CN115438747A - Abnormal account recognition model training method, device, equipment and medium - Google Patents
Abnormal account recognition model training method, device, equipment and medium Download PDFInfo
- Publication number
- CN115438747A CN115438747A CN202211202721.1A CN202211202721A CN115438747A CN 115438747 A CN115438747 A CN 115438747A CN 202211202721 A CN202211202721 A CN 202211202721A CN 115438747 A CN115438747 A CN 115438747A
- Authority
- CN
- China
- Prior art keywords
- target
- abnormal
- abnormal account
- account identification
- historical
- 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
- 230000002159 abnormal effect Effects 0.000 title claims abstract description 269
- 238000012549 training Methods 0.000 title claims abstract description 151
- 238000000034 method Methods 0.000 title claims abstract description 80
- 101100455978 Arabidopsis thaliana MAM1 gene Proteins 0.000 claims abstract description 38
- 230000006399 behavior Effects 0.000 claims description 55
- 230000006870 function Effects 0.000 claims description 26
- 238000012360 testing method Methods 0.000 claims description 20
- 230000003044 adaptive effect Effects 0.000 claims description 14
- 238000013528 artificial neural network Methods 0.000 claims description 13
- 238000012545 processing Methods 0.000 claims description 5
- 238000011002 quantification Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 description 15
- 238000001514 detection method Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 11
- 230000000694 effects Effects 0.000 description 6
- 230000002547 anomalous effect Effects 0.000 description 4
- 238000004891 communication Methods 0.000 description 3
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000006978 adaptation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
The application provides a method, a device, equipment and a medium for training an abnormal account recognition model. The method comprises the following steps: acquiring a training sample and a category label of a target service scene; inputting training samples and class labels of a target business scene into the abnormal account identification model after the initialization parameters so as to train the abnormal account identification model after the initialization parameters and obtain a target abnormal account identification model; the abnormal account identification models after the parameters are initialized are determined by adopting an MAML algorithm, and the abnormal account identification models after the parameters are initialized corresponding to a plurality of service scenes are consistent. The MAML algorithm utilizes the common characteristic features of illegal behavior transactions in different service scenes to determine the abnormal account recognition model after the initialization parameters, so that the target abnormal account recognition model can be obtained only by training the abnormal account recognition model after the initialization parameters by a small number of training samples and class labels and a few iteration times.
Description
Technical Field
The application relates to the field of artificial intelligence, in particular to a method, a device, equipment and a medium for training an abnormal account recognition model.
Background
With the development of mobile internet, various bank clients continuously appear, and hidden dangers are brought to the safety of bank accounts. Such as the unlawful act of using bank accounts. There is therefore a need to identify transaction accounts that may be involved in illicit activities.
At present, illegal behavior model rules are generally formulated by combining a business scene, batch transaction data monitoring is carried out by relying on a large data platform, relevant fields of transactions are checked, and suspicious transactions of suspected illegal behaviors are screened out according to the relevant fields of the transactions and the illegal behavior model rules.
However, the law violation behavior model rule needs to be completely redesigned for each business scene, and the common typical characteristics of the law violation behavior transactions in different business scenes cannot be utilized; and because the detection relies on the design of model rules, the manually designed rules are limited and may have holes, resulting in lower accuracy of suspicious transaction detection.
Disclosure of Invention
The application provides an abnormal account identification model training method, device, equipment and medium, which are used for solving the problems that the conventional model rules cannot utilize common typical characteristics of illegal behavior transactions in different business scenes, and the accuracy of suspicious transaction detection is low.
In a first aspect, the present application provides a method for training an abnormal account recognition model, including:
acquiring a training sample and a category label of a target service scene; the training sample comprises historical characteristic data of a historical account in a target business scene; the historical characteristic data comprises: attribute characteristic values and historical behavior characteristic values;
inputting training samples and class labels of a target business scene into the abnormal account identification model after the initialization parameters so as to train the abnormal account identification model after the initialization parameters and obtain a target abnormal account identification model, wherein the target abnormal account identification model is used for identifying whether a target account is abnormal or not in the target business scene; the abnormal account identification models after the initialization parameters are determined by adopting a model independent learning MAML algorithm, and the abnormal account identification models after the initialization parameters corresponding to a plurality of service scenes are consistent.
In a second aspect, the present application provides an abnormal account identification method, including:
acquiring current characteristic data of a target account in a target service scene; the current feature data includes: the current attribute characteristic value and the current behavior characteristic value of the target account;
inputting the current characteristic data into a target abnormal account identification model to classify the current characteristic data so as to identify whether the transaction behavior of a target account is abnormal or not; the target abnormal account identification model is obtained by training the abnormal account identification model after the parameters are initialized by adopting the abnormal account identification model training method of the first aspect.
In a third aspect, the present application provides an abnormal account recognition model training apparatus, including:
the acquisition module is used for acquiring a training sample and a category label of a target business scene; the training sample comprises historical characteristic data of a historical account in a target business scene; the historical characteristic data comprises: attribute characteristic values and historical behavior characteristic values;
the training module is used for inputting training samples and class labels of a target business scene into the abnormal account identification model after the initialization parameters so as to train the abnormal account identification model after the initialization parameters and obtain a target abnormal account identification model, and the target abnormal account identification model is used for identifying whether a target account is abnormal or not in the target business scene; the abnormal account identification models after the initialization of the parameters are determined by adopting a model independent learning MAML algorithm, and the abnormal account identification models after the initialization of the parameters corresponding to a plurality of service scenes are consistent.
In a fourth aspect, the present application provides an abnormal account identification apparatus, including:
the acquisition module is used for acquiring the current characteristic data of the target account in a target service scene; the current feature data includes: the current attribute characteristic value and the current behavior characteristic value of the target account;
the classification module is used for inputting the current characteristic data into a target abnormal account identification model to classify the current characteristic data so as to identify whether the transaction behavior of the target account is abnormal or not; the target abnormal account recognition model is obtained by training the abnormal account recognition model after the parameters are initialized by adopting the abnormal account recognition model training method of the first aspect.
