can map a large amount of acquired business data according to the map data of the relationship map and the corresponding relationship, and then perform feature extraction on a large amount of business data to obtain the corresponding feature vector and the corresponding feature value, and combine all
  • the feature combination is a two-dimensional vector to form the training data set required by the machine learning model.
  • the server further trains the training data to obtain the initial risk control analysis model.
  • the server can further optimize the parameters of the risk control analysis model using the updated data according to the preset frequency, and update the model after obtaining the optimal parameter model, thereby effectively obtaining the trained risk control analysis model.
  • the step of using a risk control analysis model to analyze the change data and the relationship map data to obtain the analysis result specifically includes the following content:
  • Step 402 Input the business data and the relationship map data into the risk control analysis model, and extract the feature vector corresponding to the business data and the map element characteristics in the relationship map data.
  • Step 404 Determine a risk factor for changing the map data corresponding to the business data according to the characteristic variables and the characteristics of the map elements.
  • Step 406 Use the risk control analysis model to calculate multiple risk index values according to the risk factors and the characteristics of the map elements.
  • Step 408 Generate analysis results according to multiple risk index values, and add corresponding risk level labels to the analysis results.
  • the risk control server obtains business data and establishes a relationship map, and stores the business data and relationship map data in the preset risk control database, and then detects the data in the risk control database.
  • the business data in the risk control database is detected
  • the feature vector refers to the vector used to represent the data feature after linear transformation and feature extraction are performed on the business data and the relationship map data.
  • the map element features include entity data features and relationship features.
  • the entity data features and relationship features represent the feature values corresponding to the entity data and the associated relationship data, and the feature values are used to represent the data features and relationship features.
  • the risk control server inputs business data and relationship map data into the risk control analysis model.
  • the risk control analysis model extracts features from the business data and relationship map data, extracts the feature vector corresponding to the business data, and extracts the relationship map
  • the graph element characteristics in the data the risk control server conducts risk analysis on the feature vector, the entity data feature and the relationship feature through the risk control analysis model, and determines the risk factor for changing the map data corresponding to the business data according to the feature variable and the graph element feature.
  • the risk factor can represent the impact of changing business data on the relationship graph data.
  • the monitoring server uses the risk control analysis model to calculate multiple risk indicator values based on the risk factors and map element characteristics, uses multiple risk indicator values to generate corresponding analysis results in a preset manner, and adds the corresponding risk indicator values to the analysis results Risk control label. Specifically, when there is index data in the analysis result that does not reach the preset index, it indicates that there is a risk, and a risk label is added to the analysis result. When there is a risk tag in the analysis result, the corresponding risk control report is generated according to the preset method and sent to the corresponding monitoring terminal, and an early warning is given.
  • the risk control analysis model is used to conduct risk control analysis on the changed data in real time.
  • the model has been obtained by learning and training a large amount of business data and view data, so that the risk control analysis model can be used to analyze and make decisions on the changed business data and corresponding entity data, so that the risk control analysis model can be based on data characteristics
  • the inherent inevitable connection can analyze the degree of risk corresponding to the updated business data, and then can accurately and effectively monitor the risks in the financial business, and effectively improve the efficiency of risk control.
  • the risk control server can also analyze specific enterprises and monitor the capabilities of small and medium-sized enterprises.
  • the monitoring terminal may send a risk control request to the risk control server, and the risk control request includes the user identification.
  • the risk control server receives the risk control request sent by the monitoring terminal, it obtains the business data and related data in the corresponding financial field according to the user ID, obtains the preset relationship analysis model, and analyzes the business data and related data through the relationship analysis model. Multiple entity data and the relationship between multiple entity data; establish a relationship graph based on the relationship between multiple entity data.
  • the risk control server further performs risk control analysis on the business data and relationship map data identified by the user through the preset risk control analysis model, and calculates multiple risk index values corresponding to the business data and relationship map data according to the preset indicators, and uses Multiple risk index values generate corresponding analysis results in a preset manner.
  • the risk control server may further generate a user risk control portrait corresponding to the user identifier in a preset manner based on the analysis result and the relationship map data.
  • loan risk indicators can include non-performing loan ratio, loan weighted risk, loan diversification ratio, non-performing loan provision coverage ratio, etc.
  • the potential risks of the company can be accurately and effectively monitored.
  • a risk control analysis and processing device for business data including: a data monitoring module 502, a data acquisition module 504, a risk control analysis module 506, and a risk control report generation module 508, wherein :
  • the data monitoring module 502 is used to obtain the changed business data when it is detected that the business data in the risk control database changes, and determine the target user identification and business type identification according to the changed business data; the information in the risk control database is used to describe the business data , The corresponding relationship between the user ID and the relationship map data;
  • the data acquisition module 504 is configured to acquire corresponding relationship map data from the risk control database according to the target user identifier; wherein the relationship map data is business data and history generated by processing business requests based on the business server through a preset relationship analysis model Associated data, determine multiple entity data and the relationship between multiple entity data, and establish a relationship map based on the relationship;
  • the risk control analysis module 506 is used to obtain a preset risk control analysis model according to the business type identification, use the risk control analysis model to determine the risk factor of the map data corresponding to the changed business data, and calculate the risk index value according to the risk factor and the relationship map data, Generate analysis results based on risk index values;
  • the risk control report generation module 508 is configured to generate a corresponding risk control report in a preset manner according to the analysis result, and send the risk control report to the monitoring terminal.
  • the device further includes a relational data analysis module for obtaining business data generated by the business server processing the business request, and correspondingly storing the business data and the user identification carried by the business data in the risk control database;
  • the user ID obtains the corresponding historical association data, and determines multiple entity data and the association relationship between multiple entity data based on the business data and historical association data through a preset relationship analysis model; both business data and historical association data are carried Corresponding business type identification; establish a relationship map according to the association relationship, and store the relationship map and the corresponding relationship map data to the risk control database.
  • the relationship data analysis module is also used to input business data and historical associated data into the relationship analysis model, and perform feature extraction on the business data and historical associated data to obtain feature vectors of multiple entity data;
  • the analysis model analyzes the feature vector to obtain the data attributes and keyword tags of the entity data; analyzes the relationship characteristics between multiple entity data according to the business type, data attributes and keyword tags of the entity data, and determines multiple entities based on the relationship characteristics The relationship between entity data.
  • the relational data analysis module is also used to generate multiple corresponding data nodes based on multiple entity data; generate multiple data nodes based on the business type, data attributes, and keyword tags of the business data and historical associated data Descriptive information; determine the mapping relationship and relationship type of multiple data nodes according to the relationship characteristics between the description information and multiple entity data, and link multiple data nodes according to the mapping relationship and relationship type; according to the multiple data after linking Nodes and description information generate corresponding relationship graphs.
  • the data monitoring module 502 is also used to obtain updated data in the risk control database; obtain the business type of the updated data, and compare the business type of the updated data with the preset indicator type; when there is a preset indicator type When the data is updated, it means that the business data in the risk control database has changed.
  • the device further includes a model training module for acquiring multiple business data, using multiple business data to generate training set data and verification set data; clustering the training set data through a clustering algorithm, Extract the feature vector that reaches the preset threshold according to the clustering result; input the feature vector into the preset neural network model for training to obtain the initial risk control analysis model; input the validation set data into the initial risk control analysis model for training and Validation, until the validation pass rate of the validation set data meets the preset threshold, stop training and obtain the required risk control analysis model.
  • a model training module for acquiring multiple business data, using multiple business data to generate training set data and verification set data; clustering the training set data through a clustering algorithm, Extract the feature vector that reaches the preset threshold according to the clustering result; input the feature vector into the preset neural network model for training to obtain the initial risk control analysis model; input the validation set data into the initial risk control analysis model for training and Validation, until the validation pass rate of the validation set data meets the preset threshold, stop training and obtain the required risk control analysis model.
  • the risk control analysis module 506 is also used to input the business data and the relationship map data into the risk control analysis model, and extract the feature vector corresponding to the changed business data and the map element characteristics in the relationship map data; Determine the risk factors of the map data corresponding to the changed business data according to the characteristic variables and the characteristics of the map elements; use the risk control analysis model to calculate multiple risk indicator values based on the risk factors and map element characteristics; generate the analysis results based on the multiple risk indicator values, and Add the corresponding risk level label to the analysis result.
  • Each module in the above-mentioned risk control analysis and processing device for business data can 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 business data, historical related data, relational graphs and relational graph data, etc.
  • 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 fewer parts than shown in the figure, or combining some parts, or having a different arrangement of parts.
  • a computer-readable storage medium may be non-volatile or volatile, and has a computer program stored thereon.
  • the computer program is executed by a processor, The steps of the risk control analysis and processing method for business data provided in any embodiment of the present application are implemented.
  • 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.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Theoretical Computer Science (AREA)
  • Strategic Management (AREA)
  • Physics & Mathematics (AREA)
  • Economics (AREA)
  • General Physics & Mathematics (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Educational Administration (AREA)
  • Marketing (AREA)
  • General Business, Economics & Management (AREA)
  • Accounting & Taxation (AREA)
  • General Engineering & Computer Science (AREA)
  • Quality & Reliability (AREA)
  • Operations Research (AREA)
  • Finance (AREA)
  • Tourism & Hospitality (AREA)
  • Computational Linguistics (AREA)
  • Game Theory and Decision Science (AREA)
  • Animal Behavior & Ethology (AREA)
  • Databases & Information Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Technology Law (AREA)
  • Health & Medical Sciences (AREA)
  • Artificial Intelligence (AREA)
  • Audiology, Speech & Language Pathology (AREA)
  • General Health & Medical Sciences (AREA)
  • Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

