CN112668944A - Enterprise wind control method, device, equipment and medium based on big data credit investigation - Google Patents
Enterprise wind control method, device, equipment and medium based on big data credit investigation Download PDFInfo
- Publication number
- CN112668944A CN112668944A CN202110106549.9A CN202110106549A CN112668944A CN 112668944 A CN112668944 A CN 112668944A CN 202110106549 A CN202110106549 A CN 202110106549A CN 112668944 A CN112668944 A CN 112668944A
- Authority
- CN
- China
- Prior art keywords
- data
- enterprise
- credit investigation
- evaluated
- fraud
- Prior art date
- Legal status (The legal status 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 status listed.)
- Pending
Links
Images
Abstract
The embodiment of the specification discloses an enterprise wind control method, a device, equipment and a medium based on big data credit investigation, wherein the method is executed by an enterprise wind control system, the enterprise wind control system comprises a credit investigation data standardization module, a credit investigation data indexing module, an enterprise access module, an enterprise anti-fraud module and an internal credit scoring module, and the method comprises the following steps: processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module to determine a structured data item; processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises; automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result; and carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
Description
Technical Field
The specification relates to the field of enterprise risk control, in particular to an enterprise wind control method, device, equipment and medium based on big data credit investigation.
Background
Risk control refers to the risk supervisor taking various measures and methods to eliminate or reduce the various possibilities of occurrence of a risk event, or to reduce the losses incurred when a risk event occurs. In the current risk control of enterprises, credit plays an increasingly important role, and meanwhile, the credit degree of the enterprise is also an important factor for risk supervision and judgment of the enterprise.
In the prior art, the scheme for risk control of enterprises based on credit may have inaccuracy, and cannot well meet the requirements of current enterprise supervisors.
Based on this, a more accurate risk control scheme is needed for the supervisors of the existing enterprises.
Disclosure of Invention
One or more embodiments of the present specification provide an enterprise wind control method, apparatus, device, and medium based on big data credit investigation, which are used to solve the following technical problems: a more accurate risk control scheme is needed for the supervisors of existing enterprises.
One or more embodiments of the present disclosure adopt the following technical solutions:
one or more embodiments of the present specification provide an enterprise wind control method based on big data credit investigation, where the method is performed by an enterprise wind control system, the enterprise wind control system includes a credit investigation data standardization module, a credit investigation data indexing module, an enterprise admission module, an enterprise anti-fraud module, and an internal credit scoring module, and the method includes: processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module to determine a structured data item; processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises; automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result; and carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
Further, the credit investigation data acquired by the plurality of data sources is processed by the credit investigation data standardization module to determine a structured data item, which specifically includes: acquiring credit investigation data through a credit investigation data standardization module, managing the credit investigation data through a pre-established enterprise credit investigation data warehouse, and determining required business data; and carrying out enterprise credit investigation report inquiry or enterprise credit evaluation on the business data to determine a corresponding structured data item.
Further, the credit investigation data acquisition is performed through a credit investigation data standardization module, which specifically comprises the following steps: and the credit investigation data standardization module is used for carrying out the work of enterprise credit investigation data warehouse construction, data source definition, data extraction, conversion cleaning and loading.
Further, the enterprise credit investigation data warehouse comprises a source data layer, a verification data layer, a basic data layer and a product data layer; wherein the source data layer: the credit investigation system is used for uniformly and standardly processing and storing credit investigation data and updating the credit investigation data according to a preset strategy; the check data layer: the management mechanism is used for verifying the data content through a set data verification rule and a verification process, loading the verified data into the central database of the basic data layer, returning the data which does not pass the verification to the source data layer, and feeding the error data back to the corresponding data source after forming a data feedback message; the basic data layer: the data processing device is used for storing the data which passes the verification of the verification data layer, returning the error data checked out through the post verification to the verification data layer, and finally forming an error feedback message to be fed back to the reporting mechanism; the product data layer: the product data extraction logic and mining analysis are used for loading the product data into the product data layer according to the types of service products in the evaluated enterprise, and the product data required by the service are prepared for users.
Further, the technology applied by the enterprise admission module comprises a rule engine and a threshold system; the rule engine comprises a feature library design mechanism, a rule matching optimization mechanism and a rule judgment mechanism; the threshold system comprises an expert threshold and a dynamic threshold based on the behavior of the enterprise to be evaluated; the expert threshold is a static threshold based on multi-dimensional indexes, and corresponding wind control measures are executed on enterprises and enterprise legal representatives with obvious risks; the dynamic threshold based on the behavior of the evaluated enterprise is based on the behavior of the evaluated enterprise, one or more of historical tax payment indexes, financial statement indexes, business registration and change indexes of the evaluated enterprise are utilized, a cluster analysis model is adopted to carry out user classification and deep feature mining, different risk levels are distributed to various types of evaluated enterprises, and the dynamic threshold is determined according to the corresponding risk levels of the evaluated enterprises.
Further, fraud risk evaluation is carried out on the evaluated enterprise meeting the admission requirement through a built-in model of an enterprise anti-fraud module, the probability of fraud occurrence of the evaluated enterprise is determined, and a fraud risk score of the evaluated enterprise is formed, and the method specifically comprises the following steps: obtaining information which is beneficial to anti-fraud decision making from high-dimensional data through the enterprise anti-fraud module, training an anti-fraud model based on one or more machine learning algorithms in rule engine and enterprise indexes and anomaly detection, relationship maps and deep learning, carrying out fraud risk evaluation on the evaluated enterprise in real time, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise; the enterprise anti-fraud module comprises an access plate, a data storage plate, a fraud detection plate and a machine learning plate; the access plate provides API service access functions, including anti-fraud index access, anti-fraud rule engine analysis based on Rete algorithm, feature processing, model loading prediction and data precipitation; the data storage plate provides support for a relational database and a distributed database, the relational database stores configuration information and enterprise information of an anti-fraud rule engine, and the distributed database is used for storing model training characteristics after stream-oriented computation; the fraud identification rule in the fraud detection plate comprises one or more of a list detection index, a comprehensive detection index, an account detection index and a strategy monitoring index; the machine learning board block is used for improving the evaluation effect of the anti-fraud model so as to identify a new fraud risk.
Furthermore, the enterprise wind control system also comprises an enterprise internal credit scoring module; the enterprise internal credit rating module is an enterprise credit rating card formed based on expert rating and machine learning technology, and predicts the credit default probability of the rated enterprise to form enterprise credit rating.
One or more embodiments of the present specification provide an enterprise wind control device based on big data credit, including: the data processing unit is used for processing credit investigation data acquired by a plurality of data sources through the credit investigation data standardization module and determining a structured data item; the index determining unit is used for processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises; the result screening unit is used for automatically checking the credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module and screening the evaluated enterprises meeting the access requirements according to the checking result; and the risk evaluation unit is used for carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a model built in the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise and forming a fraud risk score of the evaluated enterprise.