In a fifth aspect, the present application provides an electronic device, comprising: a processor, and a memory communicatively coupled to the processor;
a circuit interconnection between the processor and the memory;
the memory stores computer execution instructions;
the processor executes computer-executable instructions stored in the memory to implement the abnormal account recognition model training method of the first aspect or the abnormal account recognition method of the second aspect.
In a sixth aspect, the present application provides a computer-readable storage medium, in which computer-executable instructions are stored, and the computer-executable instructions are executed by a processor to implement the abnormal account recognition model training method according to the first aspect or the abnormal account recognition method according to the second aspect.
The abnormal account identification model training method, the abnormal account identification model training device, the abnormal account identification model training equipment and the abnormal account identification model training medium obtain training samples and class labels of a target business scene; the training sample comprises historical characteristic data of a historical account in a target business scene; the historical characteristic data comprises: attribute characteristic values and historical behavior characteristic values; inputting training samples and class labels of a target business scene into the abnormal account identification model after the initialization parameters so as to train the abnormal account identification model after the initialization parameters and obtain a target abnormal account identification model, wherein the target abnormal account identification model is used for identifying whether a target account is abnormal or not in the target business scene; the abnormal account identification models after the initialization parameters are determined by adopting a model independent learning MAML algorithm, and the abnormal account identification models after the initialization parameters corresponding to a plurality of service scenes are consistent. Due to the fact that the model independent learning MAML algorithm is adopted, the typical characteristics common to illegal action and transactions in different service scenes are utilized, and the parameters in the abnormal account identification model after the parameters are initialized have strong enough adaptability. Therefore, the target abnormal account identification model can be obtained only by slightly correcting the parameters in the abnormal account identification model after the parameters are initialized by a small number of training samples and class labels of the target business scene and a very small number of iterations.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the application and, together with the description, serve to explain the principles of the application.
Fig. 1 is a flowchart of an abnormal account recognition model training method according to an embodiment of the present application;
fig. 2 is a flowchart of an abnormal account recognition model training method provided in the second embodiment of the present application;
FIG. 3 is a schematic diagram of historical data set allocation provided by an embodiment of the present application;
fig. 4 is a schematic diagram of an abnormal account recognition model training method according to an embodiment of the present application;
fig. 5 is a flowchart of an abnormal account identification method provided in the third embodiment of the present application;
fig. 6 is a schematic structural diagram of an abnormal account recognition model training device according to a fourth embodiment of the present application;
fig. 7 is a schematic structural diagram of an abnormal account identification apparatus according to a fourth embodiment of the present application;
fig. 8 is a schematic structural diagram of an electronic device according to a fifth embodiment of the present application.
Specific embodiments of the present application have been shown by way of example in the drawings and will be described in more detail below. The drawings and written description are not intended to limit the scope of the inventive concepts in any manner, but rather to illustrate the concepts of the application by those skilled in the art with reference to specific embodiments.
Detailed Description
Reference will now be made in detail to the exemplary embodiments, examples of which are illustrated in the accompanying drawings. The following description refers to the accompanying drawings in which the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present application. Rather, they are merely examples of apparatus and methods consistent with certain aspects of the present application, as detailed in the appended claims.
The terms "first", "second", etc. are used for descriptive purposes only and are not to be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated. In the description of the following examples, "plurality" means two or more unless specifically limited otherwise.
The prior art to which the present invention relates will be explained and analyzed in detail below.
At present, illegal behavior model rules are generally formulated by combining a business scene, batch transaction data monitoring is carried out by relying on a large data platform, relevant fields of transactions are checked, and suspicious transactions of suspected illegal behaviors are screened out according to the relevant fields of the transactions and the illegal behavior model rules.
However, the rule of the illegal activity model needs to be completely redesigned for each business scene, and the common typical characteristics of illegal activity transactions in different business scenes cannot be utilized; and because the detection relies on the design of model rules, the manually designed rules are limited and may have holes, the accuracy of suspicious transaction detection is low.
The inventor finds out in research that the task of screening out suspicious transactions of suspected illegal behaviors is locked and regarded as a two-classification problem, a target abnormal account identification model is adopted to classify target accounts, whether suspicious transactions exist or not is determined according to whether the target accounts are abnormal or not, and the accuracy of suspicious transaction detection can be improved. And before the abnormal account identification model is trained aiming at a certain business scene, the abnormal account identification model after the initialization parameters is made to learn the common typical characteristics of illegal behavior transactions under each business scene in advance, so that the speed of training the abnormal account identification model after the initialization parameters can be improved. And the parameters in the abnormal account identification model after the parameters are initialized can have strong enough adaptability by adopting the model independent element learning MAML algorithm. On the basis, only a small number of training samples and class labels of the target service scene and very few iteration times are needed, and the target abnormal account identification model can be obtained by slightly correcting the parameters.
The following describes the technical solution of the present application and how to solve the above technical problems in detail by specific embodiments. These several specific embodiments may be combined with each other below, and details of the same or similar concepts or processes may not be repeated in some embodiments. Embodiments of the present application will be described below with reference to the accompanying drawings.
Example one
Fig. 1 is a flowchart of an abnormal account recognition model training method provided in an embodiment of the present application, and the embodiment of the present application provides an abnormal account recognition model training method for solving the problems that the existing model rules cannot utilize typical characteristics common to illegal transactions in different business scenarios, and the accuracy of suspicious transaction detection is low. The method in the embodiment is applied to an abnormal account recognition model training device, and the abnormal account recognition model training device can be located in electronic equipment. Among other things, the electronic device may be a digital computer that represents various forms. Such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers.
As shown in fig. 1, the method comprises the following specific steps:
step S101, obtaining a training sample and a class label of a target business scene.
The training sample comprises historical characteristic data of a historical account in a target service scene; the historical characteristic data comprises: attribute eigenvalues and historical behavior eigenvalues.
In this embodiment of the present application, the category label may be abnormal or normal, and the category label may also be an identifier indicating that the account corresponding to the training sample is abnormal or an identifier 5 indicating that the account corresponding to the training sample is normal, which is not limited in this embodiment of the present application. For example, the obtained class label may be 0 or 1, where a class label of 0 indicates that the account corresponding to the training sample is normal, and a class label of 1 indicates that the account corresponding to the training sample is abnormal.
