CN112950221B - Method and device for establishing wind control model and risk control method and device - Google Patents
Method and device for establishing wind control model and risk control method and device Download PDFInfo
- Publication number
- CN112950221B CN112950221B CN202110323751.7A CN202110323751A CN112950221B CN 112950221 B CN112950221 B CN 112950221B CN 202110323751 A CN202110323751 A CN 202110323751A CN 112950221 B CN112950221 B CN 112950221B
- Authority
- CN
- China
- Prior art keywords
- wind control
- target user
- control model
- sample set
- general wind
- 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
- 238000000034 method Methods 0.000 title claims abstract description 71
- 238000012954 risk control Methods 0.000 title claims abstract description 49
- 238000012549 training Methods 0.000 claims abstract description 20
- 210000002569 neuron Anatomy 0.000 claims description 51
- 230000006399 behavior Effects 0.000 claims description 37
- 230000006835 compression Effects 0.000 claims description 26
- 238000007906 compression Methods 0.000 claims description 26
- 238000012545 processing Methods 0.000 claims description 10
- 238000004590 computer program Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 description 14
- 238000010586 diagram Methods 0.000 description 8
- 238000012546 transfer Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000013140 knowledge distillation Methods 0.000 description 2
- 238000013138 pruning Methods 0.000 description 2
- 238000011002 quantification Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q20/00—Payment architectures, schemes or protocols
- G06Q20/38—Payment protocols; Details thereof
- G06Q20/40—Authorisation, e.g. identification of payer or payee, verification of customer or shop credentials; Review and approval of payers, e.g. check credit lines or negative lists
- G06Q20/401—Transaction verification
- G06Q20/4016—Transaction verification involving fraud or risk level assessment in transaction processing
-
- 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
- G06N3/082—Learning methods modifying the architecture, e.g. adding, deleting or silencing nodes or connections
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Accounting & Taxation (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Security & Cryptography (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Transfer Between Computers (AREA)
Abstract
The embodiment of the specification provides a method for establishing a wind control model, a risk control method and a risk control device. In the method, a general wind control model is obtained firstly; the general wind control model is obtained by training through a first sample set, and the first sample set is generated through behavior data of at least two types of users; generating a second sample set corresponding to the target user by using the behavior data of the target user; and training the general wind control model by using the second sample set to obtain an individualized wind control model corresponding to the target user.
Description
Technical Field
One or more embodiments of the present specification relate to electronic information technology, and more particularly, to a method and apparatus for creating a wind control model and a risk control method and apparatus.
Background
With the development of internet services, more and more user services are implemented based on networks, for example, a user pays through online shopping, and for example, the user performs online bank transfer. When the user service is realized, risk control is often required, for example, when the user is judged to be a low-risk user through the wind control model, the user can complete payment without inputting a password, or the user can consume the low-risk user first, and then pay separately when the subsequent user is convenient.
However, the existing risk control method cannot provide better wind control service according to the user requirements.
Disclosure of Invention
One or more embodiments of the present specification describe a method and an apparatus for establishing a wind control model and a risk control method and apparatus, which can provide better wind control service according to user requirements.
According to a first aspect, there is provided a method of building a wind control model, comprising:
acquiring a general wind control model; the general wind control model is obtained by training through a first sample set, and the first sample set is generated through behavior data of at least two types of users;
generating a second sample set corresponding to the target user by using the behavior data of the target user;
and training the general wind control model by using the second sample set to obtain an individualized wind control model corresponding to the target user.
Wherein the feature space of the first sample set is the same as the feature space of the second sample set.
Preferably, before obtaining the universal wind control model, the method further comprises:
respectively compressing the original general wind control models obtained by training the first sample set according to preset compression requirements of each grade to obtain compressed general wind control models corresponding to each grade; each level corresponds to one type of terminal equipment configuration;
acquiring configuration information of the terminal equipment used by the target user; and
determining a target grade according to the acquired configuration information;
then, the obtained general wind control model and the general wind control model trained by using the second sample set are: and the compressed general wind control model corresponding to the target level.
The method for respectively compressing the original general wind control models obtained by training the first sample set according to the preset compression requirements of each grade comprises the following steps:
for each level, performing:
determining the number N of the neurons needing to be deleted according to the compression requirement of the level and the number M of the neurons used by the original general wind control model;
selecting N neurons from M neurons used by the original universal wind control model according to the configuration of the terminal equipment corresponding to the grade or according to the sequence from the first layer of neurons to the last layer of neurons;
deleting the N selected neurons to obtain a compressed general wind control model corresponding to the level;
wherein M and N are positive integers greater than 0, and M is greater than or equal to N.