One or more embodiments of the present specification provide an enterprise wind control device based on big data credit, including: at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to: processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module to determine a structured data item; processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises; automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result; and carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
One or more embodiments of the present specification provide a storage medium storing computer-executable instructions configured to: processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module to determine a structured data item; processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises; automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result; and carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
The embodiment of the specification adopts at least one technical scheme which can achieve the following beneficial effects: the standardization of credit investigation data of the small micro-enterprise, the indexing of the credit investigation data and the intelligent wind control of the small micro-enterprise are combined, the whole process of obtaining the credit line of the small micro-enterprise from mass data is opened, and the problem of the isolated island of the credit investigation data is fully solved; the method comprises the steps of obtaining enterprise credit investigation data from multiple sources and forming a standard enterprise credit investigation data warehouse, solving the problems that unified standard specifications are lacked among heterogeneous information systems, the degree of data publicity of government departments is low, an overall coordination mechanism is lacked, and the like, realizing interconnection and standardization management among credit investigation data, and being beneficial to a financial institution to know the default risk level of a user more comprehensively and accurately so as to give reasonable credit line; the credit risk evaluation cost is reduced, the transparency and the fairness of the credit auditing process are ensured, and the human operation risk is reduced.
Drawings
In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments described in the present specification, and for those skilled in the art, other drawings can be obtained according to the drawings without any creative effort. In the drawings:
fig. 1 is a flowchart of an enterprise wind control method based on big data credit according to one or more embodiments of the present disclosure;
FIG. 2 is a block diagram of an enterprise anti-fraud module according to one or more embodiments of the present disclosure;
FIG. 3 is a block diagram illustrating an internal business credit scoring module according to one or more embodiments of the present disclosure;
fig. 4 is a schematic structural diagram of an enterprise wind control device based on big data credit according to one or more embodiments of the present disclosure;
fig. 5 is a schematic structural diagram of an enterprise wind control device based on big data credit according to one or more embodiments of the present disclosure.
Detailed Description
The embodiment of the specification provides an enterprise wind control method, device, equipment and medium based on big data credit investigation.
In order to make those skilled in the art better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only a part of the embodiments of the present specification, and not all of the embodiments. All other embodiments obtained by a person of ordinary skill in the art based on the embodiments of the present specification without any creative effort shall fall within the protection scope of the present specification.
Risk control refers to the risk supervisor taking various measures and methods to eliminate or reduce the various possibilities of occurrence of a risk event, or to reduce the losses incurred when a risk event occurs. In the current risk control of enterprises, credit plays an increasingly important role, and meanwhile, the credit degree of the enterprise is also an important factor for risk supervision and judgment of the enterprise.
In the prior art, the small and micro enterprise wind control solution relies on a credit evaluation model of the small and micro enterprise to carry out credit evaluation, and depends on financial indexes more; financial information of small and micro enterprises is often difficult to obtain and has low reliability, and because operation conditions such as policy change, industry supervision fluctuation and the like are affected, financial institutions such as banks have great dilemma on wind control management of the small and micro enterprises. Based on this, a more accurate risk control scheme is needed for the supervisors of the existing enterprises.
The technical problems that information islanding phenomenon is caused due to insufficient information sharing degree among credit investigation data sources and a more accurate risk assessment control scheme is lacked can be solved through the scheme provided by the embodiment of the specification. The big data credit investigation refers to the massive, diversified and multidimensional credit data generated by an upstream data producer, which are acquired through collection and accumulation, the midstream credit investigation institution processes and processes the big data to form structured data with a utilization value, and a downstream information user evaluates the potential risk hazard which may occur after judging, evaluating and analyzing the credit data to form a final decision. The big data technology enables credit investigation data to be larger and larger in scale and wider in application range, resource sharing and supervision cooperation with government supervision departments are achieved, the defects in a traditional credit investigation system are overcome, and credit scoring under the condition of traditional data loss is made possible.
The technical solutions proposed by the embodiments of the present invention are described in detail below with reference to the accompanying drawings.
Fig. 1 is a flowchart of an enterprise wind control method based on big data credit according to an embodiment of the present disclosure. It should be noted that the enterprise wind control method is executed by an enterprise wind control system, wherein the enterprise wind control system mainly comprises a credit investigation data standardization module, a credit investigation data indexing module, an enterprise access module, an enterprise fraud prevention module and an internal credit scoring module. As shown in fig. 1, the enterprise wind control method based on big data credit investigation mainly includes the following steps:
and step S101, processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module, and determining a structured data item.
Further, in an embodiment of the present specification, the step S101 may specifically include:
acquiring credit investigation data through a credit investigation data standardization module, managing the credit investigation data through a pre-established enterprise credit investigation data warehouse, and determining required business data; and carrying out enterprise credit investigation report inquiry or enterprise credit evaluation on the business data to determine a corresponding structured data item. The credit investigation data acquisition is carried out through the credit investigation data standardization module, and the method specifically comprises the following steps: and carrying out enterprise credit investigation data warehouse construction, data source definition, data extraction, conversion cleaning and loading work through the credit investigation data standardization module.
In one embodiment of the present description, credit data is obtained from multiple data sources. Specifically, main sources of enterprise credit investigation data comprise enterprise industrial and commercial data, judicial data, tax data and reward and punishment information authorized by government; the system also comprises credit data, running data and money consumption data of enterprises and real control persons fed back by the cooperative financial institution; the system also comprises public sentiment data and credit public notice data collected by the Internet, and operator data and commercial transaction data cooperated by a third party. The method follows a new idea of establishing a novel credit system of 'all data are credit', acquires enterprise credit investigation data from various sources and forms a standard enterprise credit investigation data warehouse, solves the problems of lack of unified standard specification, low data publicity degree of government departments, lack of a coordination mechanism for overall planning and the like between heterogeneous information systems, and realizes interconnection and standardization management between credit investigation data.
In one embodiment of the present specification, the credit investigation data standardization module performs three levels of processing of data acquisition, data management and data application on the original enterprise credit data acquired by a plurality of data sources in the form of an enterprise credit investigation warehouse.
In an embodiment of the present specification, the data acquisition of the original enterprise credit data by the credit investigation data standardization module mainly includes enterprise credit investigation data warehouse construction, data source definition, data extraction, conversion cleaning and loading. Because the data sources of the original enterprise credit data are wide, and the related principles, architectures and use platforms of different systems are different, interface data, base table data and unstructured data need to be processed into structured data which is convenient to store and use; after the data source is determined, the quality of the data source needs to be controlled, and non-uniform and incomplete data sources are stored in a data warehouse according to a uniform standard through methods such as cross comparison, rule verification and the like.
In one embodiment of the present description, the credit investigation data standardization module extracts, transforms, cleans and loads (ETL) the original enterprise credit data. Specifically, an ETL strategy is established according to system characteristics, wherein the ETL strategy comprises the extraction frequency of data extraction, the granularity of data and the like, and the monitoring and tracking processing of the whole process is carried out after ETL is finished. In the data extraction stage, different updating modes are adopted for different forms of data; adopting a trigger updating mode aiming at the data in the interface form; aiming at data in a base table form, if the data volume is not large, such as public data, reward and punishment data and the like, a full-volume updating mode is adopted; if the data volume is large and the updating rule is regular, such as financial statement data, tax payment data and the like, an increment extraction mechanism based on the data timestamp is adopted. In the data conversion stage, the extracted original enterprise credit investigation data is converted into different data dimensions according to time, regions, industries and the like, and data quality verification is carried out by applying rules to eliminate data which does not meet quality requirements.