一种业务数据的风控分析处理方法,包括:当风控服务器检测到风控数据库中的业务数据发生变更时,获取变更业务数据,根据变更业务数据确定目标用户标识和业务类型标识,根据目标用户标识从风控数据库中对应的预先根据业务数据和历史关联数据确定出多个实体数据以及多个实体数据之间的关联关系,并根据关联关系建立的关系图谱数据。根据业务类型标识获取风控分析模型,利用风控分析模型确定变更业务数据对应关系图谱数据的风险因子并计算风险指标值,根据风险指标值生成分析结果,根据分析结果生成对应的风控报告并发送至监控终端。采用本方法能够准确有效地对业务数据进行全面的风控分析。

Description

业务数据的风控分析处理方法、装置和计算机设备
本申请要求于2019年6月18日提交中国专利局,申请号为201910526011.6、发明名称为“业务数据的风控分析处理方法、装置和计算机设备”的中国专利申请的优先权,其全部内容通过引用结合在本申请中。
技术领域
本申请涉及计算机技术领域,特别是涉及基于关系网络分析的一种业务数据的风控分析处理方法、装置和计算机设备。
背景技术
随着计算机技术的不断发展,互联网金融也随之迅速发展,目前互联网金融已经渗透到人们的衣食住行等方方面面,包含支付、理财、众筹、消费等功能的各类互联网金融产品和平台层出不穷。然而互联网金融存在一定的风险,因此需要建立良好的风险管理体系,以对互联网金融交易中的风险进行监控。发明人发现,现有的风控分析方式中,大多是利用风险评分模型对一些数据进行评分分析,指标计算通常是基于离线数据计算,计算数据为某个时间点之前的数据,不包含新增数据,对风险分析的数据比较有限,且分析方式较简单,风控分析结果的准确率较低,风控报告数据不完整的问题。因此,如何有效提高风控分析的准确性成为目前需要解决的技术问题。
发明内容
基于此,有必要针对上述技术问题,提供一种能够有效提高风控分析的准确性的业务数据的风控分析处理方法、装置和计算机设备。
一种能够有效提高风控分析的准确性的业务数据的风控分析处理方法,当检测到风控数据库中的业务数据发生变更时,获取变更业务数据,根据所述变更业务数据确定目标用户标识和业务类型标识;所述风控数据库中的信息用于描述业务数据、用户标识以及关系图谱数据之间的对应关系;根据所述目标用户标识获取对应的关系图谱数据;其中,所述关系图谱数据为通过预设的关系分析模型基于业务服务器对业务请求进行处理产生的业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系,并根据所述关联关系建立的关系图谱;根据所述业务类型标识获取预设的风控分析模型,利用所述风控分析模型确定所述变更业务数据对应所述关系图谱数据的风险因子,根据所述风险因子和关系图谱数据计算风险指标值,根据所述风险指标值生成分析结果;根据所述分析结果按照预设方式生成对应的风控报告,并将所述风控报告发送至监控终端。
一种业务数据的风控分析处理装置,所述装置包括:数据监测模块,用于当检测到风控数据库中的业务数据发生变更时,获取变更业务数据,根据所述变更业务数据确定目标用户标识和业务类型标识;所述风控数据库中的信息用于描述业务数据、用户标识以及关系图谱数据之间的对应关系;数据获取模块,用于根据所述目标用户标识从所述风控数据库中获取对应的关系图谱数据;其中,所述关系图谱数据为通过预设的关系分析模型基于业务服务器对业务请求进行处理产生的业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系,并根据所述关联关系建立的关系图谱;风控分析模块,用于根据所述业务类型标识获取预设的风控分析模型,利用所述风控分析模型确定所述变更业务数据对应所述关系图谱数据的风险因子,根据所述风险因子和关系图谱数据计算风险指标值,根据所述风险指标值生成分析结果;风控报告生成模块,用于根据所述分析结果按照预设方式生成对应的风控报告,并将所述风控报告发送至监控终端。
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,所述处理器执行所述计算机程序时实现本申请任意一个实施例中提供的业务数据的风控分析处理方法的步骤。
一种计算机可读存储介质,其上存储有计算机程序,所述计算机程序被处理器执行时实现本申请任意一个实施例中提供的业务数据的风控分析处理方法的步骤。
本申请通过利用关系图谱和风控分析模型实时对变更的业务数据进行风控分析,能够更全面地对用户存在的风险进行分析,由此能够准确有效地监控金融业务中存在的风险,进而有效地提高了风控的准确性和效率。
附图说明
图1为一个实施例中业务数据的风控分析处理方法的应用场景图;
图2为一个实施例中业务数据的风控分析处理方法的流程示意图;
图3为一个实施例中数据关系分析步骤的流程示意图;
图4为一个实施例中数据风控分析步骤的流程示意图;
图5为一个实施例中业务数据的风控分析处理装置的结构框图;
图6为一个实施例中计算机设备的内部结构图。
具体实施方式
为了使本申请的目的、技术方案及优点更加清楚明白,以下结合附图及实施例,对本申请进行进一步详细说明。应当理解,此处描述的具体实施例仅仅用以解释本申请,并不用于限定本申请。
本申请提供的业务数据的风控分析处理方法,可以应用于如图1所示的应用环境中。其中,终端102通过网络与风控服务器104进行通信,风控服务器104通过网络与业务服务器106进行通信。其中,终端102可以但不限于是各种个人计算机、笔记本电脑、智能手机、平板电脑和便携式可穿戴设备,业务服务器106可以是业务系统对应的服务器,风控服务器104和业务服务器106可以用独立的服务器或者是多个服务器组成的服务器集群来实现。
在一个实施例中,如图2所示,提供了一种业务数据的风控分析处理方法,以该方法应用于图1中的风控服务器为例进行说明,包括以下步骤:
步骤202,当检测到风控数据库中的业务数据发送变更时,获取变更业务数据,根据变更业务数据确定目标用户标识和业务类型标识;风控数据库中的信息用于描述业务数据、用户标识以及关系图谱数据之间的对应关系。
风控服务器可以实时对多个个人用户或企业用户的业务数据进行监控,以实时对个人或企业进行风险监控。其中,业务数据可以包括金融数据、用户数据以及业务流程数据等;业务数据可以包括多个业务服务器对用户发送的业务请求进行处理后产生的数据,也可以包括风控服务器利用网络爬虫技术从第三方平台中抓取的用户所相关联的业务数据和历史行为数据等。具体地,风控服务器获取业务服务器对业务请求进行处理产生的业务数据后,则将业务数据以及业务数据携带的用户标识对应存储至预设的风控数据库中。风控数据库中存储了业务数据、关系图谱数据以及用户标识与关系图谱数据之间的对应关系。
风控服务器可以实时获取多个用户发生的业务数据以及关联数据等,例如交易数据和征信数据等,并将实时获取的数据利用分布式文件系统存储至预设的风控数据库中。风控服务器可以预先配置多个指标类型的数据,风控服务器则根据指标类型对数据库中的业务数据进行实时检测,当风控数据库中存在预设指标类型对应的数据发生变更时,表示业务数据发生实质性变化。由此可以及时有效地检测到发生变化的重要数据。
步骤204,根据目标用户标识从风控数据库中对应的关系图谱数据;其中,关系图谱数据为通过预设的关系分析模型基于业务服务器对业务请求进行处理产生的业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系,并根据关联关系建立的关系图谱。
当检测到风控数据库中的业务数据发送变更时,获取变更业务数据,根据变更业务数据确定目标用户标识和业务类型标识,进而根据目标用户标识从风控数据库中对应的关系图谱数据。其中,关系图谱可以是知识图谱,知识图谱是指通过将应用数学、图形学、信息可视化技术、信息科学等学科的理论与方法与计量学引文分析、共现分析等方法结合,把复杂的知识领域通过数据挖掘、信息处理、知识计量和图形绘制而显示出来,以揭示知识领域的动态发展规律。关系图谱数据可以是风控服务器预先通过预设的关系分析模型基于业务服务器 对业务请求进行处理产生的业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系,并根据关联关系建立的关系图谱。例如,可以包括多个金融业务类型对应的关系图谱。
