CN115345727B - Method and device for identifying fraudulent loan application - Google Patents
Method and device for identifying fraudulent loan application Download PDFInfo
- Publication number
- CN115345727B CN115345727B CN202211000110.9A CN202211000110A CN115345727B CN 115345727 B CN115345727 B CN 115345727B CN 202211000110 A CN202211000110 A CN 202211000110A CN 115345727 B CN115345727 B CN 115345727B
- Authority
- CN
- China
- Prior art keywords
- order
- representation
- vectorization
- vectorization representation
- fusion
- 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/36—Creation of semantic tools, e.g. ontology or thesauri
- G06F16/367—Ontology
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Computational Linguistics (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The invention provides a method and a device for identifying a fraudulent loan application, wherein the method for identifying the fraudulent loan application comprises the following steps: obtaining individual data of a loan application client, wherein the individual data is data required to be input by a pre-constructed fraud identification model; vectorizing the individual data; inquiring a pre-constructed loan information knowledge graph, and acquiring vectorization representation of neighbor nodes with preset orders in the loan information knowledge graph by a loan application client; based on the vectorization representation of the loan application client and the vectorization representation of the neighbor nodes with the preset order, processing by using a two-way long and short word memory model to obtain updated vectorization representation of the loan application client; obtaining a fusion vectorization representation of the update vectorization representation by using a multi-layer perception model; and acquiring fraud probability of the loan application client based on the fusion vectorization representation and the fraud identification model. The recognition rate of fraudulent loan applications can be improved.
Description
Technical Field
The invention relates to the technical field of financial risk control, in particular to a method and a device for identifying fraudulent loan application.
Background
With the development of economy, the fraud in the financial field is increasing, and the loss of banks and financial institutions reaches tens of billions to billions each year due to fraud, and a great deal of manpower and financial resources are required to be invested each year for detecting the financial fraud and preventing the financial fraud. Among them, loan application fraud is the most dominant fraud, which is the major factor in the losses of banks and financial institutions.
At present, when fraud detection (risk analysis) is performed on a loan application, a client related to a loan application client is obtained according to loan application client data and a client relationship graph through a predetermined wind control rule, a client node sequence is formed, feature vectors of all client nodes are respectively expanded through a skip-gram model and the client relationship graph, the matching degree with the wind control rule is obtained according to the expanded feature vectors, whether the loan application is a fraudulent loan application is determined according to the matching degree, and a wind control strategy of the loan application client is obtained. However, the method carries out fraud judgment aiming at individual data evaluation of loan application clients and related clients, the loan application clients and the related clients simply form a client node sequence, the data between the nodes are independent, and the correlation characteristics and the implicit characteristics between the loan application clients and the related clients can not be mined, so that the identification rate of fraudulent loan application is lower, and the wind control level of banks and financial institutions is reduced.
Disclosure of Invention
Accordingly, the present invention is directed to a method and apparatus for identifying fraudulent loan applications, so as to improve the identification rate of the fraudulent loan applications.
In a first aspect, an embodiment of the present invention provides a method for identifying fraudulent loan applications, including:
obtaining individual data of a loan application client, wherein the individual data is data required to be input by a pre-constructed fraud identification model;
vectorizing the individual data;
inquiring a pre-constructed loan information knowledge graph, and acquiring vectorization representation of neighbor nodes with preset orders in the loan information knowledge graph by a loan application client;
based on the vectorization representation of the loan application client and the vectorization representation of the neighbor nodes with the preset order, processing by using a two-way long and short word memory model to obtain updated vectorization representation of the loan application client;
processing each order vector in the updated vectorization representation by using a multi-layer perception model to obtain each order vectorization representation;
processing the forward fusion vectorization representation of the previous step and the vectorization representation of the current step by using a long and short word memory model to obtain the forward fusion vectorization representation of the current step; processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation;
Based on each zero-order negative fusion vectorization representation and each order positive fusion vectorization representation, obtaining a fusion vectorization representation;
and acquiring fraud probability of the loan application client based on the fusion vectorization representation and the fraud identification model.
With reference to the first aspect, an embodiment of the present invention provides a first possible implementation manner of the first aspect, where the vectorizing the individual data includes:
encoding the individual data by using an evidence weight method, wherein each content item in the individual data corresponds to one code;
and splicing codes corresponding to different content items to obtain the vectorized representation of the individual data.
With reference to the first aspect, an embodiment of the present invention provides a second possible implementation manner of the first aspect, where the querying a pre-constructed loan information knowledge graph, obtaining a vectorized representation of neighboring nodes of a preset order of the loan application client in the loan information knowledge graph, includes:
inquiring a pre-constructed loan information knowledge graph;
if the loan information knowledge graph comprises the loan application client, taking the loan application client as a current node, and acquiring vectorization representation of neighbor nodes with preset orders in side connection with the current node;
If the loan information knowledge graph does not contain the loan application client, adding the loan application client into a node of the loan information knowledge graph, and setting the side contact of the loan application client and other nodes according to a side construction strategy of the loan information knowledge graph;
and obtaining the vectorization representation of the neighbor nodes with preset orders in the side connection with the loan application client from the loan information knowledge graph with the side connection.
With reference to the second possible implementation manner of the first aspect, the embodiment of the present invention provides a third possible implementation manner of the first aspect, where the preset order is 3, and obtaining a vectorized representation of a neighbor node with a preset order that has an edge connected with the current node includes:
acquiring a node in side contact with the current node to obtain a first-order neighbor node of the current node;
acquiring a node in edge connection with the first-order neighbor node to obtain a second-order neighbor node of the current node;
acquiring a node in side contact with the second-order neighbor node, and acquiring a third-order neighbor node of the current node;
a vectorized representation of each level of neighbor nodes is obtained.
With reference to the first aspect, the first possible implementation manner of the first aspect, or any one of the first possible implementation manner to the third possible implementation manner of the first aspect, the embodiment of the present invention provides a fourth possible implementation manner of the first aspect, wherein the processing, by using a multi-layer perceptual model, each order vector in the updated vectorized representation to obtain each order vectorized representation includes:
processing the vectorized representation in the updated vectorized representation by using a multi-layer perception model to obtain a zeroth order vectorized representation;
based on an activation function and the updated vectorized representation, acquiring vectorized aggregate representations of neighbor nodes of each order;
and processing the vectorized aggregate representation of each order by utilizing a multi-layer perception model to obtain the vectorized representation of the order.