It should be noted that the enterprise credit investigation data warehouse comprises a source data layer, a verification data layer, a basic data layer and a product data layer; wherein the source data layer: the credit investigation system is used for uniformly and standardly processing and storing credit investigation data and updating the credit investigation data according to a preset strategy; the check data layer: the management mechanism is used for verifying the data content through a set data verification rule and a verification process, loading the verified data into the central database of the basic data layer, returning the data which does not pass the verification to the source data layer, and feeding the error data back to the corresponding data source after forming a data feedback message; the basic data layer: the data processing device is used for storing the data which passes the verification of the verification data layer, returning the error data checked out through the post verification to the verification data layer, and finally forming an error feedback message to be fed back to the reporting mechanism; the product data layer: the product data extraction logic and mining analysis are used for loading the product data into the product data layer according to the types of service products in the evaluated enterprise, and the product data required by the service are prepared for users.
In one embodiment of the present specification, the enterprise credit investigation data is managed by a pre-established enterprise credit investigation data warehouse, wherein credit investigation data from a plurality of data sources and integrated summary data for analysis are stored in the enterprise credit investigation data warehouse, and the enterprise credit investigation data warehouse comprises a source data layer, a verification data layer, a basic data layer and a product data layer in a hierarchical view.
The source data layer is the basis of data of other layers, and the source data layer carries out unified normalized processing and storage on the data from each data source and updates the data by applying a proper strategy; when the data of the source data layer arrives, the verification data layer verifies the data content by using a set data verification rule and a verification process, and if the data passes the verification, the data is loaded into a central database; and if the data does not pass the verification, returning the data to the previous layer, and feeding back error data which does not pass the verification to the management mechanism of the corresponding data source to form a data feedback message. The basic data layer mainly stores key credit investigation service data which passes the verification. Based on the consideration of performance, the verified key credit investigation business data is generally loaded to the central database in a batch loading mode. The central database is mainly used for storing data passing the verification, returning error data checked out through the verification to the previous layer, and finally forming an error feedback message to be fed back to the reporting mechanism. After the basic data are put in storage and arranged, the basic data are loaded to a product data layer according to the types of service products (credit reports, fixed reports, summary queries and the like), established data extraction logics and mining analysis, and business data required by service are prepared for users.
In one embodiment of the present specification, the data application refers to performing enterprise credit investigation report query or enterprise credit evaluation on business data to determine corresponding structured data items. It should be noted that the enterprise credit investigation report query and the enterprise credit evaluation are core services oriented to the enterprise credit investigation data warehouse. And establishing an enterprise credit investigation data warehouse, and after data management is carried out based on the enterprise credit investigation data warehouse, providing credit products which are mainly credit reports and data which can be used for supporting enterprise credit evaluation. The credit report provides service in an autonomous query mode, and the form of the credit report can be interface data or a credit report page; the enterprise credit rating index provides services in the form of an API data interface, and the output structured data items can be directly used for credit rating or reprocessed.
And S102, processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises.
In an embodiment of the present specification, the credit investigation data indexing module further processes and derives the structured data items output by the enterprise credit investigation data warehouse by using big data processing and machine learning techniques, so as to form a standardized enterprise credit investigation data index covering the dimensions of the basic plane, performance capability, repayment capability, growth capability, business capability, etc. of the evaluated enterprise. Three methods of index derivation (Recency, Frequency, Monetary, RFM), unsupervised clustering index derivation and index filtering are mainly used.
Specifically, the index-derived RFM may implement the segmentation of the client value by characterizing the business and performance behavior of the enterprise. In the scheme, the RFM framework is applied to the aspects of change rules, tax payment and invoicing behaviors, financial statement key index changes and the like of enterprises, and a characteristic derivative main body (statistical object) taking customers, accounts and equipment as a core is defined. The specific derivation method comprises the following steps: the statistical objects include customers, accounts, and transaction devices. The statistical window comprises a short-term window and a long-term window, wherein the short-term window is in hours, such as 1 hour, 24 hours and the like, and the maximum time is not more than 48 hours; the long-term window is in months, such as 1 month and 3 months, and the maximum is not more than 36 months. Aggregation functions include continuous functions and discrete functions, such as maximum, minimum, sum, mean, and discrete functions such as count, frequency, ratio, and the like. The statistical variables comprise continuous variables and discrete variables, wherein the continuous variables refer to variables which can be randomly valued in a certain interval, such as real tax payment and the like; the discrete variable refers to a variable with a limited fixed value and can be listed, such as an enterprise type and the like.
Specifically, the unsupervised clustering index derivation method is mainly based on the relationship and characteristics between indexes, finds out the characteristics that are not easy to be induced between high-dimensional indexes, and generates the indexes with high predictive distinguishability by applying a KMeans clustering algorithm. This process results in a large number of indicators that are not processed in the derivation stage and that are automatically screened for subsequent feature selection. It should be noted that different types of indicators have different derivation directions, and specific derivation ways include the following two ways: time series based index clustering and multi-dimensional portrait based index clustering. The index clustering based on the time series aims at enterprise credit investigation indexes with the time series, clustering is carried out according to the time dimension of the enterprise credit investigation indexes, and enterprise and time characteristics possibly with abnormal risks are mined; the index clustering based on the multi-dimensional portrait aims at the multi-dimensional indexes such as financial indexes and tax payment indexes of enterprises, clusters the index set on a specific time section, and marks corresponding service labels on the enterprises according to the clustering result to serve as new enterprise indexes.
Specifically, the target of index filtering is as follows: firstly, selecting indexes of a data set according to a certain rule, wherein the selected indexes have the characteristics of high predictability, high stability, high business interpretability and the like; and then, carrying out service classification on the screened indexes according to service experience to form the dimensionalities of the enterprise basic plane, the performance capability, the debt paying capability, the growth performance, the operation capability and the like. According to different index types, different methods are adopted, specifically, if the index is a continuous index, a variance filtering method is applied; if the index is a discrete index, the same value filtering method is applied. The index filtering step comprises the following steps: calculating a correlation coefficient between indexes, and deleting one of index groups with too high correlation coefficient; calculating a correlation coefficient between the index and the target variable, and deleting the index with the too small correlation coefficient; and performing index importance sorting by using an XGboost algorithm, and removing indexes with index importance smaller than a threshold value.
And S103, automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result.
In one embodiment of the present description, the techniques employed by the enterprise admission module include a rules engine and a threshold system; the rule engine comprises a feature library design mechanism, a rule matching optimization mechanism and a rule judgment mechanism; the threshold system comprises an expert threshold and a dynamic threshold based on the behavior of the enterprise to be evaluated; the expert threshold is a static threshold based on multi-dimensional indexes, and corresponding wind control measures are executed on enterprises and enterprise legal representatives with obvious risks; the dynamic threshold based on the behavior of the evaluated enterprise is based on the behavior of the evaluated enterprise, one or more of historical tax payment indexes, financial statement indexes, business registration and change indexes of the evaluated enterprise are utilized, a cluster analysis model is adopted to carry out user classification and deep feature mining, different risk levels are distributed to various types of evaluated enterprises, and the dynamic threshold is determined according to the corresponding risk levels of the evaluated enterprises.