风控服务器可以获取业务数据和历史关联数据,进一步获取预设的关系分析模型,通过关系分析模型分析出业务数据和历史关联数据中包含的实体数据,并分析出实体数据以及对应的数据属性和关键字标签,根据业务类型、数据属性和关键字标签分析多个实体数据之间的关联关系,并根据多个实体数据之间的关联关系建立关系图谱,将关系图谱和对应的关系图谱数据存储至风控数据库。
步骤206,根据业务类型标识获取预设的风控分析模型,利用风控分析模型确定变更业务数据对应关系图谱数据的风险因子,根据风险因子和关系图谱数据计算风险指标值,根据风险指标值生成分析结果。
步骤208,根据分析结果按照预设方式生成对应的风控报告,并将风控报告发送至监控终端。
风控服务器检测到风控数据库中的业务数据存在变更时,根据用户标识获取变更业务数据和对应的关系图谱数据,并进一步获取预设的风控分析模型,通过风控分析模型中的预设指标对变更数据和关系图谱数据进行分析。其中,风控分析模型可以是风控服务器预先构建的基于对大量业务数据和关系图谱数据进行分析和训练得到的模型。
具体地,风控服务器将业务数据和关系图谱数据输入至风控分析模型中,对业务数据和关系图谱数据进行聚类分析和特征提取,提取出满足预设阈值的特征向量,并提取出关系图谱数据中的图谱元素特征。利用风控分析模型确定变更业务数据对应关系图谱数据的风险因子,根据风险因子和关系图谱数据计算风险指标值,根据风险指标值生成分析结果,并根据多个风控指标值在分析结果中添加相应的风险标签。风控服务器进一步根据分析结果按照预设方式生成对应的风控报告,并将风控报告发送至监控终端。以使得监控终端根据风控报告对多个用户进行有效地风险管控。通过利用关系图谱和风控分析模型实时对变更的业务数据进行风控分析,由此能够准确有效地监控多个用户存在的风险,进而有效地提高了风控的效率。
上述业务数据分析处理方法中,风控服务器实时对风控数据库中的数据进行检测,当检测到风控数据库中的业务数据发生变更时,根据变更业务数据确定目标用户标识和业务类型标识,风控数据库中的信息用于描述业务数据、用户标识以及关系图谱数据之间的对应关系;风控服务器则根据目标用户标识从风控数据库中获取对应的关系图谱数据;关系图谱数据为通过预设的关系分析模型基于业务服务器对业务请求进行处理产生的业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系,并根据关联关系建立的 关系图谱,由此能够有效地构建出与用户标识相关联的关系图谱。风控服务器进而根据业务类型标识获取预设的风控分析模型,利用风控分析模型确定变更业务数据对应关系图谱数据的风险因子,根据风险因子和关系图谱数据计算风险指标值,根据风险指标值生成分析结果,根据分析结果按照预设方式生成对应的风控报告,并将风控报告发送至监控终端,以使得监控终端根据风控报告对该用户进行风险决策等处理。通过利用关系图谱和风控分析模型实时对变更的业务数据进行风控分析,能够更全面地对用户存在的风险进行分析,由此能够准确有效地监控金融业务中存在的风险,进而有效地提高了风险监控的准确性和风控的效率。
在一个实施例中,如图3所示,建立关系图谱数据的步骤,具体包括以下内容:
步骤302,获取业务服务器对业务请求进行处理产生的业务数据,并将业务数据以及业务数据携带的用户标识对应存储至风控数据库。
风控服务器可以实时对多个个人用户或企业用户的业务数据进行监控,以实时对个人或企业进行风险监控。其中,业务数据可以包括金融数据、用户数据以及业务流程数据等;业务数据可以包括多个业务服务器对用户发送的业务请求进行处理后产生的数据,也可以包括风控服务器利用网络爬虫技术从第三方平台中抓取的用户所相关联的业务数据和历史行为数据等。具体地,风控服务器获取业务服务器对业务请求进行处理产生的业务数据后,则将业务数据以及业务数据携带的用户标识对应存储至预设的风控数据库中。
步骤304,根据用户标识获取对应的历史关联数据,并通过预设的关系分析模型基于业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系;业务数据和历史关联数据均携带相应的业务类型标识。
风控服务器还可以进一步根据用户标识获取对应的业务数据和历史关联数据,业务数据和历史关联数据均携带相应的业务类型标识。业务类型标识用于描述业务数据对应的业务类型,以区分不同类型的业务所产生的业务数据。其中,历史关联数据可以包括用户的历史行为数据、业务关联数据以及其他关联数据等,历史关联数据的来源可以包括多个业务服务器对用户发送的业务请求进行处理的历史业务数据,也可以包括风控服务器利用网络爬虫技术从第三方平台中抓取的用户所相关联的业务数据和历史行为数据等。
其中,实体可以是具体的人、事、物,也可以是抽象的概念或联系,表示客观存在并可相互区别的事物;实体数据表示数据库中用于描述现实世界中的对象或概念对应的数据。其中,关系图谱可以是知识图谱,知识图谱是指通过将应用数学、图形学、信息可视化技术、信息科学等学科的理论与方法与计量学引文分析、共现分析等方法结合,把复杂的知识领域通过数据挖掘、信息处理、知识计量和图形绘制而显示出来,以揭示知识领域的动态发展规律。
风控服务器可以预先构建多个关系图谱。例如,可以包括多个金融业务类型对应的关系图谱。风控服务器获取业务数据和历史关联数据后,进一步获取预设的关系分析模型。关系分析模型可以是风控服务器通过预先对大量历史业务数据和关联数据进行特征提取,分析这些数据之间的数据特征和关系特征,并对分析出的历史数据之间的数据特征和关系特征进行不断学习和训练,分析数据特征之间的联系,从而训练得到关系分析模型,关系分析模型可以是基于深度学习的神经网络模型。通过关系分析模型分析出业务数据和历史关联数据中包含的实体数据,并分析出实体数据以及对应的数据属性和关键字标签。
具体地,风控服务器可以首先对业务数据和历史关联数据进行特征提取,例如可以采用聚类算法对业务数据和历史关联数据进行聚类分析,并提取出满足预设阈值的特征变量,并通过数据抽取、转换、加载等预处理后,分析出业务数据和历史关联数据中的实体数据,并分析出实体数据以及对应的数据属性和关键字标签,并根据业务类型、数据属性和关键字标签分析多个实体数据之间的关联关系。通过利用预先训练的关系分析模型对业务数据和历史关联数据进行分析,由于关系分析模型中部署了相关的数据特征分析的模型算法,从而可以有效地分析出多个实体数据和多个实体数据之间的关联关系。
步骤306,根据关联关系建立关系图谱,并将关系图谱和对应的关系图谱数据存储至风控数据库。
进一步地,风控服务器根据多个实体数据生成对应的多个数据节点;根据业务数据和历史关联数据的业务类型、数据属性和关键字标签生成多个数据节点的描述信息;根据描述信息和多个实体数据之间的关系特征确定多个数据节点的映射关系和关系类型,并根据映射关系和关系类型将多个数据节点进行链接;根据多个数据节点和关联关系生成对应的关系图谱,并采用图库的方式对构建的关系图谱以及对应的关系图谱数据存储至风控数据库中。由此能够有效地建立一个或多个用户之间的关系图谱。
在一个实施例中,确定出多个实体数据之间的关联关系的步骤,具体包括以下内容:将业务数据和历史关联数据输入至关系分析模型中,对业务数据和历史关联数据进行特征提取,得到多个实体数据的特征向量。通过关系分析模型对特征向量进行分析,得到实体数据的数据属性和关键字标签。根据实体数据的业务类型、数据属性和关键字标签分析多个实体数据之间的关系特征,根据关系特征确定出多个实体数据之间的关联关系。
风控服务器可以预先构建多个关系图谱。例如,可以包括多个金融业务类型对应的关系图谱。风控服务器获取业务数据和历史关联数据后,进一步获取预设的关系分析模型,关系分析模型可以是基于深度学习的神经网络模型。通过关系分析模型分析出业务数据和历史关联数据中包含的实体数据,并分析出实体数据以及对应的数据属性和关键字标签。
具体地,风控服务器可以首先对业务数据和历史关联数据进行特征提取,例如可以采用聚类算法对业务数据和历史关联数据进行聚类分析,并提取出满足预设阈值的特征变量,并通过数据抽取、转换、加载等预处理后,分析出业务数据和历史关联数据中的实体数据,并分析出实体数据以及对应的数据属性和关键字标签,并通过关系抽取技术根据业务类型、数据属性和关键字标签把实体间的关系从业务数据和历史关联数据中提取出来,提取出多个实体数据之间的关系特征,进而根据多个实体数据之间的关系特征确定出多个实体数据之间的关联关系。风控服务器则根据多个实体数据之间的关联关系建立关系图谱,并采用图库的方式对构建的关系图谱以及对应的关系图谱数据存储至风控数据库中。