With reference to the fourth possible implementation manner of the first aspect, the embodiment of the present invention provides a fifth possible implementation manner of the first aspect, wherein the preset order is 3, and the processing is performed on the forward fusion vectorization representation of the previous order and the current order vectorization representation by using a long and short word memory model to obtain the forward fusion vectorization representation of the current order; and processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation, comprising:
Processing the zeroth order vectorization representation by using a long and short word memory model to obtain a zeroth order forward fusion vectorization representation;
processing the zero-order forward fusion vectorization representation and the first-order vectorization representation by using a long and short word memory model to obtain the first-order forward fusion vectorization representation;
processing the first-order forward fusion vectorization representation and the second-order vectorization representation by using a long and short word memory model to obtain the second-order forward fusion vectorization representation;
processing the second-order forward fusion vectorization representation and the third-order vectorization representation by using a long and short word memory model to obtain a third-order forward fusion vectorization representation;
processing the third-order vectorization representation by using a long and short word memory model to obtain a third-order negative fusion vectorization representation;
processing the third-order negative fusion vectorization representation and the second-order vectorization representation by using a long and short word memory model to obtain the second-order negative fusion vectorization representation;
processing the second-order negative fusion vectorization representation and the first-order vectorization representation by using a long and short word memory model to obtain the first-order negative fusion vectorization representation;
And processing the first-order negative fusion vectorization representation and the zeroth-order vectorization representation by using the long and short word memory model to obtain the zeroth-order negative fusion vectorization representation.
With reference to the first aspect, the first possible implementation manner of the first aspect, or any one of the first possible implementation manner to the third possible implementation manner of the first aspect, the embodiment of the present invention provides a sixth possible implementation manner of the first aspect, wherein the processing, based on the vectorized representation of the loan application client and the vectorized representation of the neighboring node of the preset order, using a two-way long and short word memory model, obtains an updated vectorized representation of the loan application client, includes:
for the vectorized representation of the neighbor nodes of each order, calculating the mean value of the vectorized representation of the neighbor nodes of the order;
constructing a vectorization sequence according to the vectorization representation of the loan application client and the average value of the vectorization representation of the neighbor nodes of each order;
and processing the vectorization sequence by using a two-way long and short word memory model to obtain the updated vectorization representation.
In a second aspect, an embodiment of the present invention further provides an apparatus for identifying a fraudulent loan application, including:
The data acquisition module is used for acquiring individual data of a loan application client, wherein the individual data is data required to be input by a pre-constructed fraud identification model;
the vectorization module is used for vectorizing the individual data;
the multi-information acquisition module is used for inquiring a pre-constructed loan information knowledge graph and acquiring vectorization representation of neighbor nodes with preset orders in the loan information knowledge graph of the loan application client;
the vectorization updating module is used for processing the vectorization representation of the loan application client and the vectorization representation of the neighbor node with the preset order by using a two-way long and short word memory model to obtain updated vectorization representation of the loan application client;
the vectorization fusion module is used for respectively processing each order vector in the updated vectorization representation by utilizing a multi-layer perception model to obtain each order vectorization representation; processing the forward fusion vectorization representation of the previous step and the vectorization representation of the current step by using a long and short word memory model to obtain the forward fusion vectorization representation of the current step; processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation; based on each zero-order negative fusion vectorization representation and each order positive fusion vectorization representation, obtaining a fusion vectorization representation;
And the probability generation module is used for acquiring the fraud probability of the loan application client based on the fusion vectorization representation and the fraud identification model.
In a third aspect, embodiments of the present application provide a computer device including a memory, a processor, and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the above method when executing the computer program.
In a fourth aspect, embodiments of the present application provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method described above.
According to the method and the device for identifying the fraudulent loan application, provided by the embodiment of the invention, the individual data of a loan application client is obtained, wherein the individual data is data which is required to be input by a pre-built fraudulent identification model; vectorizing the individual data; inquiring a pre-constructed loan information knowledge graph, and acquiring vectorization representation of neighbor nodes with preset orders in the loan information knowledge graph by a loan application client; based on the vectorization representation of the loan application client and the vectorization representation of the neighbor nodes with the preset order, processing by using a two-way long and short word memory model to obtain updated vectorization representation of the loan application client; processing each order vector in the updated vectorization representation by using a multi-layer perception model to obtain each order vectorization representation; processing the forward fusion vectorization representation of the previous step and the vectorization representation of the current step by using a long and short word memory model to obtain the forward fusion vectorization representation of the current step; processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation; based on each zero-order negative fusion vectorization representation and each order positive fusion vectorization representation, obtaining a fusion vectorization representation; and acquiring fraud probability of the loan application client based on the fusion vectorization representation and the fraud identification model. In this way, the updated vectorization representation of the neighbor node with the side relation with the loan application client is obtained, and the fusion vectorization representation is obtained based on the previous-order forward fusion vectorization representation, the next-order forward fusion vectorization representation and the current-order vectorization representation in the updated vectorization representation.
In order to make the above objects, features and advantages of the present invention more comprehensible, preferred embodiments accompanied with figures are described in detail below.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and therefore should not be considered as limiting the scope, and other related drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a method for identifying fraudulent loan applications, according to an embodiment of the invention;
FIG. 2 illustrates a schematic diagram of acquiring a fusion vectorized representation provided by an embodiment of the present invention;
FIG. 3 is a flowchart illustrating a method for identifying fraudulent loan applications, in accordance with an embodiment of the invention;
FIG. 4 is a schematic diagram of an apparatus for identifying fraudulent loan applications, according to an embodiment of the invention;
fig. 5 is a schematic structural diagram of a computer device 500 according to an embodiment of the present application.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, are intended to be within the scope of the present invention.