In the embodiment of the application, one group of training samples and class labels correspond to one historical account in a target business scene. The training samples and the class labels can be obtained according to historical transaction flow data of the historical account and attributes of the historical transaction flow data.
Step S102, inputting training samples and class labels of the target business scene into the abnormal account recognition model after the initialization parameters, so as to train the abnormal account recognition model after the initialization parameters and obtain the target abnormal account recognition model.
The target abnormal account identification model is used for identifying whether a target account is abnormal or not in a target service scene; the abnormal account identification model after the initialization of the parameters is determined by adopting a model independent learning MAML algorithm, and the abnormal account identification models after the initialization of the parameters corresponding to a plurality of service scenes are consistent.
It should be understood that the abnormal account identification model after the model independent element learning MAML algorithm determines the initialization parameters has learned the common typical characteristics of the abnormal transaction account and the normal transaction account of each business scene, and the initialization parameters are not randomly determined. The abnormal account identification models after the initialization parameters learn common typical characteristics of the abnormal transaction accounts and the normal transaction accounts of all the service scenes, so that the abnormal account identification models after the initialization parameters corresponding to the service scenes are consistent, and the abnormal account identification models after the initialization parameters are trained when any service scene of the target service scene is selected, so that the target abnormal account identification models suitable for the target service scene are obtained.
In the implementation of the application, the abnormal account identification model after the initialization parameters are determined by the model independent learning MAML algorithm, the initialized parameters in the abnormal account identification model after the parameters are initialized are the optimized initialization parameters, the parameter fine adjustment of the abnormal account identification model is carried out through model training on the basis of the optimized initialization parameters, the convergence condition can be quickly reached only by a small number of training samples and class labels, and the training of the abnormal account identification model after the parameters are initialized is completed to obtain the target abnormal account identification model.
The abnormal account identification model training method provided by the embodiment of the application obtains a training sample and a category label of a target business scene; the training sample comprises historical characteristic data of a historical account in a target business scene; the historical characteristic data comprises: attribute characteristic values and historical behavior characteristic values; inputting training samples and class labels of a target business scene into the abnormal account identification model after the initialization parameters so as to train the abnormal account identification model after the initialization parameters and obtain a target abnormal account identification model, wherein the target abnormal account identification model is used for identifying whether a target account is abnormal or not in the target business scene; the abnormal account identification model after the parameters are initialized is determined by adopting a model independent learning MAML algorithm, and the abnormal account identification models after the parameters are initialized corresponding to a plurality of service scenes are consistent. Due to the fact that the model irrelevant element learning MAML algorithm is adopted, the common typical characteristics of illegal behavior transactions under different service scenes are utilized, and the parameters in the abnormal account identification model after the parameters are initialized have strong enough adaptability. Therefore, the target abnormal account identification model can be obtained only by slightly correcting the parameters in the abnormal account identification model after the parameters are initialized by a small number of training samples and class labels of the target business scene and a very small number of iterations.
In an optional implementation manner, the specific manner of executing step S101 to obtain the training sample and the class label of the target service scenario includes:
step S1011, obtaining transaction data of a plurality of historical accounts within preset days, corresponding account characteristics and category labels from a database corresponding to the target service scenario. The accounts with abnormal category labels are included in the plurality of accounts.
In this embodiment of the application, the transaction data may include transaction flow information within a preset number of days, the corresponding account characteristics may include a line to which the account belongs, and the corresponding category label may be abnormal or non-abnormal. For example, the transaction data may include transaction running information for a predetermined number of days before the anomalous account produces the anomalous transaction activity, including the day of the anomalous transaction activity.
Specifically, after determining that some transaction flow information is abnormal or determining that an account has abnormal transaction behavior, the corresponding category label of the corresponding account may be determined to be abnormal, and the transaction data of the account within a preset number of days, the corresponding account characteristics, and the category label are stored in the database corresponding to the corresponding service scenario. After the target service scene is determined, transaction data of a plurality of historical accounts within preset days, corresponding account characteristics and category labels can be obtained from a database corresponding to the target service scene.
In the embodiment of the application, the training of the abnormal account identification model after the parameters are initialized can be completed only by a small amount of transaction data of the account with abnormal category labels, corresponding account characteristics and labels.
Step S1012, extracting data characteristics of the transaction data corresponding to each historical account to obtain historical behavior characteristic values of a plurality of historical accounts in the target service scenario.
For example, the extracted data characteristics of the transaction data corresponding to each historical account may include: the number of transactions over a preset number of days, the amount of transactions over a preset number of days, the number of different counterparties over a preset number of days, etc., the number of transactions between certain preset time periods. For example, the number of trades between certain preset time periods may be from 20 to 24 trades in 5 days.
It should be understood that, the transaction data corresponding to the historical account may also be subjected to a quantification process, for example, a line to which the transaction-partner account belongs included in the transaction data is subjected to a quantification process, and the line to which the transaction-partner account belongs after the quantification process is determined as the historical behavior feature value of the historical account.
Step S1013, quantifying the account features corresponding to the historical accounts to obtain attribute feature values of the multiple historical accounts in the target service scene.
It should be understood that the account characteristics corresponding to the account may include a line to which the account belongs, and may further include other characteristics, for example, an account identifier, information related to an account issuer, an account opening duration, and the like, which is not limited in this embodiment of the present application.
For example, the account features corresponding to the account may include lines to which the account belongs, each line to which the account belongs and the unique identifier corresponding to the line may be stored in a certain storage space in an associated manner, when the account features corresponding to each historical account are quantified, the unique identifier corresponding to the line to which the account belongs may be acquired from the storage space, and the unique identifier corresponding to the line to which the account belongs is determined as the attribute feature value of the account, so as to perform the quantification process on the line to which the account belongs. Wherein, the corresponding unique identification is a specific numerical value.
Step 1014, determining the historical behavior characteristic value and the attribute characteristic value corresponding to each historical account as each training sample, and determining the category label corresponding to each historical account as the category label corresponding to the training sample.