例如,信贷业务对应的关系图谱中,信贷的核心主体是贷款申请者,贷款申请者可能是个人也可能是公司,通过申请者的基本信息、行为信息、经营状况、社会关系等评估贷款的风险。则信贷相关的实体可以为:人、企业、银行账户、银行、抵押物、申请事件、诉讼事件等,以及电话、邮件、地址等基本信息实体。
在一个实施例中,根据所述关联关系建立关系图谱的步骤包括:根据多个实体数据生成对应的多个数据节点;根据业务数据和历史关联数据的业务类型、数据属性和关键字标签生成多个数据节点的描述信息;根据描述信息和多个实体数据之间的关系特征确定多个数据节点的映射关系和关系类型,并根据映射关系和关系类型将多个数据节点进行链接;根据多个数据节点和关联关系生成对应的关系图谱。
风控服务器获取业务数据和历史关联数据后,获取预设的关系分析模型,通过关系分析模型分析出业务数据和历史关联数据中包含的实体数据,并分析出实体数据以及对应的数据属性和关键字标签,进而根据业务类型、数据属性和关键字标签分析多个实体数据之间的关联关系。
进一步地,风控服务器通过关系分析模型分析出业务数据和历史关联数据中包含的实体数据后,根据实体数据生成对应的多个数据节点,并根据业务数据和历史关联数据的业务类型、数据属性和关键字标签生成数据节点的描述信息。根据多个实体数据之间的关系特征和描述信息生成多个数据节点的映射关系,其中,映射关系包括对应的关系类型。风控服务器则根据多个数据节点之间的映射关系和对应的关系类型将多个数据节点进行链接,并根据链接后的多个数据节点和描述信息按照预设方式生成对应的关系图谱。
例如,构建风控领域的关系图谱可以先构建出本体模型,即关系分析模型,以及分析出实体数据之间的关系特征。风控服务器可以利用预设的关系图谱用节点和实体数据之间的关系组成对应的图谱,具体地,风控服务器可以节点为主体目标对不同来源的数据进行映射与合并,通过数据属性描述不同数据对于 的节点,并利用关系描述各个节点数据之间的关联关系。风控服务器可以通过节点链接技术将节点的多种类型数据进行关联存储,并利用图数据库将原本没有联系的数据连通,将离散的数据进行整合,通过不同知识的关联性形成网状的知识结构,以有效地建立关系图谱从而提供更有价值的决策支持。
在一个实施例中,检测到风控数据库中的业务数据发生变更的步骤包括:获取风控数据库中的更新数据;获取更新数据的业务类型,将更新数据的业务类型与预设指标类型进行比较;当存在预设指标类型的更新数据时,则表示风控数据库中的业务数据存在变更。
风控服务器可以实时监控风控数据库中的业务数据以及关联数据等是否发生变化。风控服务器可以实时获取风控数据库中的更新数据,其中,更新数据可以包括新增、修改的业务数据等。并获取更新数据的业务类型,将更新数据的业务类型与预设指标类型进行检测,当检测到存在预设指标类型的更新数据时,则表示风控数据库中的业务数据存在实质的变更。
具体地,风控服务器可以实时获取多个用户发生的业务数据以及关联数据等,例如交易数据和征信数据等,并将实时获取的数据利用分布式文件系统存储至预设的风控数据库中。风控服务器可以预先配置多个指标类型的数据,风控服务器则根据指标类型对数据库中的业务数据进行实时检测,当风控数据库中存在预设指标类型对应的数据发生变更时,表示业务数据发生实质性变化。例如,可以通过工具监控数据库中新增、修改等数据的变化,同步到大数据平台,根据预设的指标类型的计算数据是否发生实质变更。例如,可以利用kafka与spark的整合及时获取数据库中变动的数据,由此可以及时有效地检测到发生变更的重要数据。
在一个实施例中,根据业务类型标识获取预设的风控分析模型之前,该方法还包括:获取多个业务数据,利用多个业务数据生成训练集数据和验证集数据;通过聚类算法对训练集数据进行聚类分析,根据聚类结果提取达到预设阈值的特征向量;将特征向量输入至预设的神经网络模型中进行训练,得到初始风控分析模型;将验证集数据输入至初始风控分析模型中进行训练和验证,直到验证集数据的验证通过率满足预设阈值时,则停止训练,得到所需的风控分析模型。
风控服务器可以预先获取大量的业务数据以及关联数据等,并利用多个业务数据生成训练集数据和验证集数据。其中,训练集数据可以是已经添加了人工标注的数据集。风控服务器则进一步通过对训练集数据进行大数据分析,例如,可以通过聚类算法对训练集数据进行聚类分析和特征提取,根据聚类结果提取训练集数据的特征向量,并将特征向量输入至预设的神经网络模型中进行训练,通过预设的分析模型计算每个特征向量的重要性,得到初始风控分析模型。其中,预设的神经网络模型可以是基于决策树的神经网络模型或基于深度 学习的神经网络模型等。
风控服务器训练的到初始风控分析模型后,则进一步利用验证集数据对风控分析模型进行验证训练,当验证集数据中验证通过率满足预设阈值时,对应的类别概率在预设范围内的数量达到预设数据时,表示风控分析模型已经训练成功,则停止训练,得到所需的风控分析模型。
由于风控分析模型是利用大量数据进行训练,分析出这些数据内在必然的联系和数据之间的特征,进而根据这些数据内在必然的联系和数据之间的特征训练的到所需的风控分析模型,通过训练完成的风控分析模型对业务数据和关系图谱数据进行分析,从而能够有效地得到对应的分析结果数据。
例如,风控服务器可以将获取的大量业务数据根据关系图谱的图谱数据以及对应的关系进行数据映射,进而对大量的业务数据进行特征提取,得到对应的特征向量和对应的特征值,并将所有特征组合为二维向量,形成机器学习模型所需的训练数据集。服务器进一步通过对训练数据进行训练,以得到初始风控分析模型。服务器还可以进一步根据预设频率利用更新的数据对风控分析模型进行优化调参,当获得最优参数的模型后,对模型进行更新,由此可以有效地得到训练完成的风控分析模型。
在一个实施例中,如图4所示,利用风控分析模型对变更数据和关系图谱数据进行分析得到分析结果的步骤,具体包括以下内容:
步骤402,将业务数据和关系图谱数据输入至风控分析模型中,提取出业务数据对应的特征向量,以及关系图谱数据中的图谱元素特征。
步骤404,根据特征变量和图谱元素特征确定变更业务数据对应关系图谱数据的风险因子。
步骤406,利用风控分析模型根据风险因子和图谱元素特征计算多项风险指标值。
步骤408,根据多项风险指标值生成分析结果,并在分析结果中添加相应的风险等级标签。
风控服务器获取业务数据和建立关系图谱,并将业务数据以及关系图谱数据存储至预设的风控数据库后,对风控数据库中的数据进行检测,当检测到风控数据库中的业务数据存在变更时,根据用户标识获取对应的业务数据和关系图谱数据;根据业务类型获取预设的风控分析模型,通过风控分析模型中的预设指标对变更数据和关系图谱数据进行分析。
其中,特征向量指对业务数据和关系图谱数据进行线性变换和特征提取后,用于表示数据特征的向量。图谱元素特征包含实体数据特征和关系特征,实体数据特征和关系特征即表示实体数据和关联关系数据对应的特征值,并将特征值用以表示数据特征和关系特征。
具体地,风控服务器将业务数据和关系图谱数据输入至风控分析模型中, 风控分析模型对业务数据和关系图谱数据进行特征提取,提取出业务数据对应的特征向量,并提取出关系图谱数据中的图谱元素特征;风控服务器通过风控分析模型对特征向量、实体数据特征和关系特征进行风险分析,根据特征变量和图谱元素特征确定变更业务数据对应关系图谱数据的风险因子,其中,风险因子可以表示变更业务数据对关系图谱数据所产生的影响。监控服务器进而利用风控分析模型根据风险因子和图谱元素特征计算多项风险指标值,利用多项风险指标值按照预设方式生成对应的分析结果,并在分析结果中添加与风险指标值相应的风控标签。具体地,当分析结果中存在指标数据未达到预设指标时,表示存在风险,则在分析结果中添加风险标签。当分析结果中存在风险标签时,按照预设方式生成对应的风控报告发送至对应的监控终端,并进行预警提示,通过风控分析模型实时对变动的数据进行风控分析,由于风控分析模型是已经通过对大量业务数据和视图数据进行学习和训练得到的,由此可以通过风控分析模型对变更业务数据和对应的实体数据进行分析和决策,从而能够使得风控分析模型根据数据特征内在必然的联系,分析出更新业务数据对应的风险程度,进而能够准确有效地监控金融业务中存在的风险,有效地提高了风控的效率。