The method is applied to the terminal equipment;
or,
the method is applied to a server, and the method further comprises the following steps: and issuing the individualized wind control model corresponding to the target user to the terminal equipment of the target user.
According to a second aspect, there is provided a risk control method comprising:
determining a risk control decision of the target user according to the obtained safety appeal data of the target user;
performing risk control on the target user according to the risk control decision of the target user and the risk value output by the personalized wind control model of the target user; the target user personalized wind control model is obtained by using the method for establishing the wind control model provided by the embodiment of the specification.
The method for acquiring the safety appeal data of the target user comprises the following steps:
providing a questionnaire about security appeal to a target user, and acquiring security appeal data of the target user based on an answer of the questionnaire of the target user;
and/or the presence of a gas in the gas,
and modifying the record and/or the complaint data according to the password of the target user to obtain the safety appeal data of the target user.
The method is applied to the terminal equipment or the server.
According to a third aspect, there is provided an apparatus for creating a wind control model, comprising:
the general wind control model acquisition module is configured to acquire a general wind control model; the general wind control model is obtained by training through a first sample set, and the first sample set is generated through behavior data of at least two types of users;
the personalized sample generation module is configured to generate a second sample set corresponding to the target user by utilizing the behavior data of the target user;
and the personalized wind control model acquisition module is configured to train the general wind control model by using the second sample set to obtain a personalized wind control model corresponding to the target user.
Wherein the feature space of the first sample set is the same as the feature space of the second sample set.
Further comprising:
the compression module is configured to respectively compress the original general wind control models obtained by training the first sample set according to preset compression requirements of each grade to obtain compressed general wind control models corresponding to each grade; each level corresponds to one type of terminal equipment configuration;
the issuing module is configured to acquire configuration information of the terminal equipment used by the target user and determine a target grade according to the acquired configuration information; and issuing the compressed general wind control model corresponding to the target grade to the general wind control model acquisition module.
The compression module is configured to perform, for each level:
determining the number N of the neurons needing to be deleted according to the compression requirement of the level and the number M of the neurons used by the original general wind control model;
selecting N neurons from M neurons used by the original general wind control model according to the configuration of the terminal equipment corresponding to the grade or the sequence from the first layer neuron to the last layer neuron;
deleting the N selected neurons to obtain a compressed general wind control model corresponding to the level;
wherein M and N are positive integers greater than 0, and M is greater than or equal to N.
According to a fourth aspect, there is provided a risk control device comprising:
the personalized wind control decision acquisition module is configured to determine a risk control decision of the target user according to pre-acquired security appeal data of the target user;
the risk processing module is configured to carry out risk control on the target user according to a risk control decision of the target user and a risk value output by the personalized wind control model of the target user; the personalized wind control model of the target user is sent by a device for establishing the wind control model in the embodiment of the specification.
Wherein the personalized wind control decision acquisition module is further configured to: providing a questionnaire about security appeal to a target user, and acquiring security appeal data of the target user based on an answer of the questionnaire of the target user; and/or modifying the record and/or the complaint data according to the password of the target user to obtain the security appeal data of the target user.
According to a fifth aspect, there is provided a computer readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method of the first or second aspect.
According to a sixth aspect, there is provided a computing device comprising a memory having stored therein executable code and a processor that when executing the executable code implements the method of the first or second aspect.
According to the method and the device provided by the embodiment of the specification, firstly, behavior data of various types of users are used for generating a first sample set, the first sample set can guarantee enough sample quantity and sample types, the general wind control model trained according to the first sample set is more general, then, personalized behavior data of a target user are used for generating a second sample set, the second sample set can better accord with behavior data characteristics of the target user, the general wind control model is adjusted according to the second sample set, and therefore the personalized wind control model corresponding to the target user is obtained finally. The finally obtained wind control model can meet the requirements on the number of samples and the universality of the model, can meet the personalized requirements of the user, and better meets the personalized behavior characteristics of the user. Subsequently, a more accurate risk value of the user can be calculated according to the wind control model.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a schematic diagram of a system architecture to which one embodiment of the present description is applied.
FIG. 2 illustrates a flow diagram of a method of building a wind control model according to one embodiment of the present description.
FIG. 3 is a flow diagram of a method for compressing an original generic model by a server in one embodiment of the present disclosure.
FIG. 4 illustrates a flow diagram of a risk control method according to one embodiment of the present description.
Fig. 5 is a schematic structural diagram of an apparatus for creating a wind control model according to an embodiment of the present disclosure.