Financial risk prevention is mainly aimed at solving the problem of fraud loans (fraudulent loan applications), and therefore, customers who may not have repayment capability are also removed from the system in order to identify malicious fraud loan behaviors. Moreover, as fraud technology advances, the concealment of fraud increases, such that the identification rate of fraud based solely on the customer and the individual profile of the customer associated with the customer is becoming lower and the risk of concealment cannot be identified. In the embodiment of the invention, the individual data of each loan application client is analyzed to construct the loan information knowledge graph comprising each loan application client and the correlation, the information fusion processing is carried out on the client and the individual data of the client associated with the client by utilizing a multi-level information fusion technology based on a graph calculation method and the loan information knowledge graph, so that the associated characteristics and the hidden characteristics between the client and the associated client are mined, the fraud recognition is carried out based on the associated characteristics and the hidden characteristics, the hidden fraud risk can be realized, and the recognition accuracy of the fraudulent loan application is improved.
The embodiment of the invention provides a method and a device for identifying fraudulent loan application, which are described below through the embodiment.
FIG. 1 is a flow chart of a method for identifying fraudulent loan applications, according to an embodiment of the invention. As shown in fig. 1, the method includes:
in an embodiment of the present invention, as an alternative embodiment, the individual data includes, but is not limited to: liability information, credit history information, social relationship information, and historical loan information, wherein,
the liability information includes, but is not limited to: sum of the number of occurrences in the first time threshold, monthly income, and total amount of telephone charges in the second time threshold. As an alternative embodiment, the first time threshold is set to 12 months and the second time threshold is set to 5 months;
credit history information includes, but is not limited to: whether the identity card hits a p2p blacklist, the loan application times of all institutions within a third time threshold from the present, the loan application times of non-silver institutions within a fourth time threshold from the present, and the number of non-silver institutions of the loan application within a fourth time threshold from the present, wherein as an optional embodiment, the third time threshold is set to 3 months, and the fourth time threshold is set to 12 months;
Social relationship information includes, but is not limited to: the total number of contacts, the number of times of contact between the service personnel, the silent duration of the mobile phone and the use duration of the mobile phone in a fifth time threshold from the present, wherein the fifth time threshold is set to be 3 months as an optional embodiment;
historical loan information includes, but is not limited to: loan amount, partner merchant type, merchant level, and rate of payment.
102, vectorizing the individual data;
in an embodiment of the present invention, as an optional embodiment, the vectorizing the individual data includes:
encoding the individual data by using a evidence weight (WOE, weight of Evidence) method, wherein each content item in the individual data corresponds to a code;
in the embodiment of the present invention, for example, for the liability information as above, three content items are included, which are respectively: sum of the number of occurrences in the first time threshold, monthly income, and total amount of telephone charges in the second time threshold.
And splicing codes corresponding to different content items to obtain the vectorized representation of the individual data.
In the embodiment of the invention, WOE encoding is carried out on each piece of information contained in the individual data to obtain the vectorized representation of the loan application client.
In the embodiment of the invention, the vectorization of the loan application client i is expressed as follows:
n i ＝[w i1 ,w i2 ,...,w ik ,...w in ]
wherein, the liquid crystal display device comprises a liquid crystal display device,
w ik WOE codes corresponding to the kth content item of the loan application client i, and n is the number of the content items.
in the embodiment of the present invention, as an optional embodiment, querying a pre-constructed loan information knowledge graph, and obtaining a vectorized representation of neighbor nodes of a preset order of the loan application client in the loan information knowledge graph, where the vectorized representation includes:
inquiring a pre-constructed loan information knowledge graph;
if the loan information knowledge graph comprises the loan application client, taking the loan application client as a current node, and acquiring vectorization representation of neighbor nodes with preset orders in side connection with the current node;
if the loan information knowledge graph does not contain the loan application client, adding the loan application client into a node of the loan information knowledge graph, and setting the side contact of the loan application client and other nodes according to a side construction strategy of the loan information knowledge graph;
And obtaining the vectorization representation of the neighbor nodes with preset orders in the side connection with the loan application client from the loan information knowledge graph with the side connection.
In the embodiment of the invention, each node in the loan information knowledge graph corresponds to a loan application client, the attribute of the node is vectorized representation, and the nodes are connected through edges.
In an embodiment of the present invention, as an alternative embodiment, the edge construction policy includes, but is not limited to: the nodes have communication connection, common guarantors, common identification numbers and common loan merchants. For example, if any one of the edge construction policies exists between two nodes, the two nodes are connected by an edge.
In the embodiment of the invention, the edges (connecting lines) in the loan information knowledge graph represent the relations among the nodes, wherein the relations among the nodes are unoriented, and the information of other nodes related to the loan application client is acquired, so that the information of the loan application client can be further expanded, and the loan application client can be more comprehensively reflected.
In the embodiment of the present invention, as an optional embodiment, the obtaining, by using a preset order of 3, a vectorized representation of a neighbor node of the preset order that has an edge connection with the current node includes:
Acquiring a node in side contact with the current node to obtain a first-order neighbor node of the current node;
acquiring a node in edge connection with the first-order neighbor node to obtain a second-order neighbor node of the current node;
acquiring a node in side contact with the second-order neighbor node, and acquiring a third-order neighbor node of the current node;
a vectorized representation of each level of neighbor nodes is obtained.
104, processing by using a two-way long and short word memory model based on the vectorization representation of the loan application client and the vectorization representation of the neighbor nodes with the preset orders to obtain an updated vectorization representation of the loan application client;
in the embodiment of the present invention, as an optional embodiment, based on the vectorized representation of the loan application client and the vectorized representation of the neighbor node of the preset order, the processing is performed by using a two-way long and short word memory model, so as to obtain an updated vectorized representation of the loan application client, which includes:
for the vectorized representation of the neighbor nodes of each order, calculating the mean value of the vectorized representation of the neighbor nodes of the order;
constructing a vectorization sequence according to the vectorization representation of the loan application client and the average value of the vectorization representation of the neighbor nodes of each order;
And processing the vectorization sequence by using a two-way long and short word memory model to obtain the updated vectorization representation.