Specifically, the enterprise access module performs first-step screening on enterprises entering the wind control process based on artificial intelligence automation rule collision, performs black and white list collision, false application identification, information counterfeiting identification, black product package excavation and the like in a real-time manner through rule judgment, and filters high-risk enterprises which do not accord with the access rules. It should be noted that the technology applied by the enterprise admission module includes a rule engine and a threshold system, wherein the rule engine includes a feature library design, a rule matching optimization and a rule judgment mechanism; the threshold system comprises expert thresholds and dynamic thresholds based on enterprise behavior.
In one embodiment of the present specification, the rules engine consists of an inference engine, a fact set, a rule base. Wherein the fact set refers to related data of the data layer, and the related data is loaded into the working memory. Variables needed in the rule base are stored as characteristic factors through preprocessing so as to improve the multiplexing rate of the variables and the simplification degree of the rules. In addition, according to the wind control requirement, the characteristic factors in the characteristic library are divided into user characteristic factors and global characteristic factors. The user characteristic factors take enterprise unified social credit codes as main keys, and the characteristic data of each dimension of the small and micro enterprises are aggregated; the global feature factors are combined and calculated from other dimensions required for abstracting the global data.
In one embodiment of the present specification, Rete algorithm is used in the rule pattern matching for improving matching efficiency and reducing temporal redundancy caused by repeated calculation. When the number of rules and fact samples is large, each fact data needs to be matched with an Alpha node in the Rete network. Therefore, through setting the pre-matching module, a plurality of rules are aggregated into a small number of rule sets, and part of normal data is filtered out in the pre-matching stage through rule set screening, so that the matching times of facts and nodes are reduced, and the rule matching efficiency is improved. It should be noted that, the implementation method is to divide the rules containing multiple identical conditional atoms into the same rule group, and the conditional atom with the largest occurrence number in the rule group is used as the characteristic condition of the rule group; the global data can filter partial data through screening of the rule group in the pre-matching module, and only the rule judgment in the rule group of the residual samples is executed.
In one embodiment of the present specification, an effective admission rule system includes identifying at risk users and factual risk interception measures. Therefore, the admission rule system needs to reduce the risk false alarm rate and the missing report rate to be within an acceptable range, and improves the air control effectiveness and the user experience. It should be noted that the rule evaluation mechanism is based on two data sources, namely, the triggering frequency distribution obtained according to the wind control score, and the final request result obtained by responding to the wind control measure after the rule is triggered. And calculating a precision rate (p) and a recall rate (r) based on the triggering times of each rule output by the rule engine, wherein p is TP/(TP + FP) and r is TP/(TP + FN), wherein TP represents the clients actually having defaults, FP represents the clients wrongly determined as defaults, and FN represents the clients wrongly determined as non-defaults. The system matches real-time wind control measures such as short message verification, manual electric core, on-site verification, application rejection and the like according to the value returned by each request.
In an embodiment of the present specification, a threshold system is designed to mainly solve the drawback caused by one-time threshold in the conventional wind control system, and a total of three modules are provided based on a scoring mechanism. On the basis of expert threshold values, an enterprise behavior scoring mechanism and dynamic threshold value adjustment are added. On one hand, the expert threshold is used for preliminary filtering, and corresponding wind control measures are executed on enterprises and legal representatives of the enterprises which obviously have risks based on the static threshold of the multi-dimensional index; the expert threshold is used for carrying out threshold determination on the single indexes one by one based on an expert consultation method, and has objectivity and representativeness. On the other hand, the dynamic threshold based on the enterprise behavior is a comprehensive method for dynamically adjusting the threshold based on the enterprise behavior, and the specific implementation mode is divided into the following three steps: firstly, based on enterprise behaviors, utilizing historical tax payment indexes, financial statement indexes, business registration and change indexes and the like of enterprises, and adopting models such as cluster analysis, random forests and the like to classify users and mine depth features; secondly, establishing an access risk evaluation system of the enterprise, calculating by using a model, and distributing different risk levels to various types of enterprises; and finally, calculating the sample by adopting the trained admission risk evaluation model, realizing high-availability personalized intelligent admission control, and improving the wind control processing efficiency.
And step S104, carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
In an embodiment of the present specification, the step S104 may specifically include:
obtaining information which is beneficial to anti-fraud decision making from high-dimensional data through the enterprise anti-fraud module, training an anti-fraud model based on one or more machine learning algorithms in rule engine and enterprise indexes and anomaly detection, relationship maps and deep learning, carrying out fraud risk evaluation on the evaluated enterprise in real time, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise; the enterprise anti-fraud module comprises an access plate, a data storage plate, a fraud detection plate and a machine learning plate; the access plate provides API service access functions, including anti-fraud index access, anti-fraud rule engine analysis based on Rete algorithm, feature processing, model loading prediction and data precipitation; the data storage plate provides support for a relational database and a distributed database, the relational database stores configuration information and enterprise information of an anti-fraud rule engine, and the distributed database is used for storing model training characteristics after stream-oriented computation; the fraud identification rule in the fraud detection plate comprises one or more of a list detection index, a comprehensive detection index, an account detection index and a strategy monitoring index; the machine learning board block is used for improving the evaluation effect of the anti-fraud model so as to identify a new fraud risk.
Specifically, information which is helpful for anti-fraud decision making is obtained from high-dimensional data, an anti-fraud model is trained based on machine learning algorithms such as a rule engine, enterprise indexes and anomaly detection, a relational graph and deep learning, a pre-loan anti-fraud machine learning model set based on enterprise behavior portraits is constructed, potential fraud risks of an enterprise to be evaluated are pre-judged in real time, and the accuracy of fraud risk detection is improved. Fig. 2 is a schematic structural diagram of an enterprise anti-fraud module provided by an embodiment of the present specification, where the enterprise anti-fraud module 200 includes: access panel 202, data storage panel 204, fraud detection panel 206, machine learning panel 208, and fraud risk handling panel 210. The various blocks of the enterprise anti-fraud module are described in detail below.
First, the access board 202 provides API service access functions, including anti-fraud indicator access, anti-fraud rule engine analysis based on Rete algorithm, feature processing, model loading prediction, and data precipitation. It should be noted that, in order to ensure real-time analysis performance, a monitoring function is also provided, and any problem found in the monitoring process calls a corresponding handler.
Secondly, the data storage module 204 provides relational database support and distributed database support; the relational data mainly stores configuration information of an anti-fraud rule engine, enterprise information and other data, the distributed database is used for storing model training characteristics after flow type calculation, provides photographing characteristic support for model training, stores models after batch updating, supports online prediction and model updating, and provides protection for the whole intelligent wind control system.