According to the abnormal account identification model training method provided by the embodiment of the application, transaction data, corresponding account characteristics and category labels of a plurality of historical accounts within preset days are obtained from a database corresponding to a target service scene; the accounts comprise accounts with abnormal category labels; extracting data characteristics of transaction data corresponding to each historical account to obtain historical behavior characteristic values of a plurality of historical accounts in a target service scene; quantifying the account characteristics corresponding to the historical accounts to obtain attribute characteristic values of the historical accounts in a target service scene; and determining the historical behavior characteristic value and the attribute characteristic value corresponding to each historical account as each training sample, and determining the category label corresponding to each historical account as the category label corresponding to the training sample. The method and the device can realize quick acquisition of the training samples and the category labels of the target business scene, can filter the transaction data and the characteristic attributes irrelevant to abnormal transaction behaviors in the account characteristics, and ensure the high efficiency of abnormal account identification.
In an optional implementation manner, the specific manner of executing step S102 to input the training sample and the class label of the target service scenario into the abnormal account recognition model after the initialization parameter, so as to train the abnormal account recognition model after the initialization parameter, and obtain the target abnormal account recognition model includes:
inputting training samples and class labels of a target service scene into the deep neural network DNN model after initializing parameters; training the deep neural network DNN model with the initialized parameters by adopting a training sample and a class label of a target service scene; judging whether a loss function in the trained DNN model reaches a preset threshold value or not; and if the loss function is determined to reach the preset threshold value, determining the DNN model with the minimum loss function as the target abnormal account identification model.
The method for training the deep neural network DNN model after initializing the parameters is not limited in the embodiment of the present application, and for example, a gradient descent algorithm may be used to train the DNN model.
In the embodiment of the application, before training the abnormal account recognition model after the initialization parameters, the method further comprises the step of constructing the abnormal account recognition model. The abnormal account identification model is a deep neural network DNN model, the number of input nodes is the same as the number of the characteristics in the historical characteristic data, the number of output nodes is 2, and the classification result is abnormal or normal. The DNN model uses fully connected layers to fit the classifier. Specifically, the input layer is used for receiving historical characteristic data of training samples, each hidden layer in the middle is subjected to nonlinear transformation of an activation function, and network parameters of each layer lIncluding connection weight W (l) And bias term b (l) . The classification result given by the final output layer has the following two possibilities: the transaction behavior of the account is abnormal, or the transaction behavior of the account is normal.
Optionally, a Softmax layer may be further included in the DNN model, and the probability of obtaining the abnormal transaction behavior of the account and the probability of the normal transaction behavior of the account may be output.
Optionally, a test sample and a category label of the target service scene may also be obtained, and the test sample of the target service scene is input into the target abnormal account identification model to determine whether the classification accuracy of the target abnormal account identification model on the target account reaches a preset accuracy threshold. If the target account identification model does not reach the preset accuracy threshold, the target abnormal account identification model needs to be trained again so as to update parameters in the target abnormal account identification model and improve the classification accuracy of the target abnormal account identification model to the target account; if the preset accuracy threshold is reached, the target account can be classified by adopting the target abnormal account identification model. The manner of obtaining the test sample and the category label of the target service scenario is similar to the manner of obtaining the training sample and the category label of the target service scenario, and is not repeated here.
According to the abnormal account identification model training method provided by the embodiment of the application, the abnormal account identification model after the parameters are initialized is a deep neural network DNN model after the parameters are initialized, and training samples and class labels of a target service scene are input into the deep neural network DNN model after the parameters are initialized; training the deep neural network DNN model with the initialized parameters by adopting a training sample and a class label of a target service scene; judging whether a loss function in the trained DNN model reaches a preset threshold value or not; and if the loss function is determined to reach the preset threshold, determining the DNN model with the minimum loss function as the target abnormal account identification model. Because the DNN model has a large number of hidden layers and a large number of intermediate nodes, and can store more learning rules, the learning depth is relatively deep, and the accuracy of classifying the accounts can be further improved.
Example two
Fig. 2 is a flowchart of an abnormal account recognition model training method provided in the second embodiment of the present application, and on the basis of the first embodiment, the present embodiment relates to a specific process of determining an abnormal account recognition model after initialization parameters are determined by using a model independent learning MAML algorithm.
As shown in fig. 2, the method comprises the following specific steps:
step S201, acquiring historical transaction data sets under a plurality of service scenes.
The historical transaction data set comprises transaction data samples and category labels corresponding to a plurality of service scenes; the transaction data sample comprises historical characteristic data of the account under the corresponding business scene.
In the embodiment of the application, the mode of acquiring historical transaction data sets under a plurality of service scenes is similar to the mode of acquiring training samples and category labels of target service scenes, and the historical transaction data sets under the plurality of service scenes are acquired by acquiring transaction data of a plurality of historical accounts within preset days, corresponding account characteristics and category labels from a database corresponding to each service scene.
Step S202, initializing parameters in the abnormal account identification model according to the historical transaction data set by adopting a model independent meta learning (MAML) algorithm so as to obtain the abnormal account identification model after the parameters are initialized.
Specifically, parameters of an abnormal account identification model which is preferred under a plurality of service scenes, namely shared initialization parameters, are determined according to a historical transaction data set by adopting an MAML algorithm, and the abnormal account identification model is initialized by adopting the shared initialization parameters so as to obtain the abnormal account identification model after the initialization parameters.
According to the abnormal account identification model training method, historical transaction data sets under multiple service scenes are obtained; the historical transaction data set comprises transaction data samples and category labels corresponding to a plurality of service scenes; the transaction data sample comprises historical characteristic data of an account under a corresponding service scene; and initializing parameters in the abnormal account identification model according to the historical transaction data set by adopting a model independent learning MAML algorithm so as to obtain the abnormal account identification model after the parameters are initialized. Because the parameters in the abnormal account identification model are initialized according to the historical transaction data set by adopting the model independent learning MAML algorithm, the typical characteristics of illegal behavior transactions in different business scenes are utilized, and the parameters in the abnormal account identification model after the parameters are initialized have strong enough adaptability. Only a small number of training samples and class labels of the target service scene and a very small number of iterations are needed, and the target abnormal account identification model can be obtained by slightly correcting the parameters in the abnormal account identification model after the parameters are initialized.