In the embodiment of the invention, the update vector is expressed as follows by using a two-way long and short word memory (BILSTM, bi-directional Long Short Term Memory) model:
in the method, in the process of the invention,
n j a vectorized representation of a j-th order neighbor node to the current node;
C N(1) the number of the first-order neighbor nodes of the current node;
n (1) is a first-order neighbor node of the current node;
n m a vectorized representation of an mth order neighbor node to the current node;
C N(k) the number of the k-th order neighbor nodes of the current node;
n (k) is the kth order neighbor node of the current node.
in an embodiment of the present invention, as an optional embodiment, the obtaining, using a multi-layer perceptual model, a fused vectorized representation of the updated vectorized representation includes:
processing each order vector in the updated vectorization representation by using a multi-layer perception model to obtain each order vectorization representation;
Processing the forward fusion vectorization representation of the previous step and the vectorization representation of the current step by using a long and short word memory model to obtain the forward fusion vectorization representation of the current step; processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation;
and obtaining the fusion vectorization representation based on the zero-order negative fusion vectorization representation and the positive fusion vectorization representation of each order.
In the embodiment of the present invention, as an optional embodiment, each order vector in the updated vectorized representation is respectively processed by using a multi-layer perceptual model to obtain each order vectorized representation, including:
processing the vectorized representation in the updated vectorized representation by using a multi-layer perception model to obtain a zeroth order vectorized representation;
based on an activation function and the updated vectorized representation, acquiring vectorized aggregate representations of neighbor nodes of each order;
and processing the vectorized aggregate representation of each order by utilizing a multi-layer perception model to obtain the vectorized representation of the order.
In an embodiment of the present invention, the multi-layer perceptron (MLP, multilayer Perceptron) model is a multi-layer perceptron.
In the embodiment of the present invention, as an optional embodiment, the vectorized aggregate representation of each level of neighbor nodes is obtained using the following formula:
in the method, in the process of the invention,
g (N (i)) is a vectorized aggregate representation of the ith order neighbor node;
N i is the ith order neighbor node;
h ij a vectorized representation of the ith order, jth neighbor node;
w j a multi-layer perception model parameter matrix;
c i the number of the neighbor nodes is the ith order;
σ () is an activation function.
FIG. 2 illustrates a schematic diagram of acquiring a fusion vectorized representation provided by an embodiment of the present invention. As shown in fig. 2, in the embodiment of the present invention, a multi-layer perceptual model is used to process a vectorized aggregate representation of a first-order neighbor node to obtain a first-order vectorized representation (x 1); processing the vectorized aggregate representation of the second-order neighbor nodes by using a multi-layer perception model to obtain a second-order vectorized representation (x 2); and processing the vectorized aggregate representation of the third-order neighbor node by using the multi-layer perception model to obtain a third-order vectorized representation (x 3). As an alternative embodiment, after the multi-section information extraction, stitching is performed in order from the zeroth order to the kth order, resulting in a multi-order vectorized representation sequence [ x0, x1, x2, xk ].
In the embodiment of the invention, as an optional embodiment, the preset order is 3, and the forward fusion vectorization representation of the previous order and the vectorization representation of the current order are processed by utilizing a long and short word memory model to obtain the forward fusion vectorization representation of the current order; and processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation, comprising:
processing the zeroth order vectorization representation by using a long and short word memory model to obtain a zeroth order forward fusion vectorization representation;
processing the zero-order forward fusion vectorization representation and the first-order vectorization representation by using a long and short word memory model to obtain the first-order forward fusion vectorization representation;
processing the first-order forward fusion vectorization representation and the second-order vectorization representation by using a long and short word memory model to obtain the second-order forward fusion vectorization representation;
processing the second-order forward fusion vectorization representation and the third-order vectorization representation by using a long and short word memory model to obtain a third-order forward fusion vectorization representation;
processing the third-order vectorization representation by using a long and short word memory model to obtain a third-order negative fusion vectorization representation;
Processing the third-order negative fusion vectorization representation and the second-order vectorization representation by using a long and short word memory model to obtain the second-order negative fusion vectorization representation;
processing the second-order negative fusion vectorization representation and the first-order vectorization representation by using a long and short word memory model to obtain the first-order negative fusion vectorization representation;
and processing the first-order negative fusion vectorization representation and the zeroth-order vectorization representation by using the long and short word memory model to obtain the zeroth-order negative fusion vectorization representation.
In an embodiment of the present invention, as an alternative embodiment, the fusion vectorization representation is obtained using the following formula:
H＝(h 1 ，h 2 ，...，h k )
in the method, in the process of the invention,
h t a vectorized representation is fused for the t th order;
h is fusion vectorization representation;
w t a long and short word memory model;
k is a preset order.
In the embodiment of the invention, the hidden state of the BILSTM model is obtained:
h t-1 、/>
h t+1 、h t the hidden fraud risk identification can be realized, and the accuracy of loan application client identification can be effectively improved.
And 106, acquiring fraud probability of the loan application client based on the fusion vectorization representation and the fraud identification model.
In an embodiment of the present invention, as an alternative embodiment, the fraud recognition model calculates the fraud probability using the following formula:
in the method, in the process of the invention,
h is fusion vectorization representation;
W f 、b f model parameters are identified for fraud.
In the embodiment of the present invention, as an optional embodiment, if the acquired fraud probability exceeds a preset probability threshold, it is determined that the loan applicant client is a fraudulent loan applicant client.
In the embodiment of the present invention, as an optional embodiment, the fraud identification model is constructed by:
a11, acquiring sample individual data of a sample loan application client provided with a label, and carrying out sample vectorization representation on the sample individual data;
in an embodiment of the present invention, the sample individual data includes, but is not limited to: liability information, credit history information, social relationship information, and historical loan information.
In the embodiment of the invention, each content item contained in the sample individual data is respectively encoded by using a WOE method, and different encodings are spliced to obtain the vectorized representation of the sample individual data.
In an embodiment of the present invention, a tag includes: fraud and non-fraud, i.e. the sample loan application customers are divided into fraud loan application customers and non-fraud loan application customers.
A12, inquiring the loan information knowledge graph, and acquiring vectorization representation of neighbor nodes with preset orders in the loan information knowledge graph of the sample loan application client;
a13, aiming at the vectorization representation of the neighbor node of each order, acquiring the vectorization aggregation representation of the sample of the neighbor node of the order by using a multi-layer perception model;
a14, obtaining a sample fusion vectorization representation based on the sample vectorization representation of the sample loan application client and the sample vectorization aggregation representation of each-order neighbor node;
and A15, training an initial fraud recognition model based on the sample fusion vectorization representation and the label of the sample loan application client, and obtaining the fraud recognition model when the trained initial fraud recognition model meets the preset precision.