Fig. 6 is a schematic structural diagram of an apparatus for creating a wind control model according to another embodiment of the present disclosure.
Fig. 7 shows a schematic structural diagram of a risk control device according to one embodiment of the present description.
Detailed Description
The scheme provided by the specification is described in the following with reference to the attached drawings.
In the prior art, a server establishes a universal wind control model by using strong calculation power of the server according to acquired sample data of various users, and a subsequent server calculates and evaluates risks of the various users based on the universal wind control model and outputs wind control decisions.
It can be seen that the current wind control model is a 'one thousand people one model', that is, different users use the same wind control model, and the wind control model does not consider the difference of different users in the aspect of behavior data. For example, the user group a uses the payroll system to transfer a large amount of money but does not play the game "ant garden", and the user group B does not use the payroll system to transfer a large amount of money but plays the game "ant garden". According to the method in the prior art, when the wind control model is trained, the wind control model is trained by using sample data of various types of users, including sample data for the user group a and sample data for the user group B. When the wind control model is trained, the emphasis training cannot be performed on different users, and the behavior data of the users cannot be considered more accurately, so that the risk value is not accurate enough.
In order to solve the problems in the prior art, the fact that the wind control model is trained by utilizing the personal behavior data of the user can be considered, so that the trained wind control model can better accord with the behavior characteristics of the user, and the subsequently predicted risk value can be more accurate. Meanwhile, if the wind control model is trained by only using the personal behavior data of the user, the sample data is insufficient, and the universality of the trained wind control model is poor, so that the universal wind control model can be trained by using the behavior data of various types of users, and then the universal wind control model is trained again by using the personalized behavior data of the target user, so that the finally obtained wind control model can meet the requirements on the sample and the universality, can meet the personalized requirements of the user, and better meets the personalized behavior characteristics of the user. Subsequently, a more accurate risk value of the user can be calculated according to the wind control model.
Specific implementations of the above concepts are described below.
To facilitate understanding of the present specification, a system architecture to which the present specification applies will be described first. As shown in fig. 1, the system architecture mainly includes a server and at least two terminal devices of at least two users.
The server interacts with the terminal equipment of each user through the network. The network may include various types of connections, such as wire, wireless communication links, or fiber optic cables, among others. The terminal device may be an intelligent device located at a user end, such as a mobile phone, a notebook computer, a personal digital assistant, and the like. The server may be a device arranged at the operator side for risk control.
FIG. 2 illustrates a flow diagram of a method of building a wind control model according to one embodiment. The execution subject of the method is a device for establishing a wind control model. In one embodiment of the present specification, the apparatus for creating a wind control model may be provided in the aforementioned server, or may be provided in the aforementioned terminal device. It is to be understood that the method may be performed by any computing, processing capable apparatus, device, platform, cluster of devices. Referring to fig. 2, the method includes:
step 201: acquiring a general wind control model; the general wind control model is obtained by training through a first sample set, and the first sample set is generated through behavior data of at least two types of users.
Step 203: and generating a second sample set corresponding to the target user by using the behavior data of the target user.
Step 205: and training the general wind control model by using the second sample set to obtain an individualized wind control model corresponding to the target user.
In the process shown in fig. 2, first, a first sample set is generated by using behavior data of various types of users, the first sample set can ensure sufficient sample quantity and sufficient sample types, the general wind control model trained according to the first sample set is highly accurate and is more general, and then, a second sample set is generated by using personalized behavior data of a target user, the second sample set can better meet behavior data characteristics of the target user, and the general wind control model is trained again and adjusted according to the second sample set, so that the wind control model better meets personalized requirements. The finally obtained wind control model can meet the requirements on the number of samples and the universality of the model, can meet the personalized requirements of the user, and better meets the personalized behavior characteristics of the user. Subsequently, a more accurate risk value of the user can be calculated according to the wind control model.
As described above, the execution subject of each step shown in fig. 2 may be a server or a terminal device.
When the execution subject is a server, the process of step 201 actually includes: the server collects behavior data of at least two types of users, such as behavior data of the user group A and behavior data sets of the user group B; generating a first sample set by using behavior data of at least two types of users; training the general wind control model by using the first sample set, thereby obtaining the general wind control model; in step 203, the server receives the behavior data of the target user reported by the target user through the terminal device thereof, and generates a second sample set corresponding to the target user; in step 205, after obtaining the personalized wind control model corresponding to the target user, the server issues the personalized wind control model of the target user to the terminal device of the target user. And subsequently, the terminal equipment calculates the risk value by using the wind control model and carries out wind control decision processing.