Optionally, an embodiment of initializing parameters in the abnormal account identification model according to the historical transaction data set by using a model independent meta learning MAML algorithm to obtain the abnormal account identification model after the parameters are initialized includes:
step S2021, the historical transaction data set is divided into a training data set and a testing data set.
The embodiment of the application does not limit the number of data included in the training data set and the test data set.
Exemplaryly, the obtained historical transaction data set is D = { D = { (D) i } i=1...n And n is the number of the plurality of service scenes. If the feature data of the account comprises 4 features, the historical transaction data set is divided into a training data set for inner loop adaptive trainingAnd test data set for outer loop finding shared initialization parametersWherein N is tr For training the length of the data set, N te For testing the length of the data set, v 1 ,...,v 4 Indicating the feature data to be input and L indicating the category label.
Step S2022, obtaining adaptive parameters corresponding to a plurality of service scenes according to the training data set by using an inner loop portion of the MAML algorithm.
Specifically, calculating a sub-scene loss function value of a transaction account detection model corresponding to the training data set aiming at each business scene; and acquiring the adaptive parameters corresponding to the service scene according to the scene loss function values by adopting a gradient descent algorithm.
The formula corresponding to the inner loop part of the MAML algorithm is:
wherein eta is the updating step length of the inner loop,to be based on a training data setDetermining the sub-scene loss function value corresponding to the service scene i, wherein m is the gradient descending time, i is the ith service scene,the adaptive parameter corresponding to the service scene i determined by the m-th gradient descent is shown,to differentiate Φ, the first gradient descent of the loss function is represented.
Step S2023, determining shared initialization parameters of the abnormal account identification model according to the test data set and the adaptive parameters corresponding to the plurality of service scenes by adopting an outer loop part of the MAML algorithm.
Specifically, adaptive parameters corresponding to a test data set and a plurality of service scenesAnd calculating a total loss function value, and acquiring a shared initialization parameter of the transaction account detection model by adopting a gradient descent algorithm. Wherein the total loss function value isFor the results of the inner loop according to the test data setUpdating results of the intra-scene circulation corresponding to the determined service scene iLoss function values that classify network parameters.
The formula for the outer loop portion of the MAML algorithm is:
wherein θ is a shared initialization parameter in the outer loop process, k is an outer loop step length, U is the number of gradient descent finally performed on the inner loop to obtain the adaptive parameter, and N is the number of the plurality of service scenes.
Step S2024, initialize the parameters in the abnormal account identification model to the shared initialization parameters.
Specifically, the determined shared initialization parameters are substituted into the abnormal account identification model to obtain the abnormal account identification model after the initialization parameters.
In the embodiment of the application, after the abnormal account identification model after the initialization parameters are determined, a small number of training samples and category labels of a target service scene are adopted, and a small number of rounds of updating training are performed, so that the parameters of the target abnormal account identification model can be obtained through rapid convergence, and the target abnormal account identification model is obtained. The concrete formula is as follows:
wherein,identify the parameters of the model, θ, for the target anomalous account * For the determined shared initialization parameter, γ is the step size of the test procedure update.
Fig. 3 is a schematic diagram of historical data set allocation provided in an embodiment of the present application, and as shown in fig. 3, the historical data set includes a target data set corresponding to a target service scenario and historical transaction data sets in a plurality of service scenarios; the historical transaction data set comprises historical transaction data from a service scene 1 to a service scene n; dividing historical transaction data of each business scene into training data setsAnd testing the data setThe target data set is divided into training samples and class labels, and testing samples and class labels.
Fig. 4 is a schematic diagram of a training method for an abnormal account recognition model according to an embodiment of the present application, and as shown in fig. 4, an MAML algorithm is used to determine shared initialization parameters as a meta-training network process, and training the abnormal account recognition model after initialization parameters to obtain a target abnormal account recognition model as a testing network process. In particular, the inner loop portion using the MAML algorithm is based on a training data setObtaining adaptive parameters corresponding to a plurality of service scenesThe outer loop portion using the MAML algorithm is based on the test data set D te And determining a shared initialization parameter theta of the abnormal account identification model by using adaptive parameters corresponding to a plurality of service scenes * . In determining the shared initialization parameter theta * Then, the training sample and class label D are adopted T Training the abnormal account identification model after the parameters are initialized to obtain the adaptive parameters of the target abnormal account identification modelAnd classifying the target account according to a classification function in the target abnormal account identification model, and acquiring a classification result.
Compared with the meta-training network process, the abnormal account identification model training method provided by the embodiment of the application is only oriented to a specific target service scene in the test network process, and the model initialization parameters are not randomly selected any more but are achieved in the meta-training network process. That is, the parameters are not updated in the process of testing the network, and a parameter with universality on a plurality of service scenes is selected for initialization. Therefore, only a small number of training samples and class labels of the target service scene and a very small number of iterations are needed, and the target abnormal account identification model can be obtained by slightly correcting the parameters in the abnormal account identification model after the parameters are initialized. The MAML algorithm does not completely need large data set updating parameters, but puts a long and resource-consuming process on the meta-training network process in the early stage. The learning efficiency of the target abnormal account identification model on the target business scene is improved by adopting the MAML.
EXAMPLE III
Fig. 5 is a flowchart of an abnormal account identification method provided in the third embodiment of the present application, and the embodiment of the present application provides an abnormal account identification method for solving the problems that the existing model rules cannot utilize typical characteristics common to illegal transactions in different business scenarios, and the accuracy of suspicious transaction detection is low. The method in this embodiment is applied to an abnormal account recognition device, and the abnormal account recognition device may be located in an electronic device. Among other things, the electronic device may be a digital computer that represents various forms. Such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers.
As shown in fig. 5, the method comprises the following specific steps:
step S301, obtaining current characteristic data of the target account in a target service scene.
Wherein the current feature data includes: the current attribute feature value and the current behavior feature value of the target account.
Step S302, inputting the current characteristic data into a target abnormal account identification model to classify the current characteristic data so as to identify whether the transaction behavior of the target account is abnormal or not.