Then, the fraud detection module 206 performs fraud risk monitoring on the enterprise under evaluation, and the comprehensive fraud monitoring index determines the effectiveness of monitoring the operation condition of the anti-fraud system in real time. It should be noted that, the fraud monitoring index is processed in real time by a rule engine, and the enterprise which lives high-risk fraud rules is filtered. In one embodiment of the present specification, the fraud identification rules mainly include: list class detection indexes, comprehensive class detection indexes, account class detection indexes and strategy class monitoring indexes. The list detection indexes refer to collision of enterprises based on black and grey lists formed inside and outside, such as lost letters, multi-head loan, illegal violation, punishment, operators and the like, so that the enterprises with bad history are comprehensively detected; the comprehensive detection indexes focus on the overall control of the comprehensive credit condition of the enterprise, and the comprehensive detection indexes are used for carrying out statistics and analysis on data such as industry and commerce, management, basic surface, tax payment, management behaviors, historical loan behaviors and the like, abstracting the characteristic overall view of a single user or a group of users and establishing a client behavior panoramic picture; the account detection indexes focus on account grading monitoring, and account operation is detected by analyzing differences of transaction amount, account authority and account opening modes of bank accounts; strategy monitoring indexes are as follows: the method focuses on paying attention to the anti-fraud strategy and the triggering condition of the rule in real time, such as the interception rate, the hit rate, the rule response duration and the like of the anti-fraud rule, so as to achieve the purpose of analyzing and verifying the effectiveness of the model.
Secondly, the machine learning board 208 adopts a mode of combining real-time feature extraction and offline model training to improve the evaluation effect of the anti-fraud model and identify a new fraud risk as early as possible. The real-time feature extraction is used for processing the features of the real-time data stream through a stream type calculation engine, processing data according to model feature logic, storing the features and providing feature data for offline model training; the off-line model training periodically reads characteristic data in batches through a timing scheduling strategy, automatically trains the model, and updates the trained model. In one embodiment of the present description, a machine learning plate-supported model includes: unsupervised learning enterprise anomaly detection, intelligent anti-fraud association maps and supervised machine learning fraud probability prediction. The unsupervised learning enterprise anomaly detection abstracts the characteristic overall appearance of a single enterprise or a group of enterprises through an unsupervised clustering analysis algorithm, can quickly measure the similarity degree of an evaluated enterprise and a real non-fraudulent enterprise in fraud detection, provides a big data basis for judging enterprise fraud risk, and further moves a risk identification link; the intelligent anti-fraud association map builds an association map network based on the relations of business addresses, contacts, contact calls, investment and financing, guarantee and the like, defines the relation between network fixed points by a similarity algorithm by combining a user portrait model and a characteristic engineering technology, builds an enterprise map network, finds the association relation between related attribute sets in a large amount of data, mines potential fraud groups, provides reference basis for rule formulation, and optimizes intelligent anti-fraud static protection rules; the supervised machine learning fraud probability prediction is to finish anomaly detection labeling based on historical real enterprise fraud performance data and an anomaly detection model, and adopt a strategy of fusing a plurality of algorithms on the basis of fully excavating fraud samples in deep network fraud detection, wherein the strategies comprise machine learning algorithms such as logistic regression and support vector machines, and deep learning algorithms such as convolutional neural networks and multilayer fully-connected neural networks, so that the model effect is gradually optimal, and the risk prediction is more accurate.
Finally, in fraud risk handling block 210, the handling policies and rules for fraud risk may be used to clarify the acceptable level and manner of handling of fraud risk. It should be noted that, the fraud handling manner adopted in an embodiment of the present specification is determined according to the fraud risk level, specifically: for the high-level suspected fraud risk, intercepting and blocking in real time by formulating an anti-fraud strategy; for the suspected fraud risk of the middle level, the risk level is balanced by sacrificing the user experience, the risk level is reduced, and a manual review link is added; for low-level suspected fraud risk, reminding the customer by means of outgoing call, short message reminding and the like; and manual investigation and analysis are carried out on the factual fraud clues and cases fed back by each channel.
In an embodiment of the present specification, in step S104, fraud risk evaluation is performed on the evaluated enterprise meeting the admission requirement through a model built in an enterprise fraud prevention module, a probability of fraud occurring in the evaluated enterprise is determined, and a fraud risk score of the evaluated enterprise is formed, and then, an enterprise internal credit scoring module is an enterprise credit scoring card formed based on expert scoring and machine learning technology, and a credit default probability of the evaluated enterprise is predicted, so as to form an enterprise credit rating.
Fig. 3 is a schematic structural diagram of an internal enterprise credit scoring module according to an embodiment of the present disclosure, and the internal enterprise credit scoring module is described below with reference to fig. 3.
In one embodiment of the present description, the enterprise internal credit scoring module 300 includes a feature screening tile 302, a sample equalization tile 304, a base classifier pool tile 306, and a credit prediction module 308.
Specifically, in one embodiment of the subject specification, a feature filter panel 302 is developed based upon the primary factors that affect enterprise credit. The feature screening plate 302 adopts a machine learning automated screening method to analyze the significance of the influence of each factor on the enterprise credit, a perfect and stable index system is formed, in order to improve the model prediction efficiency and the generalization capability of the model on unknown data, the feature screening plate 302 selects beneficial features from a sample space, abandons irrelevant or redundant features, and performs feature screening mainly through a four-step feature screening scheme: firstly, aiming at the credit characteristics of an original 2000-dimensionality enterprise, carrying out single-characteristic coarse-grained screening based on WOE; secondly, automatically screening the credit features of the 500-dimension enterprises screened in the first step based on XGboost and a random forest algorithm; thirdly, aiming at the 100-dimentional enterprise credit features screened in the second step, iterative feature screening based on Recursive Feature Elimination (RFE) is carried out; and fourthly, screening the credit characteristics of the 50-dimensional enterprise screened in the third step based on the characteristics of deep business understanding.
In one embodiment of the present description, the sample equalization plate 304 employs a sample synthesis based SMOTE oversampling technique to balance the number of two types of samples by constructing a new few types of samples. Because the SMOTE algorithm has a poor effect on high-dimensional data sets, the indexes need to be screened first, and then the samples need to be expanded. One embodiment of the present specification is directed to a problem that the number of valid samples in the enterprise credit scoring database is insufficient, and a SMOTE algorithm is used to generate a simulation sample for increasing the capacity of the valid samples in a sample set, so as to optimize an original sample set.
In an embodiment of the present specification, with the establishment of a high-accuracy enterprise credit scoring model as a target, a base classifier pool is established to form a base classifier pool plate block 306, the accuracy, the difference and the false positive loss rate are respectively used as selection criteria, the complementarity and the difference between the base classifiers are considered, a global search algorithm is designed to output an optimal base classifier, a proper method is selected as a fusion method of the base classifiers, and the output of an optimal combination model is finally realized. The key technology applied by the base classifier pool plate block 306 comprises automatic classifier screening and classifier fusion, and the implementation process is as follows: first, classifier filtering criteria are determined, and in the present embodiment, the criteria of classifier filtering include classification accuracy of the classifier itself, hypothesis conditions, and principles of the classifier. In the calculation process, the larger the difference of the base classifier is, the better the fitting effect of the model is, the stronger the generalization capability is and the lower the noise influence is, so that the classification precision and the algorithm rule are used as the screening standard in the classifier selection process. It should be noted that, in the embodiments of the present specification, the classifiers in the base classifier pool include a logistic regression LR, a support vector machine SVM, a multilayer fully-connected neural network MLP, an XGBoost, a random forest, and the like. Secondly, fusing classifiers; in the process of fusing the classifiers, the basic standard of selection is the precision and different scores of the classifiersAnd a mode of similarity difference fusion, namely performing method fusion by using a selective integration model. The classifier fusion process processes the classifier fusion process using the Stacking method for verifying the validity of the classifier selection. The Stacking fusion method comprises the following specific steps: the training set D is first broken down into k similarly sized but mutually disjoint subsets D1,D2,D3,…,Dk(ii) a Secondly, order
In that
Training a base classifier to classify DjAs a test set, obtain LjAt DjOutput of
Deriving k basis classifiers and k corresponding outputs
The k outputs plus the original class labels form a new training set Dn(ii) a At DnAnd training the final classifier L, wherein the output of the L is the final result.