进一步地,风控服务器还可以对特定企业进行分析,对中小企业能力进行监控。具体地,监控终端可以向风控服务器发送风控请求,风控请求中包括了用户标识。风控服务器接收到监控终端发送的风控请求后,根据用户标识获取对应的金融领域的业务数据以及关联数据,获取预设的关系分析模型,通过关系分析模型分析出业务数据和关联数据中的多个实体数据以及多个实体数据之间的关联关系;根据多个实体数据之间的关联关系建立关系图谱。风控服务器进一步通过预设的风控分析模型对该用户标识的业务数据和关系图谱数据进行风控分析,并根据预设指标计算业务数据和关系图谱数据对应的多项风险指标值,并利用多项风险指标值按照预设方式生成对应的分析结果。风控服务器还可以进一步根据分析结果和关系图谱数据按照预设方式生成该用户标识对应的用户风控画像。
例如,对多个企业进行风险监控时,贷款风险指标可以包括不良贷款比率、贷款加权风险度、贷款分散化比率、不良贷款拨备覆盖率等。将关系图谱中贷款人节点和相关指标相结合,设定报警阈值,通过机器学习等技术,找到隐蔽的风险结构,指标特征,能够快速找出相关责任方和其关联方,形成报告供业务人员进行调查。通过对特定企业用户分析出对应的关系图谱和企业画像,能够准确有效地对企业潜在的风险进行监控。
应该理解的是,虽然图2-4的流程图中的各个步骤按照箭头的指示依次显示,但是这些步骤并不是必然按照箭头指示的顺序依次执行。除非本文中有明确的说明,这些步骤的执行并没有严格的顺序限制,这些步骤可以以其它的顺序执行。而且,图2-4中的至少一部分步骤可以包括多个子步骤或者多个阶段, 这些子步骤或者阶段并不必然是在同一时刻执行完成,而是可以在不同的时刻执行,这些子步骤或者阶段的执行顺序也不必然是依次进行,而是可以与其它步骤或者其它步骤的子步骤或者阶段的至少一部分轮流或者交替地执行。
在一个实施例中,如图5所示,提供了一种业务数据的风控分析处理装置,包括:数据监测模块502、数据获取模块504、风控分析模块506和风控报告生成模块508,其中:
数据监测模块502,用于当检测到风控数据库中的业务数据发生变更时,获取变更业务数据,根据变更业务数据确定目标用户标识和业务类型标识;风控数据库中的信息用于描述业务数据、用户标识以及关系图谱数据之间的对应关系;
数据获取模块504,用于根据目标用户标识从风控数据库中获取对应的关系图谱数据;其中,关系图谱数据为通过预设的关系分析模型基于业务服务器对业务请求进行处理产生的业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系,并根据关联关系建立的关系图谱;
风控分析模块506,用于根据业务类型标识获取预设的风控分析模型,利用风控分析模型确定变更业务数据对应关系图谱数据的风险因子,根据风险因子和关系图谱数据计算风险指标值,根据风险指标值生成分析结果;
风控报告生成模块508,用于根据分析结果按照预设方式生成对应的风控报告,并将风控报告发送至监控终端。
在其中一个实施例中,该装置还包括关系数据分析模块,用于获取业务服务器对业务请求进行处理产生的业务数据,并将业务数据以及业务数据携带的用户标识对应存储至风控数据库;根据用户标识获取对应的历史关联数据,并通过预设的关系分析模型基于业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系;业务数据和历史关联数据均携带相应的业务类型标识;根据关联关系建立关系图谱,并将关系图谱和相应的关系图谱数据存储至风控数据库。
在其中一个实施例中,关系数据分析模块还用于将业务数据和历史关联数据输入至关系分析模型中,对业务数据和历史关联数据进行特征提取,得到多个实体数据的特征向量;通过关系分析模型对特征向量进行分析,得到实体数据的数据属性和关键字标签;根据实体数据的业务类型、数据属性和关键字标签分析多个实体数据之间的关系特征,根据关系特征确定出多个实体数据之间的关联关系。
在其中一个实施例中,关系数据分析模块还用于根据多个实体数据生成对应的多个数据节点;根据业务数据和历史关联数据的业务类型、数据属性和关键字标签生成多个数据节点的描述信息;根据描述信息和多个实体数据之间的 关系特征确定多个数据节点的映射关系和关系类型,并根据映射关系和关系类型将多个数据节点进行链接;根据链接后的多个数据节点和描述信息生成对应的关系图谱。
在其中一个实施例中,数据监测模块502还用于获取风控数据库中的更新数据;获取更新数据的业务类型,将更新数据的业务类型与预设指标类型进行比较;当存在预设指标类型的更新数据时,则表示风控数据库中的业务数据存在变更。
在其中一个实施例中,该装置还包括模型训练模块,用于获取多个业务数据,利用多个业务数据生成训练集数据和验证集数据;通过聚类算法对训练集数据进行聚类分析,根据聚类结果提取达到预设阈值的特征向量;将特征向量输入至预设的神经网络模型中进行训练,得到初始风控分析模型;将验证集数据输入至初始风控分析模型中进行训练和验证,直到验证集数据的验证通过率满足预设阈值时,则停止训练,得到所需的风控分析模型。
在其中一个实施例中,风控分析模块506还用于将业务数据和关系图谱数据输入至风控分析模型中,提取出变更业务数据对应的特征向量,以及关系图谱数据中的图谱元素特征;根据特征变量和图谱元素特征确定变更业务数据对应关系图谱数据的风险因子;利用风控分析模型根据风险因子和图谱元素特征计算多项风险指标值;根据多项风险指标值生成分析结果,并在分析结果中添加相应的风险等级标签。
关于业务数据的风控分析处理装置的具体限定可以参见上文中对于业务数据的风控分析处理方法的限定,在此不再赘述。上述业务数据的风控分析处理装置中的各个模块可全部或部分通过软件、硬件及其组合来实现。上述各模块可以硬件形式内嵌于或独立于计算机设备中的处理器中,也可以以软件形式存储于计算机设备中的存储器中,以便于处理器调用执行以上各个模块对应的操作。
在一个实施例中,提供了一种计算机设备,该计算机设备可以是服务器,其内部结构图可以如图6所示。该计算机设备包括通过系统总线连接的处理器、存储器、网络接口和数据库。其中,该计算机设备的处理器用于提供计算和控制能力。该计算机设备的存储器包括非易失性存储介质、内存储器。该非易失性存储介质存储有操作系统、计算机程序和数据库。该内存储器为非易失性存储介质中的操作系统和计算机程序的运行提供环境。该计算机设备的数据库用于存储业务数据、历史关联数据、关系图谱和关系图谱数据等。该计算机设备的网络接口用于与外部的终端通过网络连接通信。该计算机程序被处理器执行时以实现本申请任意一个实施例中提供的业务数据的风控分析处理方法的步骤。
本领域技术人员可以理解,图6中示出的结构,仅仅是与本申请方案相关 的部分结构的框图,并不构成对本申请方案所应用于其上的计算机设备的限定,具体的计算机设备可以包括比图中所示更多或更少的部件,或者组合某些部件,或者具有不同的部件布置。
在一个实施例中,提供了一种计算机可读存储介质,所述计算机可读存储介质可以是非易失性,也可以是易失性,其上存储有计算机程序,计算机程序被处理器执行时实现本申请任意一个实施例中提供的业务数据的风控分析处理方法的步骤。
本领域普通技术人员可以理解实现上述实施例方法中的全部或部分流程,是可以通过计算机程序来指令相关的硬件来完成,所述的计算机程序可存储于一非易失性计算机可读取存储介质中,该计算机程序在执行时,可包括如上述各方法的实施例的流程。其中,本申请所提供的各实施例中所使用的对存储器、存储、数据库或其它介质的任何引用,均可包括非易失性和/或易失性存储器。非易失性存储器可包括只读存储器(ROM)、可编程ROM(PROM)、电可编程ROM(EPROM)、电可擦除可编程ROM(EEPROM)或闪存。易失性存储器可包括随机存取存储器(RAM)或者外部高速缓冲存储器。作为说明而非局限,RAM以多种形式可得,诸如静态RAM(SRAM)、动态RAM(DRAM)、同步DRAM(SDRAM)、双数据率SDRAM(DDRSDRAM)、增强型SDRAM(ESDRAM)、同步链路(Synchlink)DRAM(SLDRAM)、存储器总线(Rambus)直接RAM(RDRAM)、直接存储器总线动态RAM(DRDRAM)、以及存储器总线动态RAM(RDRAM)等。
以上实施例的各技术特征可以进行任意的组合,为使描述简洁,未对上述实施例中的各个技术特征所有可能的组合都进行描述,然而,只要这些技术特征的组合不存在矛盾,都应当认为是本说明书记载的范围。
以上所述实施例仅表达了本申请的几种实施方式,其描述较为具体和详细,但并不能因此而理解为对发明专利范围的限制。应当指出的是,对于本领域的普通技术人员来说,在不脱离本申请构思的前提下,还可以做出若干变形和改进,这些都属于本申请的保护范围。因此,本申请专利的保护范围应以所附权利要求为准。