In the embodiment of the invention, when an initial fraud recognition model is trained, a sample fusion vectorization representation of a sample loan application client is used as the input of the initial fraud recognition model, a tag of the sample loan application client is used for carrying out counter-propagation correction on fraud probability prediction output by the initial fraud recognition model, the training is stopped until the precision of the model meets the preset precision, and the initial fraud recognition model when the training is stopped is used as the fraud recognition model for fraud recognition.
The invention is further elucidated below in connection with the drawings and the specific embodiments.
FIG. 3 is a flowchart illustrating a method for identifying fraudulent loan applications, according to an embodiment of the invention. As shown in fig. 3, the method includes:
in the embodiment of the present invention, the 16 kinds of information include: the sum of the number of times of appearance in the first time threshold, the total amount of telephone charge in the second time threshold, whether the identity card hits the p2p blacklist, the number of loan application times of all institutions in the third time threshold, the number of loan application times of non-silver institutions in the fourth time threshold, the number of non-silver institutions in the fourth time threshold, the total number of contacts in the fifth time threshold, the contact number of service personnel, the mobile phone silence duration, the mobile phone use duration, the loan amount, the type of cooperative merchants, the merchant level and the payment ratio.
in the embodiment of the invention, each information corresponds to a vector code, that is, a vectorized representation is performed to represent the client information. For each type of information, vector encoding of that information can be obtained by processing using the WOE method.
in the embodiment of the invention, 4 side relations are predefined: and establishing a loan information knowledge graph according to whether the predefined side relationship exists in the 16-dimensional vector codes of each sample client or not. In the loan information knowledge graph, sample clients with side relations are connected, sample clients without side relations are not connected, and nodes in the loan information knowledge graph are sample clients.
in the embodiment of the invention, 0, 1, 2 and 3-order neighbor node information is respectively extracted for each node from a loan information knowledge graph. Fusing the neighbor node information of 0, 1, 2 and 3 orders to obtain a 128-dimensional fused vector representation (fused vector representation). And inputting the 128-dimensional fusion vectorization representation into an initial fraud identification model to obtain the fraud probability of the client.
in the embodiment of the invention, after 128-dimensional fusion vectorization representation of a sample client (node) is input into an initial fraud recognition model, the fraud probability of the sample client can be obtained, the parameters of the initial fraud recognition model are adjusted in a gradient descent mode through calculating the fraud probability and the actual fraud probability (0 or 1) of the sample client, so that the probability difference between the fraud probability of the sample client and the actual fraud probability obtained by using the initial fraud recognition model with the adjustment parameters is smaller than a preset probability threshold, then the 128-dimensional fusion vectorization representation of the next sample client is input into the initial fraud recognition model, and training is continued until the 128-dimensional fusion vectorization representation of the sample client for testing is input into the trained initial fraud recognition model, and the error of the fraud probability predicted by the model is smaller than the probability threshold.
And 307, inputting the updated 128-dimensional fusion vectorization representation into a fraud identification model to obtain the fraud probability of the loan application client.
In the embodiment of the invention, hidden fraud risks are mined through multi-level information fusion, so that fraud identification capability, especially identification capability for hidden fraudulent clients, is improved, and financial loss is reduced.
FIG. 4 is a schematic diagram of an apparatus for identifying fraudulent loan applications, according to an embodiment of the invention;
the data acquisition module 401 is configured to acquire individual data of a loan application client, where the individual data is data that a pre-constructed fraud recognition model requires to be input;
in an embodiment of the present invention, as an alternative embodiment, the individual data includes, but is not limited to: liability information, credit history information, social relationship information, and historical loan information, wherein,
the liability information includes, but is not limited to: sum of the number of occurrences in the first time threshold, monthly income, and total amount of telephone charges in the second time threshold. As an alternative embodiment, the first time threshold is set to 12 months and the second time threshold is set to 5 months;
credit history information includes, but is not limited to: whether the identity card hits a p2p blacklist, the loan application times of all institutions within a third time threshold from the present, the loan application times of non-silver institutions within a fourth time threshold from the present, and the number of non-silver institutions of the loan application within a fourth time threshold from the present, wherein as an optional embodiment, the third time threshold is set to 3 months, and the fourth time threshold is set to 12 months;
Social relationship information includes, but is not limited to: the total number of contacts, the number of times of contact between the service personnel, the silent duration of the mobile phone and the use duration of the mobile phone in a fifth time threshold from the present, wherein the fifth time threshold is set to be 3 months as an optional embodiment;
historical loan information includes, but is not limited to: loan amount, partner merchant type, merchant level, and rate of payment.
A vectorization module 402, configured to perform vectorization on the individual data;
in an embodiment of the present invention, as an alternative embodiment, the vectorization module 402 includes:
a coding unit (not shown in the figure) for coding the individual data by using an evidence weight method, wherein each content item in the individual data corresponds to a code;
and the vectorization unit is used for splicing codes corresponding to different content items to obtain vectorization representation of the individual data.
The multi-information obtaining module 403 is configured to query a pre-constructed loan information knowledge graph, and obtain a vectorized representation of neighbor nodes of a preset order in the loan information knowledge graph for the loan application client;
in an embodiment of the present invention, as an optional embodiment, the multiple information obtaining module 403 includes:
A query unit (not shown) for querying a previously constructed loan information knowledge map;
a first vector obtaining unit, configured to obtain a vectorized representation of a neighboring node of a preset order that is in edge contact with a current node, if the loan information knowledge graph includes the loan application client, with the loan application client as the current node;
a node adding unit, configured to add the loan application client to a node of the loan information knowledge graph, if the loan information knowledge graph does not include the loan application client, and to set an edge contact between the loan application client and other nodes according to an edge construction policy of the loan information knowledge graph;
and the second vector acquisition unit is used for acquiring the vector representation of the neighbor nodes with preset orders, which are in side contact with the loan application client, from the loan information knowledge graph with side contact.