When the execution subject is a terminal device, in step 201, the terminal device receives the general wind control model issued by the server, so as to obtain the general wind control model. The training process of the general wind control model is still carried out in the server, that is, the server generates a first sample set and trains the general wind control model by using the first sample set; in step 203, the terminal device collects behavior data of a target user using the terminal device and generates a second sample set only reflecting behavior data characteristics of the target user; in step 205, the terminal device finally obtains an individualized wind control model corresponding to the target user.
In one embodiment of the present specification, the feature space of the first sample set is the same as the feature space of the second sample set. That is, the data type of the data included in the first sample set is the same as the data type of the data included in the second sample set. For example, the first sample set includes 1 thousand types of data of 100 ten thousand users (e.g., 1 thousand types of data such as large amount transfer data, data of game "ant garden", online shopping data, credit card payment data, etc.), and the second sample set includes 1 thousand types of data of the same type of data of the target user, i.e., 1 user, except that if the target user does not have a certain type of data record, the data value of the data corresponding to the type of data in the second sample set is 0, for example, for the target user in the user group a, the data value of the data type of game "ant garden" in the second sample set is 0. The characteristic space of the first sample set is the same as that of the second sample set, so that the original parameter distance between the trained personalized wind control model and the general model is not too far, but the personalized requirements of the target user are met.
When the process shown in fig. 2 is executed by the terminal device, because the processing power of the terminal device, such as the computing power, is lower than that of the server, there may be a case where the terminal device cannot directly run the original common model trained by the server. Therefore, in an embodiment of this specification, the server may compress the original general model trained by using the first sample set, and send the compressed general model to the terminal device of the target user, so that, in step 201, the terminal device obtains the compressed general model, thereby completing hardware adaptation, ensuring that the terminal device has sufficient processing capability to run the compressed general model, and completing subsequent risk control processing.
In one embodiment of the present specification, before step 201, a process of compressing, by a server, an original generic model to obtain a generic model suitable for a terminal device of a target user may be referred to in fig. 3, including:
step 301: at least two levels and compression requirements of each level are preset, wherein each level corresponds to one type of terminal equipment configuration.
For example, 10 types of terminal device configurations may be categorized according to configuration information (hardware configuration information and/or software configuration information) of various terminal devices searched in history, and different types of terminal device configurations embody processing capabilities of different terminal devices. Therefore, for the 10 types of terminal device configurations, corresponding 10 levels can be divided, and the compression requirement of each level is different. For example, the terminal device configuration type 1 corresponds to the level 1, the level 1 corresponds to the compression requirement 1 "to compress the original general wind control model to 10%", the terminal device configuration type 2 corresponds to the level 2, the level 2 corresponds to the compression requirement 2 "to compress the original general wind control model to 20%", and so on, the terminal device configuration type 10 corresponds to the level 10, and the level 10 corresponds to the compression requirement 10 "to compress the original general wind control model to 100%" (that is, the original general wind control model is maintained).
Step 303: the server trains the original generic wind control model using the first set of samples.
Step 305: and the server compresses the original general wind control model in grades according to the preset compression requirements of each grade to obtain the compressed general wind control model corresponding to each grade.
For example, in this step, 10 compressed general wind control models corresponding to 10 levels are obtained.
Step 307: the server obtains the configuration information of the terminal device used by the target user.
Step 309: and the server determines the target level according to the acquired configuration information.
Step 311: and the server issues the compressed general wind control model corresponding to the target grade to the terminal equipment of the target user.
For example, according to the configuration information of the terminal device used by the target user, the corresponding terminal device configuration type 2 is determined, the target level corresponding to the terminal device configuration type 2 is level 2, and the server issues the general wind control model which corresponds to the level 2 and is compressed to 20% to the terminal device.
Through the processing of fig. 3, in the process shown in fig. 2, the generic wind control models processed by the terminal device (the obtained generic wind control model and the generic wind control model trained by using the second sample set) are: a compressed generic wind control model corresponding to the target level, i.e. a model adapted to the configuration information of the terminal device.
In one embodiment of the present specification, the configuration information of the terminal device referred to in fig. 3 may only include hardware configuration information of the terminal device, such as information of CPU, memory brand, model and the like of the terminal device. The server may obtain the hardware configuration information of the terminal device when the terminal device is active, for example, when a user logs in or registers with the terminal device.