In an embodiment of the present specification, the pre-credit calculation module 308 generates a credit score of the evaluated enterprise with a score interval of 0 to 100 points for the default probability of the evaluated enterprise, and corresponds to a certain credit level when the default probability of the evaluated enterprise is within a certain threshold interval according to the principle that the higher the credit level is, the smaller the default probability is.
Specifically, taking the average tax intake of the enterprise in the last two years as a basic quota, and then converting the credit level of the enterprise to be evaluated into a quota adjustment coefficient to obtain the customer rating; when the client is rated as AAA and AA, the credit line is expressed as the average value of the gross taxes of the enterprise in the last two years 10; when the customer rating is A, the credit line is expressed as the average value of the gross taxes of the enterprise in the last two years, 8; when the customer is rated as BBB, the credit line is expressed as the average value of the gross taxes of the enterprise in the last two years 5; when the customer is rated as BB, the credit line is expressed as the average value of the gross taxes of the enterprise in the last two years, namely 2; when the customer rating is B and below, the credit line is 0.
Fig. 4 is a schematic structural diagram of an enterprise wind control device based on big data credit according to an embodiment of the present disclosure. Enterprise wind control device based on big data credit includes: the data processing unit 402 is configured to process credit investigation data acquired by multiple data sources through a credit investigation data standardization module, and determine a structured data item; the index determining unit 404 is configured to process the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises; the result screening unit 406 is used for automatically checking credit investigation data indexes corresponding to the evaluated enterprises through the enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result; and the risk scoring unit 408 is used for performing fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a model built in the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
Fig. 5 is a schematic structural diagram of an enterprise wind control device based on big data credit according to an embodiment of the present disclosure. Enterprise wind control equipment based on big data credit investigation includes: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to: processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module to determine a structured data item; processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises; automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result; and carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
One embodiment of the present specification provides a storage medium storing computer-executable instructions, wherein the computer-executable instructions are configured to: processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module to determine a structured data item; processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises; automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result; and carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
The technical scheme provided by one or more embodiments of the specification combines standardization of credit investigation data of the small micro-enterprise, indexing of the credit investigation data and intelligent wind control of the small micro-enterprise based on the intelligent wind control solution of the small micro-enterprise for big data credit investigation, and the whole process of obtaining the credit line of the small micro-enterprise from mass data is opened, so that the problem of islanding of the credit investigation data is fully solved, and the purpose of solving the problem of difficult financing of the small micro-enterprise by using a big data technology is achieved. Has the following beneficial effects: following a new idea of establishing a novel credit system of 'all data are credit', acquiring enterprise credit investigation data from multiple sources and forming a standard enterprise credit investigation data warehouse, solving the problems of lack of unified standard specification, low data publicity degree of government departments, lack of an overall coordination mechanism and the like among heterogeneous information systems, and realizing interconnection and standardization management among credit investigation data; based on automatic feature processing, a rule engine and a machine learning algorithm, rich enterprise credit indexes, access verification systems and anti-fraud and credit scoring models are constructed; by means of technologies such as big data and real-time data processing, a set of intelligent small and micro enterprise business wind control schemes are formed through practices in links such as systematic deployment, application strategies and model iterative optimization; the risk control analysis is carried out by comprehensively applying the big data technology, the impact of the risks such as default, fraud and the like caused by information asymmetry on the financing ecology of small and micro enterprises is solved, and the financial institutions are helped to know the default risk level of the users more comprehensively and accurately, so that reasonable credit line is given; in addition, the technical scheme provided by the specification is realized automatically, the dispatching and credit risk evaluation cost of financial institutions is reduced, the service efficiency of financing of small and micro enterprises is improved, the transparency and fairness of the credit auditing process are ensured, and the risk of manual operation is reduced.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the embodiments of the apparatus, the device, and the nonvolatile computer storage medium, since they are substantially similar to the embodiments of the method, the description is simple, and for the relevant points, reference may be made to the partial description of the embodiments of the method.
The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.
The above description is merely one or more embodiments of the present disclosure and is not intended to limit the present disclosure. Various modifications and alterations to one or more embodiments of the present description will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement or the like made within the spirit and principle of one or more embodiments of the present specification should be included in the scope of the claims of the present specification.
Claims (10)
1. An enterprise wind control method based on big data credit investigation is characterized in that the method is executed by an enterprise wind control system, the enterprise wind control system comprises a credit investigation data standardization module, a credit investigation data indexing module, an enterprise access module, an enterprise anti-fraud module and an internal credit scoring module, and the method comprises the following steps:
processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module to determine a structured data item;
processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises;
automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result;
and carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
2. The method according to claim 1, wherein the step of processing credit investigation data acquired from a plurality of data sources by the credit investigation data standardization module to determine the structured data item specifically comprises:
acquiring credit investigation data through a credit investigation data standardization module, managing the credit investigation data through a pre-established enterprise credit investigation data warehouse, and determining required business data;
and carrying out enterprise credit investigation report inquiry or enterprise credit evaluation on the business data to determine a corresponding structured data item.
3. The method according to claim 2, wherein the obtaining of the credit investigation data through the credit investigation data standardization module specifically comprises:
and the credit investigation data standardization module is used for carrying out the work of enterprise credit investigation data warehouse construction, data source definition, data extraction, conversion cleaning and loading.
4. The method of claim 2, wherein the enterprise credit investigation data repository comprises a source data layer, a verification data layer, a base data layer, and a product data layer; wherein the content of the first and second substances,
the source data layer: the credit investigation system is used for uniformly and standardly processing and storing credit investigation data and updating the credit investigation data according to a preset strategy;
the check data layer: the management mechanism is used for verifying the data content through a set data verification rule and a verification process, loading the verified data into the central database of the basic data layer, returning the data which does not pass the verification to the source data layer, and feeding the error data back to the corresponding data source after forming a data feedback message;
the basic data layer: the data processing device is used for storing the data which passes the verification of the verification data layer, returning the error data checked out through the post verification to the verification data layer, and finally forming an error feedback message to be fed back to the reporting mechanism;
the product data layer: the product data extraction logic and mining analysis are used for loading the product data into the product data layer according to the types of service products in the evaluated enterprise, and the product data required by the service are prepared for users.