Claims (21)

  1. 一种业务数据的风控分析处理方法,所述方法包括:
    当检测到风控数据库中的业务数据发生变更时,获取变更业务数据,根据所述变更业务数据确定目标用户标识和业务类型标识;所述风控数据库中的信息用于描述业务数据、用户标识以及关系图谱数据之间的对应关系;
    根据所述目标用户标识从所述风控数据库中对应的关系图谱数据;其中,所述关系图谱数据为通过预设的关系分析模型基于业务服务器对业务请求进行处理产生的业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系,并根据所述关联关系建立的关系图谱;
    根据所述业务类型标识获取预设的风控分析模型,利用所述风控分析模型确定所述变更业务数据对应所述关系图谱数据的风险因子,根据所述风险因子和关系图谱数据计算风险指标值,根据所述风险指标值生成分析结果;
    根据所述分析结果按照预设方式生成对应的风控报告,并将所述风控报告发送至监控终端。
  2. 根据权利要求1所述的方法,其中所述建立关系图谱数据的步骤包括:
    获取业务服务器对业务请求进行处理产生的业务数据,并将所述业务数据以及所述业务数据携带的用户标识对应存储至风控数据库;
    根据用户标识获取对应的历史关联数据,并通过预设的关系分析模型基于所述业务数据和所述历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系;所述业务数据和历史关联数据均携带相应的业务类型标识;
    根据所述关联关系建立关系图谱,并将所述关系图谱和相应的关系图谱数据存储至所述风控数据库。
  3. 根据权利要求2所述的方法,其中所述确定出多个实体数据以及多个实体数据之间的关联关系的步骤,包括:
    将所述业务数据和历史关联数据输入至所述关系分析模型中,对所述业务数据和历史关联数据进行特征提取,得到多个实体数据的特征向量;
    通过所述关系分析模型对所述特征向量进行分析,得到所述实体数据的数据属性和关键字标签;
    根据所述实体数据的业务类型、数据属性和关键字标签分析多个实体数据之间的关系特征,根据所述关系特征确定出所述多个实体数据之间的关联关系。
  4. 根据权利要求2所述的方法,其中所述根据所述关联关系建立关系图谱的步骤包括:
    根据所述多个实体数据生成对应的多个数据节点;
    根据所述业务数据和历史关联数据的业务类型、数据属性和关键字标签生成所述多个数据节点的描述信息;
    根据所述描述信息和多个实体数据之间的关系特征确定多个数据节点的映射关系和关系类型,并根据所述映射关系和关系类型将所述多个数据节点进行 链接;
    根据链接后的多个数据节点和描述信息生成对应的关系图谱。
  5. 根据权利要求1所述的方法,其中检测到所述风控数据库中的业务数据发生变更的步骤包括:
    获取所述风控数据库中的更新数据;
    获取所述更新数据的业务类型,将所述更新数据的业务类型与预设指标类型进行比较;
    当存在预设指标类型的更新数据时,则表示所述风控数据库中的业务数据存在变更。
  6. 根据权利要求1所述的方法,其中所述根据所述业务类型标识获取预设的风控分析模型之前,所述方法还包括:
    获取多个业务数据,利用所述多个业务数据生成训练集数据和验证集数据;
    通过聚类算法对所述训练集数据进行聚类分析,根据聚类结果提取达到预设阈值的特征向量;
    将所述特征向量输入至预设的神经网络模型中进行训练,得到初始风控分析模型;
    将所述验证集数据输入至所述初始风控分析模型中进行训练和验证,直到所述验证集数据的验证通过率满足预设阈值时,则停止训练,得到所需的风控分析模型。
  7. 根据权利要求1至6任意一项所述的方法,其中所述利用所述风控分析模型对变更数据和关系图谱数据进行分析得到分析结果,包括:
    将所述变更业务数据和关系图谱数据输入至风控分析模型中,提取出所述变更业务数据对应的特征向量,以及所述关系图谱数据中的图谱元素特征;
    根据所述特征变量和所述图谱元素特征确定所述变更业务数据对应所述关系图谱数据的风险因子;
    利用所述风控分析模型根据所述风险因子和图谱元素特征计算多项风险指标值;
    根据所述多项风险指标值生成分析结果,并在所述分析结果中添加相应的风险等级标签。
  8. 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:
    当检测到风控数据库中的业务数据发生变更时,获取变更业务数据,根据所述变更业务数据确定目标用户标识和业务类型标识;所述风控数据库中的信息用于描述业务数据、用户标识以及关系图谱数据之间的对应关系;
    根据所述目标用户标识从所述风控数据库中对应的关系图谱数据;其中,所述关系图谱数据为通过预设的关系分析模型基于业务服务器对业务请求进行 处理产生的业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系,并根据所述关联关系建立的关系图谱;
    根据所述业务类型标识获取预设的风控分析模型,利用所述风控分析模型确定所述变更业务数据对应所述关系图谱数据的风险因子,根据所述风险因子和关系图谱数据计算风险指标值,根据所述风险指标值生成分析结果;
    根据所述分析结果按照预设方式生成对应的风控报告,并将所述风控报告发送至监控终端。
  9. 根据权利要求1所述的计算机设备,其中所述建立关系图谱数据的步骤包括:
    获取业务服务器对业务请求进行处理产生的业务数据,并将所述业务数据以及所述业务数据携带的用户标识对应存储至风控数据库;
    根据用户标识获取对应的历史关联数据,并通过预设的关系分析模型基于所述业务数据和所述历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系;所述业务数据和历史关联数据均携带相应的业务类型标识;
    根据所述关联关系建立关系图谱,并将所述关系图谱和相应的关系图谱数据存储至所述风控数据库。
  10. 根据权利要求9所述的计算机设备,其中所述确定出多个实体数据以及多个实体数据之间的关联关系的步骤,包括:
    将所述业务数据和历史关联数据输入至所述关系分析模型中,对所述业务数据和历史关联数据进行特征提取,得到多个实体数据的特征向量;
    通过所述关系分析模型对所述特征向量进行分析,得到所述实体数据的数据属性和关键字标签;
    根据所述实体数据的业务类型、数据属性和关键字标签分析多个实体数据之间的关系特征,根据所述关系特征确定出所述多个实体数据之间的关联关系。
  11. 根据权利要求9所述的计算机设备,其中所述根据所述关联关系建立关系图谱的步骤包括:
    根据所述多个实体数据生成对应的多个数据节点;
    根据所述业务数据和历史关联数据的业务类型、数据属性和关键字标签生成所述多个数据节点的描述信息;
    根据所述描述信息和多个实体数据之间的关系特征确定多个数据节点的映射关系和关系类型,并根据所述映射关系和关系类型将所述多个数据节点进行链接;
    根据链接后的多个数据节点和描述信息生成对应的关系图谱。
  12. 根据权利要求8所述的计算机设备,其中检测到所述风控数据库中的业务数据发生变更的步骤包括:
    获取所述风控数据库中的更新数据;
    获取所述更新数据的业务类型,将所述更新数据的业务类型与预设指标类型进行比较;
    当存在预设指标类型的更新数据时,则表示所述风控数据库中的业务数据存在变更。
  13. 根据权利要求8所述的计算机设备,其中所述根据所述业务类型标识获取预设的风控分析模型之前,所述方法还包括:
    获取多个业务数据,利用所述多个业务数据生成训练集数据和验证集数据;
    通过聚类算法对所述训练集数据进行聚类分析,根据聚类结果提取达到预设阈值的特征向量;
    将所述特征向量输入至预设的神经网络模型中进行训练,得到初始风控分析模型;
    将所述验证集数据输入至所述初始风控分析模型中进行训练和验证,直到所述验证集数据的验证通过率满足预设阈值时,则停止训练,得到所需的风控分析模型。
  14. 根据权利要求8至13任意一项所述的计算机设备,其中所述利用所述风控分析模型对变更数据和关系图谱数据进行分析得到分析结果,包括:
    将所述变更业务数据和关系图谱数据输入至风控分析模型中,提取出所述变更业务数据对应的特征向量,以及所述关系图谱数据中的图谱元素特征;
    根据所述特征变量和所述图谱元素特征确定所述变更业务数据对应所述关系图谱数据的风险因子;
    利用所述风控分析模型根据所述风险因子和图谱元素特征计算多项风险指标值;
    根据所述多项风险指标值生成分析结果,并在所述分析结果中添加相应的风险等级标签。
  15. 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:
    当检测到风控数据库中的业务数据发生变更时,获取变更业务数据,根据所述变更业务数据确定目标用户标识和业务类型标识;所述风控数据库中的信息用于描述业务数据、用户标识以及关系图谱数据之间的对应关系;
    根据所述目标用户标识从所述风控数据库中对应的关系图谱数据;其中,所述关系图谱数据为通过预设的关系分析模型基于业务服务器对业务请求进行处理产生的业务数据和历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系,并根据所述关联关系建立的关系图谱;
    根据所述业务类型标识获取预设的风控分析模型,利用所述风控分析模型确定所述变更业务数据对应所述关系图谱数据的风险因子,根据所述风险因子和关系图谱数据计算风险指标值,根据所述风险指标值生成分析结果;
    根据所述分析结果按照预设方式生成对应的风控报告,并将所述风控报告发送至监控终端。
  16. 根据权利要求15所述的计算机可读存储介质,其中所述建立关系图谱数据的步骤包括:
    获取业务服务器对业务请求进行处理产生的业务数据,并将所述业务数据以及所述业务数据携带的用户标识对应存储至风控数据库;
    根据用户标识获取对应的历史关联数据,并通过预设的关系分析模型基于所述业务数据和所述历史关联数据,确定出多个实体数据以及多个实体数据之间的关联关系;所述业务数据和历史关联数据均携带相应的业务类型标识;
    根据所述关联关系建立关系图谱,并将所述关系图谱和相应的关系图谱数据存储至所述风控数据库。
  17. 根据权利要求16所述的计算机可读存储介质,其中所述确定出多个实体数据以及多个实体数据之间的关联关系的步骤,包括:
    将所述业务数据和历史关联数据输入至所述关系分析模型中,对所述业务数据和历史关联数据进行特征提取,得到多个实体数据的特征向量;
    通过所述关系分析模型对所述特征向量进行分析,得到所述实体数据的数据属性和关键字标签;
    根据所述实体数据的业务类型、数据属性和关键字标签分析多个实体数据之间的关系特征,根据所述关系特征确定出所述多个实体数据之间的关联关系。
  18. 根据权利要求16所述的计算机可读存储介质,其中所述根据所述关联关系建立关系图谱的步骤包括:
    根据所述多个实体数据生成对应的多个数据节点;
    根据所述业务数据和历史关联数据的业务类型、数据属性和关键字标签生成所述多个数据节点的描述信息;
    根据所述描述信息和多个实体数据之间的关系特征确定多个数据节点的映射关系和关系类型,并根据所述映射关系和关系类型将所述多个数据节点进行链接;
    根据链接后的多个数据节点和描述信息生成对应的关系图谱。
  19. 根据权利要求15所述的计算机可读存储介质,其中检测到所述风控数据库中的业务数据发生变更的步骤包括:
    获取所述风控数据库中的更新数据;
    获取所述更新数据的业务类型,将所述更新数据的业务类型与预设指标类型进行比较;
    当存在预设指标类型的更新数据时,则表示所述风控数据库中的业务数据存在变更。
  20. 根据权利要求15所述的计算机可读存储介质,其中所述根据所述业务 类型标识获取预设的风控分析模型之前,所述方法还包括:
    获取多个业务数据,利用所述多个业务数据生成训练集数据和验证集数据;
    通过聚类算法对所述训练集数据进行聚类分析,根据聚类结果提取达到预设阈值的特征向量;
    将所述特征向量输入至预设的神经网络模型中进行训练,得到初始风控分析模型;
    将所述验证集数据输入至所述初始风控分析模型中进行训练和验证,直到所述验证集数据的验证通过率满足预设阈值时,则停止训练,得到所需的风控分析模型。
  21. 根据权利要求15至20任意一项所述的计算机可读存储介质,其中所述利用所述风控分析模型对变更数据和关系图谱数据进行分析得到分析结果,包括:
    将所述变更业务数据和关系图谱数据输入至风控分析模型中,提取出所述变更业务数据对应的特征向量,以及所述关系图谱数据中的图谱元素特征;
    根据所述特征变量和所述图谱元素特征确定所述变更业务数据对应所述关系图谱数据的风险因子;
    利用所述风控分析模型根据所述风险因子和图谱元素特征计算多项风险指标值;
    根据所述多项风险指标值生成分析结果,并在所述分析结果中添加相应的风险等级标签。
PCT/CN2020/086088 2019-06-18 2020-04-22 业务数据的风控分析处理方法、装置和计算机设备 WO2020253358A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201910526011.6A CN110390465A (zh) 2019-06-18 2019-06-18 业务数据的风控分析处理方法、装置和计算机设备
CN201910526011.6 2019-06-18