In an embodiment of this specification, in the step 305, for each preset level, the original generic wind control model is compressed, and a specific implementation process for each level may include:
step 3051: and determining the number N of the neurons needing to be deleted according to the compression requirement of the level and the number M of the neurons used by the original universal wind control model.
For example, if the level 9 requires compression to 90%, and the number of neurons used in the original generic wind control model is 1 ten thousand, it is determined that the number of neurons that need to be deleted is 1 thousand.
Step 3052: and selecting N neurons from M neurons used by the original general wind control model according to a preset strategy.
Here, the preset strategy may be determined according to actual business requirements, so as to decide which neurons in the artificial neural network used by the generic wind control model to delete. For example, which neurons are to be deleted may be determined according to the terminal device configuration (such as a CPU, memory information, and the like) corresponding to the current level, that is, the process in this step is: and selecting N neurons from M neurons used by the original universal wind control model according to the terminal equipment configuration corresponding to the grade.
Of course, the selection and deletion can also be directly performed according to the sequence of the number of layers of the neurons, and then the process of the step is as follows: and selecting M neurons used by the original universal wind control model according to the sequence from the first layer of neurons to the last layer of neurons until N neurons needing to be deleted are selected.
Step 3053: deleting the N selected neurons to obtain a compressed general wind control model corresponding to the level;
wherein M and N are positive integers greater than 0, and M is greater than or equal to N.
In other embodiments of the present disclosure, step 305 compresses the original wind control model, and other methods such as low-rank Approximation (low-rank Approximation), Network pruning (Network pruning), Network quantification (Network quantification), knowledge distillation (knowledge distillation), compact Network design (compact Network design), etc. may also be used.
In the above embodiments of the present specification, an individualized wind control model for a target user is established, and subsequently, wind control data of the target user may be calculated by using the individualized wind control model to obtain a risk value of the target user. After the risk value of the target user is obtained, the target user needs to be subjected to risk control by using a preset risk control decision.
In the prior art, the risk control decision is also 'one thousand people for one model', that is, different users use the wind control decision obtained by the same wind control judgment standard. For example, when the user risk value calculated by the wind control model is 0.4 and the user is a low-risk user, the same standard is adopted regardless of the user security appeal, and the face authentication is not performed again.
However, the security appeal varies from user to user. For example, a user with a high security appeal hopes that a face authentication interface appears in the system every time a password is modified, so that identity authentication can be performed conveniently; when a user with moderate security appeal modifies a password every time, a face authentication interface appears when the system judges that the user is at risk; users with low security appeal want to improve efficiency and avoid cumbersome operations, so that a face authentication interface appears only when the system judges a high risk every time a password is modified. In the prior art, the safety appeal of the user is not considered, so that better risk control service cannot be improved according to the personalized requirement of the user.
To solve this problem, in an embodiment of the present specification, a risk control decision of each user may be determined according to the obtained security appeal data of the user, that is, the risk control decision of different users is different. On the basis of fig. 2, after the personalized wind control model of the target user is obtained, the risk control decision of the target user can be determined according to the pre-obtained security appeal data of the target user, so that the wind control model and the risk control decision used in the risk control process are the personalized wind control model and the personalized risk control decision determined according to the behavior data of the target user, and therefore the finally obtained risk control result is a personalized wind control result according with the characteristics of the behavior data of the user, the user requirements are met better, and better wind control service can be improved.
In one embodiment of the present description, referring to fig. 4, a risk control method includes:
step 401: and determining a risk control decision of the target user according to the obtained safety appeal data of the target user.
Step 403: performing risk control on the target user according to the risk control decision of the target user and the risk value output by the personalized wind control model of the target user; the personalized wind control model of the target user is obtained by the method of any embodiment in the specification.
In step 401, the method for acquiring the security appeal data of the target user may include:
providing a questionnaire about security appeal to a target user, and acquiring security appeal data of the target user based on an answer of the questionnaire of the target user;
and/or the presence of a gas in the atmosphere,
and modifying the record and/or the complaint data according to the password of the target user to obtain the safety appeal data of the target user.
When the process of creating the wind control model in fig. 2 is executed by the server, the risk control process in fig. 4 may be executed by the server, or may be executed by the terminal device of the target user. When the process of establishing the wind control model in fig. 2 is performed by the terminal device, the risk control process in fig. 4 may be performed by the terminal device of the target user.
An embodiment of the present specification further provides an apparatus for creating a wind control model, and referring to fig. 5, the apparatus 500 includes:
a general wind control model obtaining module 501 configured to obtain a general wind control model; the general wind control model is obtained by training through a first sample set, and the first sample set is generated through behavior data of at least two types of users;
a personalized sample generation module 502 configured to generate a second sample set corresponding to a target user by using behavior data of the target user;
an individualized wind control model obtaining module 503 is configured to train the general wind control model by using the second sample set, so as to obtain an individualized wind control model corresponding to the target user.