5. The method of claim 1, wherein the technologies employed by the enterprise admission module include a rule engine and a threshold hierarchy; wherein the content of the first and second substances,
the rule engine comprises a feature library design mechanism, a rule matching optimization mechanism and a rule judgment mechanism;
the threshold system comprises an expert threshold and a dynamic threshold based on the behavior of the enterprise to be evaluated;
the expert threshold is a static threshold based on multi-dimensional indexes, and corresponding wind control measures are executed on enterprises and enterprise legal representatives with obvious risks;
the dynamic threshold based on the behavior of the evaluated enterprise is based on the behavior of the evaluated enterprise, one or more of historical tax payment indexes, financial statement indexes, business registration and change indexes of the evaluated enterprise are utilized, a cluster analysis model is adopted to carry out user classification and deep feature mining, different risk levels are distributed to various types of evaluated enterprises, and the dynamic threshold is determined according to the corresponding risk levels of the evaluated enterprises.
6. The method according to claim 1, wherein the fraud risk assessment is performed on the evaluated enterprise meeting the admission requirement through a model built in an enterprise fraud prevention module, the probability of fraud occurrence of the evaluated enterprise is determined, and a fraud risk score of the evaluated enterprise is formed, and specifically includes:
obtaining information which is beneficial to anti-fraud decision making from high-dimensional data through the enterprise anti-fraud module, training an anti-fraud model based on one or more machine learning algorithms in rule engine and enterprise indexes and anomaly detection, relationship maps and deep learning, carrying out fraud risk evaluation on the evaluated enterprise in real time, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise;
the enterprise anti-fraud module comprises an access plate, a data storage plate, a fraud detection plate and a machine learning plate;
the access plate provides API service access functions, including anti-fraud index access, anti-fraud rule engine analysis based on Rete algorithm, feature processing, model loading prediction and data precipitation;
the data storage plate provides support for a relational database and a distributed database, the relational database stores configuration information and enterprise information of an anti-fraud rule engine, and the distributed database is used for storing model training characteristics after stream-oriented computation;
the fraud identification rule in the fraud detection plate comprises one or more of a list detection index, a comprehensive detection index, an account detection index and a strategy monitoring index;
the machine learning board block is used for improving the evaluation effect of the anti-fraud model so as to identify a new fraud risk.
7. The method of claim 1, wherein the enterprise wind control system further comprises an enterprise internal credit scoring module;
the enterprise internal credit rating module is an enterprise credit rating card formed based on expert rating and machine learning technology, and predicts the credit default probability of the rated enterprise to form enterprise credit rating.
8. An enterprise wind control device based on big data credit investigation, characterized in that the device comprises:
the data processing unit is used for processing credit investigation data acquired by a plurality of data sources through the credit investigation data standardization module and determining a structured data item;
the index determining unit is used for processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises;
the result screening unit is used for automatically checking the credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module and screening the evaluated enterprises meeting the access requirements according to the checking result;
and the risk evaluation unit is used for carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a model built in the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise and forming a fraud risk score of the evaluated enterprise.
9. An enterprise wind control device based on big data credit investigation comprises:
at least one processor; and the number of the first and second groups,
a memory communicatively coupled to the at least one processor; wherein the content of the first and second substances,
the memory stores instructions executable by the at least one processor to enable the at least one processor to:
processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module to determine a structured data item;
processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises;
automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result;
and carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
10. A storage medium storing computer-executable instructions, the computer-executable instructions configured to:
processing credit investigation data acquired by a plurality of data sources through a credit investigation data standardization module to determine a structured data item;
processing the structured data items through a credit investigation data indexing module to form credit investigation data indexes corresponding to the evaluated enterprises;
automatically checking credit investigation data indexes corresponding to the evaluated enterprises through an enterprise access module, and screening the evaluated enterprises meeting the access requirements according to the checking result;
and carrying out fraud risk evaluation on the evaluated enterprise meeting the admission requirement through a built-in model of the enterprise anti-fraud module, determining the probability of fraud occurrence of the evaluated enterprise, and forming a fraud risk score of the evaluated enterprise.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110106549.9A CN112668944A (en) | 2021-01-26 | 2021-01-26 | Enterprise wind control method, device, equipment and medium based on big data credit investigation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110106549.9A CN112668944A (en) | 2021-01-26 | 2021-01-26 | Enterprise wind control method, device, equipment and medium based on big data credit investigation |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112668944A true CN112668944A (en) | 2021-04-16 |
Family
ID=75415930
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110106549.9A Pending CN112668944A (en) | 2021-01-26 | 2021-01-26 | Enterprise wind control method, device, equipment and medium based on big data credit investigation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112668944A (en) |
Cited By (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113283583A (en) * | 2021-05-18 | 2021-08-20 | 广州致景信息科技有限公司 | Method and device for predicting default rate of textile industry, storage medium and processor |
| CN113362072A (en) * | 2021-06-30 | 2021-09-07 | 平安普惠企业管理有限公司 | Wind control data processing method and device, electronic equipment and storage medium |
| CN113537807A (en) * | 2021-07-27 | 2021-10-22 | 天元大数据信用管理有限公司 | Enterprise intelligent wind control method and device |
| CN113535809A (en) * | 2021-06-24 | 2021-10-22 | 深圳市邦盛实时智能技术有限公司 | Distributed multi-kernel decision making system and method for credit granting industry |
| CN113657747A (en) * | 2021-08-12 | 2021-11-16 | 中国安全生产科学研究院 | Enterprise safety production standardization level intelligent evaluation system |
| CN113822542A (en) * | 2021-08-30 | 2021-12-21 | 天元大数据信用管理有限公司 | Enterprise credit investigation platform construction method based on government affair big data |
| CN113837862A (en) * | 2021-09-27 | 2021-12-24 | 天元大数据信用管理有限公司 | Post-credit risk early warning method, device and medium based on credit investigation |
| CN113837885A (en) * | 2021-09-27 | 2021-12-24 | 上海欣方智能系统有限公司 | Construction method of financial anti-fraud service database and financial anti-fraud service system |
| CN113873088A (en) * | 2021-10-29 | 2021-12-31 | 平安科技（深圳）有限公司 | Voice call interaction method and device, computer equipment and storage medium |