Publications (1)

Publication Number Publication Date
WO2020253358A1 true WO2020253358A1 (zh) 2020-12-24

Family

ID=68285769

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/086088 WO2020253358A1 (zh) 2019-06-18 2020-04-22 业务数据的风控分析处理方法、装置和计算机设备

Country Status (2)

Country Link
CN (1) CN110390465A (zh)
WO (1) WO2020253358A1 (zh)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113094506A (zh) * 2021-04-14 2021-07-09 每日互动股份有限公司 一种基于关系图谱的预警方法、计算机设备及存储介质
CN113326879A (zh) * 2021-05-31 2021-08-31 深圳前海微众银行股份有限公司 一种业务数据的监测方法及装置
CN113590835A (zh) * 2021-07-28 2021-11-02 上海致景信息科技有限公司 纺织行业数据的知识图谱构建方法、装置及处理器
CN113657779A (zh) * 2021-08-20 2021-11-16 杭州时趣信息技术有限公司 一种动态配置化的风控决策方法、系统、设备和存储介质

Families Citing this family (51)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110390465A (zh) * 2019-06-18 2019-10-29 深圳壹账通智能科技有限公司 业务数据的风控分析处理方法、装置和计算机设备
CN112783889A (zh) * 2019-11-07 2021-05-11 中国石油化工股份有限公司 用于建立变更风险控制措施库的方法和装置
CN110909384B (zh) * 2019-11-19 2021-11-12 支付宝(杭州)信息技术有限公司 确定泄露用户信息的业务方的方法及装置
CN110909805B (zh) * 2019-11-26 2022-02-18 西安交通大学城市学院 基于大数据与Inception V3深度网络模型的金融风控系统
CN110991871A (zh) * 2019-11-29 2020-04-10 深圳前海微众银行股份有限公司 风险监测方法、装置、设备与计算机可读存储介质
CN111143163B (zh) * 2019-12-13 2024-04-16 上海硬通网络科技有限公司 数据监控方法、装置、计算机设备和存储介质
CN111178615B (zh) * 2019-12-24 2023-10-27 成都数联铭品科技有限公司 一种企业风险识别模型的构建方法及系统
CN113037792B (zh) * 2019-12-24 2022-07-26 中国移动通信集团四川有限公司 一种业务识别参数的验证方法、装置及网关
CN111177284A (zh) * 2019-12-31 2020-05-19 清华大学 应急预案模型生成方法、装置及设备
CN111353702A (zh) * 2020-02-28 2020-06-30 中国工商银行股份有限公司 变更操作风险计算方法和装置
CN111428971A (zh) * 2020-03-05 2020-07-17 支付宝(杭州)信息技术有限公司 业务数据的处理方法、装置和服务器
CN111581222A (zh) * 2020-03-30 2020-08-25 深圳壹账通智能科技有限公司 业务数据的关联分析方法、装置、计算机设备及计算机存储介质
CN111582488A (zh) * 2020-04-23 2020-08-25 傲林科技有限公司 一种事件推演方法及装置
CN111476664A (zh) * 2020-05-04 2020-07-31 武汉众邦银行股份有限公司 一直基于实时数据流的风险感知和风险预测方法
CN111582873B (zh) * 2020-05-07 2023-01-17 支付宝(杭州)信息技术有限公司 评估交互事件的方法及装置、电子设备、存储介质
CN111784495B (zh) * 2020-06-04 2021-05-04 江苏常熟农村商业银行股份有限公司 担保圈识别方法、装置、计算机设备和存储介质
CN111782710B (zh) * 2020-06-30 2023-07-25 北京百度网讯科技有限公司 数据存储方法、装置、计算设备和介质
CN111754340B (zh) * 2020-07-03 2022-06-21 交通银行股份有限公司 一种基于图数据库的担保网络风险排查系统
CN111880989B (zh) * 2020-07-14 2024-05-17 中国银联股份有限公司 一种配置项管理方法及装置
CN111797406A (zh) * 2020-07-15 2020-10-20 智博云信息科技(广州)有限公司 一种医疗基金数据分析处理方法、装置及可读存储介质
CN111966726B (zh) * 2020-07-22 2023-09-26 武汉极意网络科技有限公司 基于不同类型客户的自适应数据分析报告生成系统及方法
CN111798162A (zh) * 2020-07-23 2020-10-20 深圳壹账通智能科技有限公司 基于神经网络的风险监测方法及装置
CN112149951A (zh) * 2020-08-11 2020-12-29 招联消费金融有限公司 风险控制方法、装置、计算机设备和存储介质
CN112015909B (zh) * 2020-08-19 2024-04-30 普洛斯科技(重庆)有限公司 知识图谱的构建方法及装置、电子设备、存储介质
CN112184235B (zh) * 2020-09-04 2022-05-13 支付宝(杭州)信息技术有限公司 风控数据变更方法和装置
CN112100451B (zh) * 2020-09-14 2023-11-17 上海飞机制造有限公司 基于图数据库搭建工业神经网络的方法
CN112184012B (zh) * 2020-09-27 2024-05-31 平安资产管理有限责任公司 一种企业风险预警方法、装置、设备及可读存储介质
CN112328802A (zh) * 2020-09-30 2021-02-05 支付宝(杭州)信息技术有限公司 数据处理方法、装置和服务器
CN112465461A (zh) * 2020-12-04 2021-03-09 金蝶云科技有限公司 业务对象信息变更方法、系统、计算机设备和存储介质
CN112634069A (zh) * 2020-12-30 2021-04-09 太平金融科技服务(上海)有限公司 核保请求处理方法、装置、计算机设备和存储介质
CN112907364A (zh) * 2021-04-01 2021-06-04 重庆度小满优扬科技有限公司 一种风控分析模型的评估方法及装置
CN113077336B (zh) * 2021-04-15 2024-05-24 北京同邦卓益科技有限公司 风控信息配置方法、装置、电子设备和计算机可读介质
CN113222223B (zh) * 2021-04-24 2023-02-21 上海钢银科技发展有限公司 实时数仓的风控联动预警方法、系统、设备及存储介质
CN113191889A (zh) * 2021-05-26 2021-07-30 中国工商银行股份有限公司 风控配置方法、配置系统、电子设备及可读存储介质
CN113112352A (zh) * 2021-05-27 2021-07-13 中国工商银行股份有限公司 风险业务检测模型训练方法、风险业务检测方法及装置
CN113282625B (zh) * 2021-05-31 2022-10-04 重庆富民银行股份有限公司 一种基于sql的api数据查询和处理系统及方法
CN113393159A (zh) * 2021-07-07 2021-09-14 上海软中信息技术有限公司 一种基于关联网络的智能风控平台系统、装置和设备
CN113656808A (zh) * 2021-08-31 2021-11-16 平安医疗健康管理股份有限公司 数据安全评估方法、装置、设备及存储介质
CN113919679B (zh) * 2021-09-30 2023-06-20 武汉金豆医疗数据科技有限公司 业务流程风险防控方法及系统
CN114091899A (zh) * 2021-11-22 2022-02-25 支付宝(杭州)信息技术有限公司 风险特征识别方法以及装置
CN114240097A (zh) * 2021-12-02 2022-03-25 支付宝(杭州)信息技术有限公司 一种风险评估的方法及装置
CN114186873B (zh) * 2021-12-14 2024-05-17 平安银行股份有限公司 加工变量验证方法、装置及相关设备