In one embodiment of the apparatus of the present specification, the feature space of the first sample set is the same as the feature space of the second sample set.
In one embodiment of the apparatus of the present disclosure, referring to fig. 6, further comprising:
the compression module 601 is configured to compress the original general wind control models obtained by training with the first sample set according to preset compression requirements of each grade to obtain compressed general wind control models corresponding to each grade; each level corresponds to one type of terminal equipment configuration;
an issuing module 602 configured to acquire configuration information of the terminal device used by the target user, and determine a target level according to the acquired configuration information; and issuing the compressed general wind control model corresponding to the target level to the general wind control model obtaining module 501.
In one embodiment of the apparatus of the present specification, the compression module 601 is configured to perform, for each level:
determining the number N of the neurons needing to be deleted according to the compression requirement of the level and the number M of the neurons used by the original general wind control model;
selecting N neurons from M neurons used by the original universal wind control model according to the configuration of the terminal equipment corresponding to the grade or according to the sequence from the first layer of neurons to the last layer of neurons;
deleting the N selected neurons to obtain a compressed general wind control model corresponding to the level;
wherein M and N are positive integers greater than 0, and M is greater than or equal to N.
An embodiment of the present specification further provides a risk control device, and referring to fig. 7, the device 700 includes:
the personalized wind control decision acquisition module 701 is configured to determine a risk control decision of a target user according to pre-acquired security appeal data of the target user;
a risk processing module 702 configured to perform risk control on the target user according to a risk control decision of the target user and a risk value output by the personalized wind control model of the target user; the personalized wind control model of the target user is sent by the personalized wind control model obtaining module 503 in the device for establishing a wind control model provided in any embodiment of the present specification.
In an embodiment of the apparatus of the present specification, the personalized wind control decision obtaining module 701 is further configured to: providing a questionnaire about security appeal to a target user, and acquiring security appeal data of the target user based on an answer of the questionnaire of the target user; and/or modifying the record and/or the complaint data according to the password of the target user to obtain the security appeal data of the target user.
According to an embodiment of another aspect, there is also provided a computer-readable storage medium having stored thereon a computer program which, when executed in a computer, causes the computer to perform the method described in connection with the embodiments of the present specification.
According to an embodiment of a further aspect, there is also provided a computing device, including a memory and a processor, the memory having stored therein executable code, the processor, when executing the executable code, implementing the method described in connection with the embodiments of the specification.
The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the apparatus embodiment, since it is substantially similar to the method embodiment, the description is relatively simple, and reference may be made to the partial description of the method embodiment for relevant points.
Those skilled in the art will recognize that, in one or more of the examples described above, the functions described in this invention may be implemented in hardware, software, firmware, or any combination thereof. When implemented in software, the functions may be stored on or transmitted over as one or more instructions or code on a computer-readable medium.
The above-mentioned embodiments, objects, technical solutions and advantages of the present invention are further described in detail, it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made on the basis of the technical solutions of the present invention should be included in the scope of the present invention.
Claims (14)
1. The method for establishing the wind control model comprises the following steps:
acquiring a general wind control model; the general wind control model is obtained by training through a first sample set, and the first sample set is generated through behavior data of at least two types of users;
generating a second sample set corresponding to the target user by using the behavior data of the target user;
training the general wind control model by using the second sample set to obtain an individualized wind control model corresponding to the target user;
before obtaining the general wind control model, the method further comprises the following steps:
respectively compressing the original general wind control models obtained by training the first sample set according to preset compression requirements of each grade to obtain compressed general wind control models corresponding to each grade; each level corresponds to one type of terminal equipment configuration;
acquiring configuration information of the terminal equipment used by the target user; and
determining a target grade according to the acquired configuration information;
then, the obtained general wind control model and the general wind control model trained by using the second sample set are: and the compressed general wind control model corresponding to the target level.
2. The method of claim 1, wherein the feature space of the first sample set is the same as the feature space of the second sample set.
3. The method of claim 1, wherein the compressing the original generic wind control models trained by using the first sample set according to preset compression requirements of each level comprises:
for each level, performing:
determining the number N of the neurons needing to be deleted according to the compression requirement of the level and the number M of the neurons used by the original general wind control model;
selecting N neurons from M neurons used by the original general wind control model according to the configuration of the terminal equipment corresponding to the grade or the sequence from the first layer neuron to the last layer neuron;
deleting the N selected neurons to obtain a compressed general wind control model corresponding to the level;
wherein M and N are positive integers greater than 0, and M is greater than or equal to N.