The target abnormal account identification model is obtained by training the abnormal account identification model after the parameters are initialized by adopting the abnormal account identification model training method provided in the first embodiment or the second embodiment.
In the embodiment of the application, the mode of acquiring the current characteristic data of the target account in the target service scene is similar to the mode of acquiring the training sample of the target service scene. For example, transaction data and corresponding account characteristics of the target account within the current day and a preset number of days before the current day may be obtained, and the current characteristic data of the target account may be obtained through preprocessing the transaction data and the corresponding account characteristics.
In the embodiment of the application, after the current characteristic data is input into the target abnormal account identification model to be classified, the classification result of the target account can be obtained. If the classification result is abnormal, the target account can be determined to have abnormal transaction behavior; if the classification result is normal, it can be determined that no abnormal transaction behavior exists in the target account.
Optionally, if it is determined that the target account has an abnormal transaction behavior, a corresponding abnormal account control policy may be executed, for example, the abnormal account may be frozen, and the abnormal account may be determined specifically according to a requirement, which is not limited in this embodiment of the present application.
According to the abnormal account identification method provided by the embodiment of the application, the current characteristic data of the target account in the target service scene is obtained; the current feature data includes: the current attribute characteristic value and the current behavior characteristic value of the target account; and inputting the current characteristic data into a target abnormal account identification model to classify the current characteristic data so as to identify whether the transaction behavior of the target account is abnormal or not. Whether the target account is abnormal or not can be quickly determined, and the speed and accuracy of classifying the target account can be improved.
According to the abnormal account identification method provided by the embodiment of the application, the current characteristic data of the target account in the target service scene is obtained; the current feature data includes: the current attribute characteristic value and the current behavior characteristic value of the target account; and inputting the current characteristic data into a target abnormal account identification model to classify the current characteristic data so as to identify whether the transaction behavior of the target account is abnormal or not. The target account is classified by adopting the target abnormal account identification model, and whether suspicious transactions exist is determined according to whether the target account is abnormal, so that the accuracy of suspicious transaction detection can be improved.
Example four
Fig. 6 is a schematic structural diagram of an abnormal account recognition model training device according to a fourth embodiment of the present application. The abnormal account recognition model training device provided by the embodiment of the application can execute the processing flow provided by the abnormal account recognition model training method. As shown in fig. 6, the abnormal account recognition model training apparatus 40 includes: an acquisition module 401 and a training module 402.
Specifically, the obtaining module 401 is configured to obtain a training sample and a category label of a target service scene; the training sample comprises historical characteristic data of a historical account in a target business scene; the historical characteristic data comprises: attribute characteristic values and historical behavior characteristic values;
a training module 402, configured to input a training sample and a category label of a target service scene into the abnormal account identification model after the initialization parameters, so as to train the abnormal account identification model after the initialization parameters, and obtain a target abnormal account identification model, where the target abnormal account identification model is used in the target service scene to identify whether a target account is abnormal or not; the abnormal account identification model after the parameters are initialized is determined by adopting a model independent learning MAML algorithm, and the abnormal account identification models after the parameters are initialized corresponding to a plurality of service scenes are consistent.
The apparatus provided in this embodiment of the present application may be specifically configured to execute the method embodiment provided in the first embodiment, and specific functions are not described herein again.
Optionally, the abnormal account recognition model training device 40 further includes: an initial parameter determination module; the initial parameter determination module is configured to: acquiring historical transaction data sets under a plurality of service scenes; the historical transaction data set comprises transaction data samples and category labels corresponding to a plurality of service scenes; the transaction data sample comprises historical characteristic data of an account under a corresponding service scene; and initializing parameters in the abnormal account identification model according to the historical transaction data set by adopting a model independent meta learning (MAML) algorithm so as to obtain the abnormal account identification model after the parameters are initialized.
Optionally, the initial parameter determining module is specifically configured to: dividing a historical transaction data set into a training data set and a testing data set; acquiring adaptive parameters corresponding to a plurality of service scenes by adopting an inner loop part of an MAML algorithm according to a training data set; determining shared initialization parameters of an abnormal account identification model according to the test data set and adaptive parameters corresponding to a plurality of service scenes by adopting an outer loop part of an MAML algorithm; parameters in the abnormal account identification model are initialized to shared initialization parameters.
Optionally, the obtaining module 401 is specifically configured to obtain, from a database corresponding to a target service scenario, transaction data of a plurality of historical accounts within preset days, corresponding account characteristics, and category labels; the accounts with abnormal category labels are included in the plurality of accounts; extracting data characteristics of transaction data corresponding to each historical account to obtain historical behavior characteristic values of a plurality of historical accounts in a target service scene; quantifying the account characteristics corresponding to the historical accounts to obtain attribute characteristic values of the historical accounts in a target service scene; and determining the historical behavior characteristic value and the attribute characteristic value corresponding to each historical account as each training sample, and determining the class label corresponding to each historical account as the class label corresponding to the training sample.
Optionally, the abnormal account identification model after the parameters are initialized is a deep neural network DNN model after the parameters are initialized; a training module 402, specifically configured to input a training sample and a class label of a target service scene into the deep neural network DNN model after the initialization parameters; training the deep neural network DNN model with the initialized parameters by adopting a training sample and a class label of a target service scene; judging whether a loss function in the trained DNN model reaches a preset threshold value or not; and if the loss function is determined to reach the preset threshold value, determining the DNN model with the minimum loss function as the target abnormal account identification model.
The apparatus provided in this embodiment of the present application may be specifically configured to execute the first method embodiment or the second method embodiment, and specific functions are not described herein again.
Fig. 7 is a schematic structural diagram of an abnormal account identification apparatus according to a fourth embodiment of the present application. The abnormal account identification device provided by the embodiment of the application can execute the processing flow provided by the abnormal account identification method. As shown in fig. 7, the abnormal account identification apparatus 50 includes: an obtaining module 501 and a classifying module 502.
Specifically, the obtaining module 501 is configured to obtain current feature data of a target account in a target service scenario; the current feature data includes: the current attribute characteristic value and the current behavior characteristic value of the target account.