CN114299764B (zh) * 2021-12-17 2022-11-08 南京莱斯信息技术股份有限公司 民航航班时刻计划与预先飞行计划联动变更一致性方法
CN114500048B (zh) * 2022-01-26 2023-10-03 南方电网数字电网研究院有限公司 基于网络安全的外部威胁情报分析方法及系统
CN114844681A (zh) * 2022-04-11 2022-08-02 中国科学院信息工程研究所 基于关联图的分析方法、系统、电子设备、存储介质
CN114693459A (zh) * 2022-04-15 2022-07-01 北京百度网讯科技有限公司 基于金融场景的风险控制方法、装置及电子设备
CN114647555B (zh) * 2022-05-13 2022-09-02 太平金融科技服务(上海)有限公司 基于多业务系统的数据预警方法、装置、设备和介质
CN117806899A (zh) * 2022-09-26 2024-04-02 华为云计算技术有限公司 数据监控分析方法、装置、服务器、运维系统及存储介质
CN115391670B (zh) * 2022-11-01 2023-04-07 南京嘉安网络技术有限公司 一种基于知识图谱的互联网行为分析方法与系统
CN115730605B (zh) * 2022-11-21 2024-02-02 暨南大学 基于多维信息的数据分析方法
CN115983636B (zh) * 2022-12-26 2023-11-17 深圳市中政汇智管理咨询有限公司 风险评估方法、装置、设备及存储介质

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108564460A (zh) * 2018-01-12 2018-09-21 阳光财产保险股份有限公司 互联网信贷场景下的实时欺诈检测方法及装置
CN109064318A (zh) * 2018-08-24 2018-12-21 苏宁消费金融有限公司 一种基于知识图谱的互联网金融风险监测系统
CN109636607A (zh) * 2018-12-18 2019-04-16 平安科技(深圳)有限公司 基于模型部署的业务数据处理方法、装置和计算机设备
CN109657837A (zh) * 2018-11-19 2019-04-19 平安科技(深圳)有限公司 违约概率预测方法、装置、计算机设备和存储介质
CN110390465A (zh) * 2019-06-18 2019-10-29 深圳壹账通智能科技有限公司 业务数据的风控分析处理方法、装置和计算机设备

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2006031306A (ja) * 2004-07-15 2006-02-02 Chugoku Electric Power Co Inc:The リスク共済システム,リスク共済方法およびリスク共済プログラム

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108564460A (zh) * 2018-01-12 2018-09-21 阳光财产保险股份有限公司 互联网信贷场景下的实时欺诈检测方法及装置
CN109064318A (zh) * 2018-08-24 2018-12-21 苏宁消费金融有限公司 一种基于知识图谱的互联网金融风险监测系统
CN109657837A (zh) * 2018-11-19 2019-04-19 平安科技(深圳)有限公司 违约概率预测方法、装置、计算机设备和存储介质
CN109636607A (zh) * 2018-12-18 2019-04-16 平安科技(深圳)有限公司 基于模型部署的业务数据处理方法、装置和计算机设备
CN110390465A (zh) * 2019-06-18 2019-10-29 深圳壹账通智能科技有限公司 业务数据的风控分析处理方法、装置和计算机设备

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113094506A (zh) * 2021-04-14 2021-07-09 每日互动股份有限公司 一种基于关系图谱的预警方法、计算机设备及存储介质
CN113094506B (zh) * 2021-04-14 2023-08-18 每日互动股份有限公司 一种基于关系图谱的预警方法、计算机设备及存储介质
CN113326879A (zh) * 2021-05-31 2021-08-31 深圳前海微众银行股份有限公司 一种业务数据的监测方法及装置
CN113590835A (zh) * 2021-07-28 2021-11-02 上海致景信息科技有限公司 纺织行业数据的知识图谱构建方法、装置及处理器
CN113657779A (zh) * 2021-08-20 2021-11-16 杭州时趣信息技术有限公司 一种动态配置化的风控决策方法、系统、设备和存储介质
CN113657779B (zh) * 2021-08-20 2024-01-09 杭州时趣信息技术有限公司 一种动态配置化的风控决策方法、系统、设备和存储介质

Also Published As

Publication number Publication date
CN110390465A (zh) 2019-10-29

Similar Documents

Publication Publication Date Title
WO2020253358A1 (zh) 业务数据的风控分析处理方法、装置和计算机设备
CN108876600B (zh) 预警信息推送方法、装置、计算机设备和介质
WO2021004333A1 (zh) 基于知识图谱的事件处理方法、装置、设备和存储介质
CN108876133B (zh) 基于业务信息的风险评估处理方法、装置、服务器和介质
WO2021004132A1 (zh) 异常数据检测方法、装置、计算机设备和存储介质
CN109165840B (zh) 风险预测处理方法、装置、计算机设备和介质
CN109858737B (zh) 基于模型部署的评分模型调整方法、装置和计算机设备
US20190311367A1 (en) System and method for using a data genome to identify suspicious financial transactions
US20190259033A1 (en) System and method for using a data genome to identify suspicious financial transactions
WO2019196546A1 (zh) 确定业务请求事件的风险概率的方法及装置
US11748416B2 (en) Machine-learning system for servicing queries for digital content
WO2021128679A1 (zh) 基于数据决策的测试数据生成方法、装置、计算机设备
KR20210116439A (ko) 자금 세탁 방지 분석을 위한 시스템 및 방법
US11538044B2 (en) System and method for generation of case-based data for training machine learning classifiers
CN110781308B (zh) 一种基于大数据构建知识图谱的反欺诈系统
CN109063921B (zh) 客户风险预警的优化处理方法、装置、计算机设备和介质
WO2020073727A1 (zh) 风险预测方法、装置、计算机设备和存储介质
WO2022001517A1 (zh) 基于谣言预测模型的信息发送方法、装置和计算机设备
CN110442713A (zh) 文章管理方法、装置、计算机设备和存储介质
Braun et al. Improving card fraud detection through suspicious pattern discovery
CN116307671A (zh) 风险预警方法、装置、计算机设备、存储介质
Ramaki et al. Credit card fraud detection based on ontology graph
CN112669039B (zh) 基于知识图谱的客户风险管控系统及方法
US20220374524A1 (en) Method and system for anamoly detection in the banking system with graph neural networks (gnns)
CN117009509A (zh) 数据安全分级方法、装置、设备、存储介质和程序产品

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: 20826936

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: 20826936

Country of ref document: EP

Kind code of ref document: A1