4. The method according to any one of claims 1 to 3, wherein the method is applied to a terminal device;
or,
the method is applied to a server, and the method further comprises the following steps: and sending the personalized wind control model corresponding to the target user to the terminal equipment of the target user.
5. A method of risk control, comprising:
determining a risk control decision of the target user according to the obtained safety appeal data of the target user;
performing risk control on the target user according to the risk control decision of the target user and the risk value output by the personalized wind control model of the target user; wherein the target user's personalized wind control model is obtained by using the method of any one of claims 1 to 4.
6. The method of claim 5, wherein the method for obtaining the security appeal data of the target user comprises:
providing a questionnaire about security appeal to a target user, and acquiring security appeal data of the target user based on an answer of the questionnaire of the target user;
and/or the presence of a gas in the gas,
and modifying the record and/or the complaint data according to the password of the target user to obtain the safety appeal data of the target user.
7. The method of any one of claims 5 to 6, wherein the method is applied to a terminal device or a server.
8. The device for establishing the wind control model comprises the following steps:
the general wind control model acquisition module is configured to acquire a general wind control model; the general wind control model is obtained by training through a first sample set, and the first sample set is generated through behavior data of at least two types of users;
the personalized sample generation module is configured to generate a second sample set corresponding to the target user by utilizing the behavior data of the target user;
the personalized wind control model acquisition module is configured to train the general wind control model by using the second sample set to obtain a personalized wind control model corresponding to the target user;
further comprising:
the compression module is configured to respectively compress the original general wind control models obtained by training the first sample set according to preset compression requirements of each grade to obtain compressed general wind control models corresponding to each grade; each level corresponds to one type of terminal equipment configuration;
the issuing module is configured to acquire configuration information of the terminal equipment used by the target user and determine a target grade according to the acquired configuration information; and issuing the compressed general wind control model corresponding to the target grade to the general wind control model acquisition module.
9. The apparatus of claim 8, wherein the feature space of the first sample set is the same as the feature space of the second sample set.
10. The apparatus of claim 8, the compression module configured to perform, for each level:
determining the number N of the neurons needing to be deleted according to the compression requirement of the level and the number M of the neurons used by the original general wind control model;
selecting N neurons from M neurons used by the original general wind control model according to the configuration of the terminal equipment corresponding to the grade or the sequence from the first layer neuron to the last layer neuron;
deleting the selected N neurons to obtain a compressed general wind control model corresponding to the grade;
wherein M and N are positive integers greater than 0, and M is greater than or equal to N.
11. A risk control device comprising:
the personalized wind control decision acquisition module is configured to determine a risk control decision of the target user according to pre-acquired security appeal data of the target user;
the risk processing module is configured to carry out risk control on the target user according to a risk control decision of the target user and a risk value output by the personalized wind control model of the target user; wherein the target user's personalized wind control model is issued by the apparatus of any of claims 8 to 10.
12. The apparatus of claim 11, wherein the personalized wind control decision acquisition module is further configured to: providing a questionnaire about security appeal to a target user, and acquiring security appeal data of the target user based on an answer of the questionnaire of the target user; and/or modifying the record and/or the complaint data according to the password of the target user to obtain the security appeal data of the target user.
13. A computer-readable storage medium, on which a computer program is stored which, when executed in a computer, causes the computer to carry out the method of any one of claims 1-7.