The classification module 502 is used for inputting the current characteristic data into the target abnormal account identification model to classify the current characteristic data so as to identify whether the transaction behavior of the target account is abnormal or not; the target abnormal account recognition model is obtained by training the abnormal account recognition model after the parameters are initialized by adopting the abnormal account recognition model training method provided in the third embodiment.
The apparatus provided in this embodiment of the present application may be specifically configured to execute the method embodiment provided in the third embodiment, and specific functions are not described herein again.
EXAMPLE five
Fig. 8 is a schematic structural diagram of an electronic device according to a fifth embodiment of the present application, and as shown in fig. 8, the present application further provides an electronic device 60, including: memory 601, processor 602.
The memory 601 is used for storing computer executable instructions and is connected to the processor 602 in communication. In particular, the program may include program code comprising computer-executable instructions. Memory 601 may comprise high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk memory.
A processor 602 for executing computer-executable instructions stored by the memory 601.
Wherein computer executable instructions are stored in the memory 601 and configured to be executed by the processor 602 to implement the method provided by any one of the embodiments of the present application. The related description may be understood by referring to the related description and effect corresponding to the steps in the drawings, and redundant description is not repeated here.
In the embodiment of the present application, the memory 601 and the processor 602 are connected by a bus. The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, an Extended ISA (Extended Industry Standard Architecture) bus, or the like. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of illustration, only one thick line is shown in FIG. 8, but that does not indicate only one bus or one type of bus.
The embodiment of the present application further provides a computer-readable storage medium, in which computer execution instructions are stored, and the computer execution instructions are executed by a processor to implement the method provided by any one of the embodiments of the present application.
The embodiment of the present application further provides a computer program product, which includes computer executable instructions, and when the computer executable instructions are executed by a processor, the method provided in any embodiment of the present application is implemented.
In the several embodiments provided in the present application, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, a division of modules is merely a division of logical functions, and an actual implementation may have another division, for example, a plurality of modules or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or modules, and may be in an electrical, mechanical or other form.
Modules described as separate parts may or may not be physically separate, and parts displayed as modules may or may not be physical modules, may be located in one place, or may be distributed on a plurality of network modules. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of the present embodiment.
In addition, functional modules in the embodiments of the present application may be integrated into one processing module, or each of the modules may exist alone physically, or two or more modules are integrated into one module. The integrated module can be realized in a hardware mode, and can also be realized in a mode of hardware and a software functional module.
Program code for implementing the methods of the present application may be written in any combination of one or more programming languages. These program codes may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable full path trajectory fusion device such that the program codes, when executed by the processor or controller, cause the functions/operations specified in the flowchart and/or block diagram to be performed. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.
In the context of this application, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
Further, while operations are depicted in a particular order, this should be understood as requiring that such operations be performed in the particular order shown or in sequential order, or that all illustrated operations be performed, to achieve desirable results. Under certain circumstances, multitasking and parallel processing may be advantageous. Likewise, while several specific implementation details are included in the above discussion, these should not be construed as limitations on the scope of the application. Certain features that are described in the context of separate embodiments can also be implemented in combination in a single implementation. Conversely, various features that are described in the context of a single implementation can also be implemented in multiple implementations separately or in any suitable subcombination.
Other embodiments of the present application will be apparent to those skilled in the art from consideration of the specification and practice of the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the invention following, in general, the principles of the application and including such departures from the present disclosure as come within known or customary practice within the art to which the invention pertains. It is intended that the specification and examples be considered as exemplary only, with a true scope and spirit of the application being indicated by the following claims.
It will be understood that the present application is not limited to the precise arrangements described above and shown in the drawings and that various modifications and changes may be made without departing from the scope thereof. The scope of the application is limited only by the appended claims.
Claims (10)
1. An abnormal account recognition model training method is characterized by comprising the following steps:
acquiring a training sample and a category label of a target service scene; the training sample comprises historical characteristic data of a historical account in a target business scene; the historical characteristic data comprises: attribute characteristic values and historical behavior characteristic values;
inputting training samples and class labels of a target business scene into the abnormal account identification model after the initialization parameters so as to train the abnormal account identification model after the initialization parameters and obtain a target abnormal account identification model, wherein the target abnormal account identification model is used for identifying whether a target account is abnormal or not in the target business scene; the abnormal account identification models after the initialization parameters are determined by adopting a model independent learning MAML algorithm, and the abnormal account identification models after the initialization parameters corresponding to a plurality of service scenes are consistent.
2. The method of claim 1, wherein determining the abnormal account identification model after the initialization parameters by using a model independent meta learning (MAML) algorithm comprises:
acquiring historical transaction data sets under a plurality of service scenes; the historical transaction data set comprises transaction data samples and category labels corresponding to a plurality of service scenes; the transaction data sample comprises historical characteristic data of an account under a corresponding service scene;
and initializing parameters in the abnormal account identification model according to the historical transaction data set by adopting a model independent learning MAML algorithm so as to obtain the abnormal account identification model after the parameters are initialized.
3. The method of claim 2, wherein initializing parameters in the abnormal account identification model according to the historical transaction data set by using a model independent learning (MAML) algorithm to obtain an abnormal account identification model after initializing parameters comprises:
dividing the historical transaction data set into a training data set and a testing data set;
acquiring adaptive parameters corresponding to a plurality of service scenes according to the training data set by adopting an inner loop part of an MAML algorithm;
determining shared initialization parameters of an abnormal account identification model according to the test data set and adaptive parameters corresponding to the plurality of service scenes by adopting an outer loop part of an MAML algorithm;
initializing parameters in the abnormal account identification model to be shared initialization parameters.
4. The method of claim 1, wherein the obtaining of the training samples and class labels of the target business scenario comprises:
acquiring transaction data of a plurality of historical accounts within preset days, corresponding account characteristics and category labels from a database corresponding to a target service scene; the accounts with abnormal category labels are included in the plurality of accounts;
extracting data characteristics of transaction data corresponding to each historical account to obtain historical behavior characteristic values of a plurality of historical accounts in a target service scene;
carrying out quantification processing on account characteristics corresponding to the historical accounts to obtain attribute characteristic values of the historical accounts in a target service scene;
and determining the historical behavior characteristic value and the attribute characteristic value corresponding to each historical account as each training sample, and determining the class label corresponding to each historical account as the class label corresponding to the training sample.