In the embodiment of the present invention, as an optional embodiment, the obtaining, by using a preset order of 3, a vectorized representation of a neighbor node of the preset order that has an edge connection with the current node includes:
acquiring a node in side contact with the current node to obtain a first-order neighbor node of the current node;
Acquiring a node in edge connection with the first-order neighbor node to obtain a second-order neighbor node of the current node;
acquiring a node in side contact with the second-order neighbor node, and acquiring a third-order neighbor node of the current node;
a vectorized representation of each level of neighbor nodes is obtained.
The vectorization updating module 404 is configured to process the vectorization representation of the loan application client and the vectorization representation of the neighbor node with the preset order by using a two-way long and short word memory model to obtain an updated vectorization representation of the loan application client;
in the embodiment of the invention, the obtained update vector is expressed as follows:
the vectorization fusion module 405 is configured to process each order vector in the updated vectorization representation by using a multi-layer perceptual model, so as to obtain each order vectorization representation; processing the forward fusion vectorization representation of the previous step and the vectorization representation of the current step by using a long and short word memory model to obtain the forward fusion vectorization representation of the current step; processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation; based on each zero-order negative fusion vectorization representation and each order positive fusion vectorization representation, obtaining a fusion vectorization representation;
In an embodiment of the present invention, as an optional embodiment, the vectorization fusion module 405 includes:
a zero-order unit (not shown in the figure) for processing the vectorized representation in the updated vectorized representation by using the multi-layer perceptual model to obtain a zeroth-order vectorized representation;
the aggregation unit is used for acquiring vectorized aggregate representations of all-order neighbor nodes based on an activation function and the updated vectorized representations;
the multi-order unit is used for processing the vectorized aggregate representation of each order by utilizing a multi-layer perception model to obtain the vectorized representation of the order;
and the fusion unit is used for acquiring the fusion vectorized representation based on the zeroth order vectorized representation and the vectorized representations of each order.
In the embodiment of the invention, as an optional embodiment, the preset order is 3, and the forward fusion vectorization representation of the previous order and the vectorization representation of the current order are processed by utilizing a long and short word memory model to obtain the forward fusion vectorization representation of the current order; and processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation, comprising:
Processing the zeroth order vectorization representation by using a long and short word memory model to obtain a zeroth order forward fusion vectorization representation;
processing the zero-order forward fusion vectorization representation and the first-order vectorization representation by using a long and short word memory model to obtain the first-order forward fusion vectorization representation;
processing the first-order forward fusion vectorization representation and the second-order vectorization representation by using a long and short word memory model to obtain the second-order forward fusion vectorization representation;
processing the second-order forward fusion vectorization representation and the third-order vectorization representation by using a long and short word memory model to obtain a third-order forward fusion vectorization representation;
processing the third-order vectorization representation by using a long and short word memory model to obtain a third-order negative fusion vectorization representation;
processing the third-order negative fusion vectorization representation and the second-order vectorization representation by using a long and short word memory model to obtain the second-order negative fusion vectorization representation;
processing the second-order negative fusion vectorization representation and the first-order vectorization representation by using a long and short word memory model to obtain the first-order negative fusion vectorization representation;
And processing the first-order negative fusion vectorization representation and the zeroth-order vectorization representation by using the long and short word memory model to obtain the zeroth-order negative fusion vectorization representation.
And a probability generation module 406, configured to obtain a fraud probability of the loan application client based on the fused vectorized representation and the fraud identification model.
In an embodiment of the present invention, as an optional embodiment, the apparatus further includes:
a model building module (not shown in the figure) for obtaining sample individual data of a sample loan application client provided with a label, and performing sample vectorization representation on the sample individual data;
inquiring the loan information knowledge graph, and acquiring vectorization representation of neighbor nodes of a preset order of the sample loan application client in the loan information knowledge graph;
aiming at the vectorization representation of the neighbor nodes of each order, acquiring the vectorization aggregation representation of the samples of the neighbor nodes of the order by using a multi-layer perception model;
based on the sample vectorization representation of the sample loan application client and the sample vectorization aggregation representation of each-order neighbor node, obtaining a sample fusion vectorization representation;
training an initial fraud recognition model based on the sample fusion vectorization representation and the label of the sample loan application client, and obtaining the fraud recognition model when the trained initial fraud recognition model meets the preset precision.
As shown in fig. 5, an embodiment of the present application provides a computer device 500 for executing the method for identifying fraudulent loan applications in fig. 1, where the device includes a memory 501, a processor 502 connected to the memory 501 via a bus, and a computer program stored on the memory 501 and executable on the processor 502, where the steps of the method for identifying fraudulent loan applications are implemented when the processor 502 executes the computer program.
In particular, the memory 501 and the processor 502 can be general-purpose memories and processors, which are not particularly limited herein, and the above-described method of identifying fraudulent loan applications can be performed when the processor 502 runs a computer program stored in the memory 501.
Corresponding to the method of identifying fraudulent loan applications in fig. 1, embodiments of the present application also provide a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, performs the steps of the method of identifying fraudulent loan applications described above.
In particular, the storage medium can be a general-purpose storage medium, such as a removable disk, a hard disk, etc., on which a computer program is executed that is capable of performing the above-described method of identifying fraudulent loan applications.
In the embodiments provided herein, it should be understood that the disclosed systems and methods may be implemented in other ways. The system embodiments described above are merely illustrative, e.g., the division of the elements is merely a logical functional division, and there may be additional divisions in actual implementation, and e.g., multiple elements or components may be combined or integrated into another system, or some features may be omitted, or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be through some communication interface, system or unit indirect coupling or communication connection, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments provided in the present application may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on such understanding, the technical solution of the present application may be embodied essentially or in a part contributing to the prior art or in a part of the technical solution, in the form of a software product stored in a storage medium, including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present application. And the aforementioned storage medium includes: a U-disk, a removable hard disk, a Read-Only Memory (ROM), a random access Memory (Random Access Memory, RAM), a magnetic disk, or an optical disk, or other various media capable of storing program codes.
It should be noted that: like reference numerals and letters in the following figures denote like items, and thus once an item is defined in one figure, no further definition or explanation of it is required in the following figures, and furthermore, the terms "first," "second," "third," etc. are used merely to distinguish one description from another and are not to be construed as indicating or implying relative importance.