14. A computing device comprising a memory and a processor, wherein the memory has stored therein executable code that, when executed by the processor, implements the method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110323751.7A CN112950221B (en) | 2021-03-26 | 2021-03-26 | Method and device for establishing wind control model and risk control method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110323751.7A CN112950221B (en) | 2021-03-26 | 2021-03-26 | Method and device for establishing wind control model and risk control method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112950221A CN112950221A (en) | 2021-06-11 |
| CN112950221B true CN112950221B (en) | 2022-07-26 |
Family
ID=76226703
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110323751.7A Active CN112950221B (en) | 2021-03-26 | 2021-03-26 | Method and device for establishing wind control model and risk control method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112950221B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113902473B (en) * | 2021-09-29 | 2024-06-14 | 支付宝(杭州)信息技术有限公司 | Training method and device of business prediction system |
Family Cites Families (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10229356B1 (en) * | 2014-12-23 | 2019-03-12 | Amazon Technologies, Inc. | Error tolerant neural network model compression |
| CN109034660B (en) * | 2018-08-22 | 2023-07-14 | 平安科技（深圳）有限公司 | Method and related device for determining risk control strategy based on prediction model |
| CN109409198B (en) * | 2018-08-31 | 2023-09-05 | 平安科技（深圳）有限公司 | AU detection method, AU detection device, AU detection equipment and AU detection medium |
| CN110163367B (en) * | 2018-09-29 | 2023-04-07 | 腾讯科技（深圳）有限公司 | Terminal deployment method and device |
| CN111144561B (en) * | 2018-11-05 | 2023-05-02 | 杭州海康威视数字技术股份有限公司 | Neural network model determining method and device |
| CN109934300B (en) * | 2019-03-21 | 2023-08-25 | 腾讯科技（深圳）有限公司 | Model compression method, device, computer equipment and storage medium |
| CN111738401A (en) * | 2019-03-25 | 2020-10-02 | 北京三星通信技术研究有限公司 | Model optimization method, grouping compression method, corresponding device and equipment |
| CN110827040A (en) * | 2019-10-31 | 2020-02-21 | 支付宝(杭州)信息技术有限公司 | Consumer appeal solution method and system |
| CN111047423A (en) * | 2019-11-01 | 2020-04-21 | 支付宝(杭州)信息技术有限公司 | Risk determination method and device and electronic equipment |
| CN111488990B (en) * | 2020-04-17 | 2022-11-29 | 苏州浪潮智能科技有限公司 | Model clipping method, device, equipment and medium based on performance perception |
| CN111860770A (en) * | 2020-06-18 | 2020-10-30 | 苏州浪潮智能科技有限公司 | Model compression method and system integrating clipping and quantization |
| CN111985650B (en) * | 2020-07-10 | 2022-06-28 | 华中科技大学 | Activity recognition model and system considering both universality and individuation |
| CN111598238A (en) * | 2020-07-17 | 2020-08-28 | 支付宝(杭州)信息技术有限公司 | Compression method and device of deep learning model |
| CN111967583A (en) * | 2020-08-13 | 2020-11-20 | 北京嘀嘀无限科技发展有限公司 | Method, apparatus, device and medium for compressing neural network |
| CN112231742B (en) * | 2020-12-14 | 2021-06-18 | 支付宝(杭州)信息技术有限公司 | Model joint training method and device based on privacy protection |
-
2021
- 2021-03-26 CN CN202110323751.7A patent/CN112950221B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN112950221A (en) | 2021-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109514586B (en) | Method and system for realizing intelligent customer service robot | |
| CN108764239B (en) | Invoice verification method and device, computer equipment and storage medium | |
| CN110602114B (en) | Block chain-based identity authentication method and device, storage medium and electronic equipment | |
| CN111754234A (en) | Air banking business processing method and device | |
| CN110287250B (en) | User grade quantification method and device | |
| CN112231663A (en) | Data acquisition method, device, equipment and storage medium combining RPA and AI | |
| CN108428175A (en) | A kind of big data analysis method and system based on consumer record | |
| CN112950221B (en) | Method and device for establishing wind control model and risk control method and device | |
| CN111048115A (en) | Voice recognition method and device | |
| CN111192151A (en) | Service processing method, device, storage medium and computer equipment | |
| CN110276608A (en) | Payment processing method, device, readable storage medium storing program for executing and computer equipment | |
| CN109697607A (en) | A kind of Electronic Transfer method, apparatus and storage medium | |
| CN111078942B (en) | Method, device and storage medium for recommending videos | |
| CN115393077B (en) | Data processing method based on loan transaction man-machine conversation system and related device | |
| CN116757855A (en) | Intelligent insurance service method, device, equipment and storage medium | |
| CN108268506A (en) | A kind of processing method of promotion message, device, terminal and computer readable storage medium | |
| CN114971017A (en) | Bank transaction data processing method and device | |
| CN114723455A (en) | Service processing method and device, electronic equipment and storage medium | |
| CN109299454A (en) | Abstraction generating method and device, storage medium and electric terminal based on chat log | |
| CN114529210A (en) | Evaluation method and device based on virtual reality | |
| CN114723206A (en) | Asset data processing method, computer equipment and readable storage medium | |
| CN110197074B (en) | User authority control method and device | |
| CN111968632A (en) | Call voice acquisition method and device, computer equipment and storage medium | |
| KR102139128B1 (en) | Platform system | |
| CN113782022B (en) | Communication method, device, equipment and storage medium based on intention recognition model |
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 |