| CN113887987A (en) * | 2021-10-15 | 2022-01-04 | 重庆葵海数字科技有限公司 | Enterprise operation risk assessment method |
| CN114091902A (en) * | 2021-11-22 | 2022-02-25 | 支付宝(杭州)信息技术有限公司 | Risk prediction model training method and device, and risk prediction method and device |
| CN115131139A (en) * | 2022-09-02 | 2022-09-30 | 创新奇智(南京)科技有限公司 | Method, device and medium for obtaining target result based on structural data |
| CN115907835A (en) * | 2022-12-30 | 2023-04-04 | 深度(山东)数字科技集团有限公司 | Big data wind control and assistant decision analysis method based on commercial draft information |
| CN116702052A (en) * | 2023-08-02 | 2023-09-05 | 云南香农信息技术有限公司 | Community social credit system information processing system and method |
| CN117591530A (en) * | 2024-01-17 | 2024-02-23 | 杭银消费金融股份有限公司 | Data cross section processing method and system |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108564286A (en) * | 2018-04-19 | 2018-09-21 | 天合泽泰（厦门）征信服务有限公司 | A kind of artificial intelligence finance air control credit assessment method and system based on big data reference |
| CN108629686A (en) * | 2018-05-09 | 2018-10-09 | 国家计算机网络与信息安全管理中心 | Internet financial company reference risk analysis method based on big data and system |
| US20190197442A1 (en) * | 2017-12-27 | 2019-06-27 | Accenture Global Solutions Limited | Artificial intelligence based risk and knowledge management |
| CN110458697A (en) * | 2019-08-19 | 2019-11-15 | 北京百度网讯科技有限公司 | Method and apparatus for assessing risk |
| CN110727922A (en) * | 2019-10-11 | 2020-01-24 | 集奥聚合(北京)人工智能科技有限公司 | Anti-fraud decision model construction method based on multi-dimensional data flow |
-
2021
- 2021-01-26 CN CN202110106549.9A patent/CN112668944A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190197442A1 (en) * | 2017-12-27 | 2019-06-27 | Accenture Global Solutions Limited | Artificial intelligence based risk and knowledge management |
| CN108564286A (en) * | 2018-04-19 | 2018-09-21 | 天合泽泰（厦门）征信服务有限公司 | A kind of artificial intelligence finance air control credit assessment method and system based on big data reference |
| CN108629686A (en) * | 2018-05-09 | 2018-10-09 | 国家计算机网络与信息安全管理中心 | Internet financial company reference risk analysis method based on big data and system |
| CN110458697A (en) * | 2019-08-19 | 2019-11-15 | 北京百度网讯科技有限公司 | Method and apparatus for assessing risk |
| CN110727922A (en) * | 2019-10-11 | 2020-01-24 | 集奥聚合(北京)人工智能科技有限公司 | Anti-fraud decision model construction method based on multi-dimensional data flow |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113283583A (en) * | 2021-05-18 | 2021-08-20 | 广州致景信息科技有限公司 | Method and device for predicting default rate of textile industry, storage medium and processor |
| CN113535809A (en) * | 2021-06-24 | 2021-10-22 | 深圳市邦盛实时智能技术有限公司 | Distributed multi-kernel decision making system and method for credit granting industry |
| CN113362072B (en) * | 2021-06-30 | 2023-09-08 | 成都一蟹科技有限公司 | Wind control data processing method and device, electronic equipment and storage medium |
| CN113362072A (en) * | 2021-06-30 | 2021-09-07 | 平安普惠企业管理有限公司 | Wind control data processing method and device, electronic equipment and storage medium |
| CN113537807A (en) * | 2021-07-27 | 2021-10-22 | 天元大数据信用管理有限公司 | Enterprise intelligent wind control method and device |
| CN113537807B (en) * | 2021-07-27 | 2024-01-23 | 天元大数据信用管理有限公司 | Intelligent wind control method and equipment for enterprises |
| CN113657747A (en) * | 2021-08-12 | 2021-11-16 | 中国安全生产科学研究院 | Enterprise safety production standardization level intelligent evaluation system |
| CN113822542A (en) * | 2021-08-30 | 2021-12-21 | 天元大数据信用管理有限公司 | Enterprise credit investigation platform construction method based on government affair big data |
| CN113837885A (en) * | 2021-09-27 | 2021-12-24 | 上海欣方智能系统有限公司 | Construction method of financial anti-fraud service database and financial anti-fraud service system |
| CN113837862A (en) * | 2021-09-27 | 2021-12-24 | 天元大数据信用管理有限公司 | Post-credit risk early warning method, device and medium based on credit investigation |
| CN113887987A (en) * | 2021-10-15 | 2022-01-04 | 重庆葵海数字科技有限公司 | Enterprise operation risk assessment method |
| CN113873088A (en) * | 2021-10-29 | 2021-12-31 | 平安科技（深圳）有限公司 | Voice call interaction method and device, computer equipment and storage medium |
| CN113873088B (en) * | 2021-10-29 | 2023-08-15 | 平安科技（深圳）有限公司 | Interactive method and device for voice call, computer equipment and storage medium |
| CN114091902A (en) * | 2021-11-22 | 2022-02-25 | 支付宝(杭州)信息技术有限公司 | Risk prediction model training method and device, and risk prediction method and device |
| CN115131139A (en) * | 2022-09-02 | 2022-09-30 | 创新奇智(南京)科技有限公司 | Method, device and medium for obtaining target result based on structural data |
| CN115907835A (en) * | 2022-12-30 | 2023-04-04 | 深度(山东)数字科技集团有限公司 | Big data wind control and assistant decision analysis method based on commercial draft information |
| CN116702052A (en) * | 2023-08-02 | 2023-09-05 | 云南香农信息技术有限公司 | Community social credit system information processing system and method |
| CN116702052B (en) * | 2023-08-02 | 2023-10-27 | 云南香农信息技术有限公司 | Community social credit system information processing system and method |
| CN117591530A (en) * | 2024-01-17 | 2024-02-23 | 杭银消费金融股份有限公司 | Data cross section processing method and system |
| CN117591530B (en) * | 2024-01-17 | 2024-04-19 | 杭银消费金融股份有限公司 | Data cross section processing method and system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112668944A (en) | Enterprise wind control method, device, equipment and medium based on big data credit investigation | |
| US11875355B2 (en) | Fast access vectors in real-time behavioral profiling in fraudulent financial transactions | |
| Patil et al. | Predictive modelling for credit card fraud detection using data analytics | |
| US11734692B2 (en) | Data breach detection | |
| Ogwueleka | Data mining application in credit card fraud detection system | |
| US20190164181A1 (en) | Reducing false positives with transaction behavior forecasting | |
| US20190279218A1 (en) | Behavior tracking smart agents for artificial intelligence fraud protection and management | |
| Gadi et al. | Credit card fraud detection with artificial immune system | |
| US20150046332A1 (en) | Behavior tracking smart agents for artificial intelligence fraud protection and management | |
| CN112767136A (en) | Credit anti-fraud identification method, credit anti-fraud identification device, credit anti-fraud identification equipment and credit anti-fraud identification medium based on big data | |
| CN111461216A (en) | Case risk identification method based on machine learning | |
| Ruyu et al. | A comparison of credit rating classification models based on spark-evidence from lending-club | |
| CN110728301A (en) | Credit scoring method, device, terminal and storage medium for individual user | |
| CN113537807A (en) | Enterprise intelligent wind control method and device | |
| García-Vico et al. | Fepds: A proposal for the extraction of fuzzy emerging patterns in data streams | |
| CN117455417B (en) | Automatic iterative optimization method and system for intelligent wind control approval strategy | |
| CN113487241A (en) | Method, device, equipment and storage medium for classifying enterprise environment-friendly credit grades | |
| Torres et al. | A proposal for online analysis and identification of fraudulent financial transactions | |
| CN105930430A (en) | Non-cumulative attribute based real-time fraud detection method and apparatus | |
| Kirkos et al. | Audit‐firm group appointment: an artificial intelligence approach | |
| KR102499182B1 (en) | Loan regular auditing system using artificia intellicence | |
| Marevac et al. | Decision-making AI for customer worthiness and viability | |
| Danenas | Intelligent financial fraud detection and analysis: a survey of recent patents | |
| SirElkhatim et al. | Prediction of banks financial distress | |
| US11328301B2 (en) | Online incremental machine learning clustering in anti-money laundering detection |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210416 |
|
| RJ01 | Rejection of invention patent application after publication |