Finally, it should be noted that: the foregoing examples are merely specific embodiments of the present application, and are not intended to limit the scope of the present application, but the present application is not limited thereto, and those skilled in the art will appreciate that while the foregoing examples are described in detail, the present application is not limited thereto. Any person skilled in the art may modify or easily conceive of the technical solution described in the foregoing embodiments, or make equivalent substitutions for some of the technical features within the technical scope of the disclosure of the present application; such modifications, changes or substitutions do not depart from the spirit and scope of the corresponding technical solutions. Are intended to be encompassed within the scope of this application. Therefore, the protection scope of the present application shall be subject to the protection scope of the claims.
Claims (9)
1. A method of identifying fraudulent loan applications, comprising:
obtaining individual data of a loan application client, wherein the individual data is data required to be input by a pre-constructed fraud identification model;
vectorizing the individual data;
Inquiring a pre-constructed loan information knowledge graph, and acquiring vectorization representation of neighbor nodes with preset orders in the loan information knowledge graph by a loan application client;
based on the vectorization representation of the loan application client and the vectorization representation of the neighbor nodes with the preset order, processing by using a two-way long and short word memory model to obtain updated vectorization representation of the loan application client;
processing each order vector in the updated vectorization representation by using a multi-layer perception model to obtain each order vectorization representation;
processing the forward fusion vectorization representation of the previous step and the vectorization representation of the current step by using a long and short word memory model to obtain the forward fusion vectorization representation of the current step; processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation;
based on each zero-order negative fusion vectorization representation and each order positive fusion vectorization representation, obtaining a fusion vectorization representation;
acquiring fraud probability of the loan application client based on the fusion vectorization representation and the fraud identification model;
The vectorization representation of the loan application client and the vectorization representation of the neighbor node of the preset order are processed by using a two-way long and short word memory model to obtain the updated vectorization representation of the loan application client, which comprises the following steps:
for the vectorized representation of the neighbor nodes of each order, calculating the mean value of the vectorized representation of the neighbor nodes of the order;
constructing a vectorization sequence according to the vectorization representation of the loan application client and the average value of the vectorization representation of the neighbor nodes of each order;
and processing the vectorization sequence by using a two-way long and short word memory model to obtain the updated vectorization representation.
2. The method of claim 1, wherein said vectorizing said individual data comprises:
encoding the individual data by using an evidence weight method, wherein each content item in the individual data corresponds to one code;
and splicing codes corresponding to different content items to obtain the vectorized representation of the individual data.
3. The method of claim 1, wherein the querying the pre-constructed loan information knowledge graph to obtain a vectorized representation of the preset order of neighbor nodes of the loan application client in the loan information knowledge graph comprises:
Inquiring a pre-constructed loan information knowledge graph;
if the loan information knowledge graph comprises the loan application client, taking the loan application client as a current node, and acquiring vectorization representation of neighbor nodes with preset orders in side connection with the current node;
if the loan information knowledge graph does not contain the loan application client, adding the loan application client into a node of the loan information knowledge graph, and setting the side contact of the loan application client and other nodes according to a side construction strategy of the loan information knowledge graph;
and obtaining the vectorization representation of the neighbor nodes with preset orders in the side connection with the loan application client from the loan information knowledge graph with the side connection.
4. A method according to claim 3, wherein the predetermined order is 3, and obtaining a vectorized representation of a predetermined order of neighboring nodes that are in edge contact with the current node comprises:
acquiring a node in side contact with the current node to obtain a first-order neighbor node of the current node;
acquiring a node in edge connection with the first-order neighbor node to obtain a second-order neighbor node of the current node;
Acquiring a node in side contact with the second-order neighbor node, and acquiring a third-order neighbor node of the current node;
a vectorized representation of each level of neighbor nodes is obtained.
5. The method according to any one of claims 1 to 4, wherein said processing each order vector in said updated vectorized representation separately using a multi-layer perceptual model to obtain each order vectorized representation comprises:
processing the vectorized representation in the updated vectorized representation by using a multi-layer perception model to obtain a zeroth order vectorized representation;
based on an activation function and the updated vectorized representation, acquiring vectorized aggregate representations of neighbor nodes of each order;
and processing the vectorized aggregate representation of each order by utilizing a multi-layer perception model to obtain the vectorized representation of the order.
6. The method according to claim 5, wherein the preset order is 3, and the processing is performed on the forward fusion vectorization representation of the previous order and the vectorization representation of the current order by using a long and short word memory model to obtain the forward fusion vectorization representation of the current order; and processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation, comprising:
Processing the zeroth order vectorization representation by using a long and short word memory model to obtain a zeroth order forward fusion vectorization representation;
processing the zero-order forward fusion vectorization representation and the first-order vectorization representation by using a long and short word memory model to obtain the first-order forward fusion vectorization representation;
processing the first-order forward fusion vectorization representation and the second-order vectorization representation by using a long and short word memory model to obtain the second-order forward fusion vectorization representation;
processing the second-order forward fusion vectorization representation and the third-order vectorization representation by using a long and short word memory model to obtain a third-order forward fusion vectorization representation;
processing the third-order vectorization representation by using a long and short word memory model to obtain a third-order negative fusion vectorization representation;
processing the third-order negative fusion vectorization representation and the second-order vectorization representation by using a long and short word memory model to obtain the second-order negative fusion vectorization representation;
processing the second-order negative fusion vectorization representation and the first-order vectorization representation by using a long and short word memory model to obtain the first-order negative fusion vectorization representation;
And processing the first-order negative fusion vectorization representation and the zeroth-order vectorization representation by using the long and short word memory model to obtain the zeroth-order negative fusion vectorization representation.