5. The method according to any one of claims 1-4, wherein the abnormal account identification model after the initialization parameters is a Deep Neural Network (DNN) model after the initialization parameters;
the method for inputting the training sample and the class label of the target service scene into the abnormal account recognition model after the initialization parameters to train the abnormal account recognition model after the initialization parameters and obtain the target abnormal account recognition model includes:
inputting training samples and class labels of a target service scene into the deep neural network DNN model after the initialization parameters;
training the deep neural network DNN model after the initialization parameters by adopting a training sample and a class label of a target service scene;
judging whether a loss function in the trained DNN model reaches a preset threshold value or not;
and if the loss function is determined to reach the preset threshold, determining the DNN model with the minimum loss function as the target abnormal account identification model.
6. An abnormal account identification method is characterized by comprising the following steps:
acquiring current characteristic data of a target account in a target service scene; the current feature data includes: the current attribute characteristic value and the current behavior characteristic value of the target account;
inputting the current characteristic data into a target abnormal account identification model to classify the current characteristic data so as to identify whether the transaction behavior of a target account is abnormal or not; the target abnormal account identification model is obtained by training the abnormal account identification model after the parameters are initialized by adopting the abnormal account identification model training method as claimed in any one of claims 1 to 5.
7. An abnormal account recognition model training device, comprising:
the acquisition module is used for acquiring a training sample and a category label of a target business scene; the training sample comprises historical characteristic data of a historical account in a target business scene; the historical characteristic data comprises: attribute characteristic values and historical behavior characteristic values;
the training module is used for inputting training samples and class labels of a target business scene into the abnormal account identification model after the initialization parameters so as to train the abnormal account identification model after the initialization parameters and obtain a target abnormal account identification model, and the target abnormal account identification model is used for identifying whether a target account is abnormal or not in the target business scene; the abnormal account identification models after the initialization parameters are determined by adopting a model independent learning MAML algorithm, and the abnormal account identification models after the initialization parameters corresponding to a plurality of service scenes are consistent.
8. An abnormal account identification apparatus, comprising:
the acquisition module is used for acquiring current characteristic data of the target account in a target service scene; the current feature data includes: the current attribute characteristic value and the current behavior characteristic value of the target account;
the classification module is used for inputting the current characteristic data into a target abnormal account identification model to classify the current characteristic data so as to identify whether the transaction behavior of the target account is abnormal or not; the target abnormal account identification model is obtained by training the abnormal account identification model after the parameters are initialized by adopting the abnormal account identification model training method as claimed in any one of claims 1 to 5.
9. An electronic device, comprising: a processor, and a memory communicatively coupled to the processor;
the memory stores computer-executable instructions;
the processor executes computer-executable instructions stored by the memory to implement the method of any one of claims 1-5 or claim 6.
10. A computer-readable storage medium having computer-executable instructions stored therein, which when executed by a processor, are configured to implement the method of any one of claims 1-5 or claim 6.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211202721.1A CN115438747A (en) | 2022-09-29 | 2022-09-29 | Abnormal account recognition model training method, device, equipment and medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211202721.1A CN115438747A (en) | 2022-09-29 | 2022-09-29 | Abnormal account recognition model training method, device, equipment and medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115438747A true CN115438747A (en) | 2022-12-06 |
Family
ID=84250694
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211202721.1A Pending CN115438747A (en) | 2022-09-29 | 2022-09-29 | Abnormal account recognition model training method, device, equipment and medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115438747A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116170829A (en) * | 2023-04-26 | 2023-05-26 | 浙江省公众信息产业有限公司 | Operation and maintenance scene identification method and device for independent private network service |
-
2022
- 2022-09-29 CN CN202211202721.1A patent/CN115438747A/en active Pending
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116170829A (en) * | 2023-04-26 | 2023-05-26 | 浙江省公众信息产业有限公司 | Operation and maintenance scene identification method and device for independent private network service |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US10943186B2 (en) | Machine learning model training method and device, and electronic device | |
WO2021174966A1 (en) | Risk identification model training method and apparatus | |
CN107391569B (en) | Data type identification, model training and risk identification method, device and equipment | |
US20220094709A1 (en) | Automatic Machine Learning Vulnerability Identification and Retraining | |
CN111898758A (en) | User abnormal behavior identification method and device and computer readable storage medium | |
CN111310814A (en) | Method and device for training business prediction model by utilizing unbalanced positive and negative samples | |
KR102168198B1 (en) | Business default prediction system and operation method thereof | |
CN103339635A (en) | Determining the vulnerability of computer software applications to privilege-escalation attacks | |
CN113408558B (en) | Method, apparatus, device and medium for model verification | |
CN113468520A (en) | Data intrusion detection method applied to block chain service and big data server | |
CN113657896A (en) | Block chain transaction topological graph analysis method and device based on graph neural network | |
CN112884569A (en) | Credit assessment model training method, device and equipment | |
CN112330355A (en) | Consumption ticket transaction data processing method, device, equipment and storage medium | |
CN114140670B (en) | Method and device for verifying ownership of model based on exogenous characteristics | |
CN115438747A (en) | Abnormal account recognition model training method, device, equipment and medium | |
CN116823428A (en) | Anti-fraud detection method, device, equipment and storage medium | |
CN111145026A (en) | Anti-money laundering model training method and device | |
CN111639688B (en) | Local interpretation method of Internet of things intelligent model based on linear kernel SVM | |
CN115935265B (en) | Method for training risk identification model, risk identification method and corresponding device | |
CN112508684A (en) | Joint convolutional neural network-based collection risk rating method and system | |
Kshatri et al. | Analytical study of some selected classification algorithms and crime prediction | |
Laimek et al. | ATM fraud detection using outlier detection | |
CN112742026B (en) | Game control method, game control device, storage medium and electronic equipment | |
CN114140246A (en) | Model training method, fraud transaction identification method, device and computer equipment | |
CN113052604A (en) | Object detection method, device, 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 |