7. An apparatus for identifying fraudulent loan applications, comprising:
the data acquisition module is used for acquiring individual data of a loan application client, wherein the individual data is data required to be input by a pre-constructed fraud identification model;
the vectorization module is used for vectorizing the individual data;
the multi-information acquisition module is used for inquiring a pre-constructed loan information knowledge graph and acquiring vectorization representation of neighbor nodes with preset orders in the loan information knowledge graph of the loan application client;
the vectorization updating module is used for processing the vectorization representation of the loan application client and the vectorization representation of the neighbor node with the preset order by using a two-way long and short word memory model to obtain updated vectorization representation of the loan application client;
the vectorization fusion module is used for respectively processing each order vector in the updated vectorization representation by utilizing a multi-layer perception model to obtain each order vectorization representation; processing the forward fusion vectorization representation of the previous step and the vectorization representation of the current step by using a long and short word memory model to obtain the forward fusion vectorization representation of the current step; processing the later-order negative fusion vectorization representation and the current-order vectorization representation to obtain the current-order negative fusion vectorization representation; based on each zero-order negative fusion vectorization representation and each order positive fusion vectorization representation, obtaining a fusion vectorization representation;
The probability generation module is used for acquiring fraud probability of the loan application client based on the fusion vectorization representation and the fraud identification model;
the vectorization updating module is used for processing the vectorization representation of the loan application client and the vectorization representation of the neighbor node with the preset order by using a two-way long and short word memory model to obtain the updated vectorization representation of the loan application client, and is specifically used for:
for the vectorized representation of the neighbor nodes of each order, calculating the mean value of the vectorized representation of the neighbor nodes of the order;
constructing a vectorization sequence according to the vectorization representation of the loan application client and the average value of the vectorization representation of the neighbor nodes of each order;
and processing the vectorization sequence by using a two-way long and short word memory model to obtain the updated vectorization representation.
8. A computer device, comprising: a processor, a memory and a bus, the memory storing machine-readable instructions executable by the processor, the processor and the memory in communication via the bus when the computer device is running, the machine-readable instructions when executed by the processor performing the steps of the method of identifying fraudulent loan applications of any of claims 1 to 6.
9. A computer readable storage medium, characterized in that it has stored thereon a computer program which, when executed by a processor, performs the steps of the method of identifying fraudulent loan applications according to any of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211000110.9A CN115345727B (en) | 2022-08-19 | 2022-08-19 | Method and device for identifying fraudulent loan application |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211000110.9A CN115345727B (en) | 2022-08-19 | 2022-08-19 | Method and device for identifying fraudulent loan application |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115345727A CN115345727A (en) | 2022-11-15 |
| CN115345727B true CN115345727B (en) | 2023-04-28 |
Family
ID=83954769
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211000110.9A Active CN115345727B (en) | 2022-08-19 | 2022-08-19 | Method and device for identifying fraudulent loan application |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115345727B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115860906A (en) * | 2022-11-22 | 2023-03-28 | 中电金信软件有限公司 | Credit risk identification method, credit risk identification device and storage medium |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110930246A (en) * | 2019-12-04 | 2020-03-27 | 深圳市新国都金服技术有限公司 | Credit anti-fraud identification method and device, computer equipment and computer-readable storage medium |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180033009A1 (en) * | 2016-07-27 | 2018-02-01 | Intuit Inc. | Method and system for facilitating the identification and prevention of potentially fraudulent activity in a financial system |
| CN110245165B (en) * | 2019-05-20 | 2023-04-11 | 平安科技（深圳）有限公司 | Risk conduction associated graph optimization method and device and computer equipment |
| CN111639193A (en) * | 2020-05-27 | 2020-09-08 | 上海明略人工智能(集团)有限公司 | Product risk assessment method and device, electronic equipment and storage medium |
| CN114462742A (en) * | 2020-11-10 | 2022-05-10 | 腾讯科技（深圳）有限公司 | Risk prompting method, device, equipment and computer storage medium |
| CN113822494B (en) * | 2021-10-19 | 2022-07-22 | 平安科技（深圳）有限公司 | Risk prediction method, device, equipment and storage medium |
| CN114282003A (en) * | 2021-11-30 | 2022-04-05 | 中国建设银行股份有限公司 | Financial risk early warning method and device based on knowledge graph |
-
2022
- 2022-08-19 CN CN202211000110.9A patent/CN115345727B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110930246A (en) * | 2019-12-04 | 2020-03-27 | 深圳市新国都金服技术有限公司 | Credit anti-fraud identification method and device, computer equipment and computer-readable storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115345727A (en) | 2022-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110009174B (en) | Risk recognition model training method and device and server | |
| Benchaji et al. | Enhanced credit card fraud detection based on attention mechanism and LSTM deep model | |
| WO2019196546A1 (en) | Method and apparatus for determining risk probability of service request event | |
| US11562372B2 (en) | Probabilistic feature engineering technique for anomaly detection | |
| CN110704730A (en) | Product data pushing method and system based on big data and computer equipment | |
| US20110166979A1 (en) | Connecting decisions through customer transaction profiles | |
| CN111260189B (en) | Risk control method, risk control device, computer system and readable storage medium | |
| CN110362689A (en) | A kind of methods of risk assessment, device, storage medium and server | |
| CN110570188A (en) | Method and system for processing transaction requests | |
| CN111428217A (en) | Method and device for identifying cheat group, electronic equipment and computer readable storage medium | |
| CN115345727B (en) | Method and device for identifying fraudulent loan application | |
| CN115545886A (en) | Overdue risk identification method, overdue risk identification device, overdue risk identification equipment and storage medium | |
| CN116307671A (en) | Risk early warning method, risk early warning device, computer equipment and storage medium | |
| CN111915218A (en) | Financial counterfeiting identification method and system based on LSTM-CNN | |
| CN110795466A (en) | Anti-fraud method based on big data processing, server and computer-readable storage medium | |
| CN113408627A (en) | Target object determination method and device and server | |
| CN112950290A (en) | Mining method and device for economic dependence clients, storage medium and electronic equipment | |
| CN115034520B (en) | Risk prediction method, device, equipment and storage medium | |
| CN111709844A (en) | Insurance money laundering personnel detection method and device and computer readable storage medium | |
| CN110570301B (en) | Risk identification method, device, equipment and medium | |
| CN113706258A (en) | Product recommendation method, device, equipment and storage medium based on combined model | |
| CN111932131A (en) | Service data processing method and device | |
| CN109800947B (en) | Loan transaction processing method and device based on machine learning and computer equipment | |
| EP4310755A1 (en) | Self learning machine learning transaction scores adjustment via normalization thereof | |
| CN117350461B (en) | Enterprise abnormal behavior early warning method, system, computer equipment and storage medium |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |