CN116776114A - Risk detection method and device, storage medium and electronic equipment - Google Patents
Risk detection method and device, storage medium and electronic equipment Download PDFInfo
- Publication number
- CN116776114A CN116776114A CN202310401375.8A CN202310401375A CN116776114A CN 116776114 A CN116776114 A CN 116776114A CN 202310401375 A CN202310401375 A CN 202310401375A CN 116776114 A CN116776114 A CN 116776114A
- Authority
- CN
- China
- Prior art keywords
- data
- control model
- wind control
- target
- features
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 58
- 238000003860 storage Methods 0.000 title claims abstract description 21
- 238000000034 method Methods 0.000 claims abstract description 65
- 238000012549 training Methods 0.000 claims abstract description 64
- 238000000605 extraction Methods 0.000 claims abstract description 63
- 238000012545 processing Methods 0.000 claims abstract description 32
- 238000005457 optimization Methods 0.000 claims abstract description 21
- 238000004590 computer program Methods 0.000 claims description 15
- 230000000873 masking effect Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 description 25
- 238000010586 diagram Methods 0.000 description 16
- 230000006870 function Effects 0.000 description 13
- 230000006872 improvement Effects 0.000 description 8
- 238000005516 engineering process Methods 0.000 description 7
- 238000012954 risk control Methods 0.000 description 4
- 238000009826 distribution Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000001010 compromised effect Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000005012 migration Effects 0.000 description 2
- 238000013508 migration Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 230000010485 coping Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000005055 memory storage Effects 0.000 description 1
- 229920001296 polysiloxane Polymers 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 239000010979 ruby Substances 0.000 description 1
- 229910001750 ruby Inorganic materials 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F21/00—Security arrangements for protecting computers, components thereof, programs or data against unauthorised activity
- G06F21/60—Protecting data
- G06F21/62—Protecting access to data via a platform, e.g. using keys or access control rules
- G06F21/6218—Protecting access to data via a platform, e.g. using keys or access control rules to a system of files or objects, e.g. local or distributed file system or database
- G06F21/6245—Protecting personal data, e.g. for financial or medical purposes
-
- 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
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Business, Economics & Management (AREA)
- General Health & Medical Sciences (AREA)
- Bioethics (AREA)
- Accounting & Taxation (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Technology Law (AREA)
- General Business, Economics & Management (AREA)
- Finance (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Medical Informatics (AREA)
- Databases & Information Systems (AREA)
- Computer Hardware Design (AREA)
- Computer Security & Cryptography (AREA)
- Software Systems (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The specification discloses a risk detection method, a risk detection device, a storage medium and electronic equipment. The risk detection method is used for privacy protection and comprises the following steps: the method comprises the steps of obtaining first service data, determining original features corresponding to the first service data, carrying out mask processing on data of partial dimensions in the first service data to obtain masked data, processing the masked data through a feature extraction network of a preset wind control model to determine target features corresponding to the masked data, inputting the target features corresponding to the masked data into a feature reconstruction network of the wind control model to obtain reconstructed features, taking deviation between the minimized original features and the reconstructed features as an optimization target, training at least the feature extraction network in the wind control model, deploying the wind control model after the training is completed, and inputting the target service data corresponding to the service request into the deployed wind control model after the service request is received, so that risk detection is carried out through the wind control model.
Description
Technical Field
The present disclosure relates to the field of internet technologies, and in particular, to a risk detection method, apparatus, storage medium, and electronic device.
Background
With the rapid development of internet technology, privacy and property security of users in the process of performing businesses such as financial transactions, financial transactions and the like are also facing increasing challenges, which requires timely risk detection and risk control during the process of performing businesses by users.
Along with the continuous updating of the risk control technology, the expression form of the risk also changes greatly along with the time, new risk forms are layered endlessly, training samples which can be used for coping with the new risk forms are fewer, the training effect of the model is poor, and the data characteristics of the risk data are difficult to be extracted effectively for carrying out accurate risk detection, so that the privacy and property safety of users face great examination.
Therefore, how to train the model through the label-free sample data so as to extract effective characteristic data to carry out accurate risk detection and ensure the privacy and property safety of users is a problem to be solved urgently.
Disclosure of Invention
The specification provides a risk detection method, a risk detection device, a storage medium and electronic equipment. And training a feature extraction network in the wind control model by taking the minimized deviation between the original features and the reconstructed features as an optimization target according to the first business data without labels.
The technical scheme adopted in the specification is as follows:
the present specification provides a method of risk detection, comprising:
acquiring first service data;
determining original characteristics corresponding to the first service data, and performing mask processing on data of partial dimensions in the first service data to obtain masked data;
processing the masked data through a characteristic extraction network of a preset wind control model to determine target characteristics corresponding to the masked data;
inputting target features corresponding to the masked data into a feature reconstruction network of the wind control model to obtain reconstructed features;
training at least a feature extraction network in the wind control model with a minimum deviation between the original features and the reconstructed features as an optimization objective;
and deploying the wind control model after training, and after receiving the service request, inputting target service data corresponding to the service request into the deployed wind control model so as to execute risk detection through the wind control model.
Optionally, the masked data is processed through a feature extraction network of a preset wind control model to determine target features corresponding to the masked data, which specifically includes:
Determining mask features corresponding to the masked data according to the masked data, wherein each feature value in the mask features is used for representing whether the data of each dimension contained in the first business data is masked or not;
adjusting the original features through the mask features to obtain adjusted features;
and inputting the adjusted features into the feature extraction network to determine target features corresponding to the masked data according to the adjusted features.
Optionally, the original feature is adjusted through the mask feature, so as to obtain an adjusted feature, which specifically includes:
determining a target position corresponding to the masked data in the first service data according to the mask characteristics;
determining a corresponding characteristic value of the data of the target position in the original characteristic as a target characteristic value;
and adjusting the target characteristic value to obtain the adjusted characteristic.
Optionally, the target feature value is adjusted to obtain the adjusted feature, which specifically includes:
and randomly adjusting the target characteristic value to the characteristic value corresponding to other data which is not in the same data line with the target position, and obtaining the adjusted characteristic.
Optionally, at least training a feature extraction network in the wind control model with the aim of minimizing the deviation between the original feature and the reconstructed feature, specifically including:
inputting the target features into a mask prediction network of the wind control model to obtain a mask prediction result, wherein the mask prediction result is used for representing the probability of masking data of each dimension in the first business data;
determining a first loss value of the wind control model according to the deviation between the original characteristic and the reconstruction characteristic, and determining a second loss value of the wind control model according to the deviation between the mask prediction result and the mask characteristic;
determining a comprehensive loss value of the wind control model according to the first loss value and the second loss value;
and training at least the feature extraction network in the wind control model by taking the minimum comprehensive loss value as an optimization target.
Optionally, before deploying the trained wind control model, the method further comprises:
acquiring second service data;
inputting the second service data into the feature extraction network as sample data to determine target features corresponding to the second service data through the feature extraction network;
Inputting target features corresponding to the second service data into a classification network of the wind control model to determine risk detection results corresponding to the second service data through the classification network;
and training at least the feature extraction network in the wind control model by taking the deviation between the minimum risk detection result and the actual label corresponding to the second service data as an optimization target.
The present specification provides an apparatus for risk detection, comprising:
the acquisition module acquires first service data;
the mask module is used for determining original characteristics corresponding to the first service data, and masking data of partial dimensions in the first service data to obtain masked data;
the processing module is used for processing the masked data through a characteristic extraction network of a preset wind control model so as to determine target characteristics corresponding to the masked data;
the reconstruction module inputs the target features corresponding to the masked data into a feature reconstruction network of the wind control model to obtain reconstruction features;
the training module is used for training at least a feature extraction network in the wind control model by taking the deviation between the minimized original features and the reconstructed features as an optimization target;
The detection module deploys the wind control model after the training is completed, and inputs target service data corresponding to the service request into the deployed wind control model after the service request is received, so as to execute risk detection through the wind control model.
Optionally, the processing module is specifically configured to determine, according to the masked data, a mask feature corresponding to the masked data, where each feature value in the mask feature is used to characterize whether data of each dimension included in the first service data is masked; adjusting the original features through the mask features to obtain adjusted features; and inputting the adjusted features into the feature extraction network to determine target features corresponding to the masked data according to the adjusted features.
Optionally, the processing module is specifically configured to determine, according to the mask feature, a target position corresponding to the masked data in the first service data; determining a corresponding characteristic value of the data of the target position in the original characteristic as a target characteristic value; and adjusting the target characteristic value to obtain the adjusted characteristic.
Optionally, the processing module is specifically configured to randomly adjust the target feature value to a feature value corresponding to other data that is not in the same data line with the target position, so as to obtain the adjusted feature.
Optionally, the training module is specifically configured to input the target feature into a mask prediction network of the wind control model to obtain a mask prediction result, where the mask prediction result is used to characterize a probability that data in each dimension of the first service data is masked; determining a first loss value of the wind control model according to the deviation between the original characteristic and the reconstruction characteristic, and determining a second loss value of the wind control model according to the deviation between the mask prediction result and the mask characteristic; determining a comprehensive loss value of the wind control model according to the first loss value and the second loss value; and training at least the feature extraction network in the wind control model by taking the minimum comprehensive loss value as an optimization target.
Optionally, before the wind control model after training is deployed, the training module is further configured to obtain second service data; inputting the second service data into the feature extraction network as sample data to determine target features corresponding to the second service data through the feature extraction network; inputting target features corresponding to the second service data into a classification network of the wind control model to determine risk detection results corresponding to the second service data through the classification network; and training at least the feature extraction network in the wind control model by taking the deviation between the minimum risk detection result and the actual label corresponding to the second service data as an optimization target.
The present specification provides a computer readable storage medium storing a computer program which when executed by a processor implements the method of risk detection described above.
The present specification provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing a method of risk detection as described above when executing the program.
The above-mentioned at least one technical scheme that this specification adopted can reach following beneficial effect:
in the risk detection method provided by the specification, first service data is acquired; the method comprises the steps of determining original features corresponding to first service data, carrying out mask processing on data of partial dimensions in the first service data to obtain masked data, processing the masked data through a feature extraction network of a preset wind control model to determine target features corresponding to the masked data, inputting the target features corresponding to the masked data into a feature reconstruction network of the wind control model to obtain reconstructed features, taking deviation between the minimized original features and the reconstructed features as an optimization target, training at least the feature extraction network in the wind control model, deploying the trained wind control model, and after a service request is received, inputting the target service data corresponding to the service request into the deployed wind control model to execute risk detection through the wind control model.
According to the method, in the process of training the wind control model, sample data with labels are not needed, but the feature extraction network of the wind control model is trained according to the deviation between the original features and the reconstructed features of the first service data, so that the training of the feature extraction network of the wind control model can be completed even under the condition that the sample data are fewer, and further risk detection and risk control are accurately carried out on the service data, and privacy and property safety of users are guaranteed.
Drawings
The accompanying drawings, which are included to provide a further understanding of the specification, illustrate and explain the exemplary embodiments of the present specification and their description, are not intended to limit the specification unduly. Attached at
In the figure:
FIG. 1 is a flow chart of a method of risk detection provided in the present specification;
FIG. 2 is a schematic diagram of a training process of a wind control model provided in the present specification;
FIG. 3 is a schematic diagram of a fine tuning process of a wind control model provided in the present specification;
FIG. 4 is a schematic diagram of an apparatus for risk detection provided herein;
Fig. 5 is a schematic diagram of an electronic device corresponding to fig. 1 provided in the present specification.
Detailed Description
For the purposes of making the objects, technical solutions and advantages of the present specification more apparent, the technical solutions of the present specification will be clearly and completely described below with reference to specific embodiments of the present specification and corresponding drawings. It will be apparent that the described embodiments are only some, but not all, of the embodiments of the present specification. All other embodiments, which can be made by one of ordinary skill in the art without undue burden from the present disclosure, are intended to be within the scope of the present disclosure.
The following describes in detail the technical solutions provided by the embodiments of the present specification with reference to the accompanying drawings.
Fig. 1 is a flow chart of a method for risk detection provided in the present specification, which includes the following steps:
s100: and acquiring first service data.
In the process of executing services such as financial transactions and financial transactions, users are usually faced with various risks, which requires risk detection on service data of the users in the process of executing the services, so as to timely control risks involved in executing the services, such as detecting whether the users are abnormal users, whether the users execute the abnormal services, whether the privacy of the users is compromised, whether the property security of the users is compromised, and the like.
In the process of performing risk detection, feature extraction is generally required to be performed through a feature extraction network in a wind control model, so that risk data can be accurately described through the extracted features, and further possible risks can be accurately identified.
Because the expression form of the risk is continuously updated, the number of the labeling sample data for training the wind control model is small, and based on the fact, the specification provides a risk detection method for training the feature extraction network of the wind control model through the label-free service data, and further carrying out risk detection through the trained wind control model.
In the present specification, an execution body for implementing a method of risk detection provided in the present specification may be a designated device such as a server, and for convenience of description, the method of risk detection provided in the present specification will be described by taking a server as an example of the execution body.
The server needs to acquire the first service data and take the service data as a training sample without a label, i.e. whether the training sample is at risk is not marked. In practical applications, the data type of the first service data may be structured data (tabular data), which includes user data such as income, age, occupation, borrowing condition, sex, etc. of the user, and transaction data such as account location, transaction amount, service type, transaction channel, etc. related to the current service of the user. Of course, other data may be included, which is not specifically limited in this specification.
S102: and determining the original characteristics corresponding to the first service data, and carrying out mask processing on the data with partial dimensions in the first service data to obtain masked data.
S104: and processing the masked data through a characteristic extraction network of a preset wind control model to determine target characteristics corresponding to the masked data.
Before the first service data is input into the wind control model, the server may pre-process the first service data, where the server may determine an original feature corresponding to the service data.
In addition, the server may perform mask processing on a part of dimensions of the first service data to obtain masked data, and determine mask features corresponding to the masked data. For example, the server may randomly mask 15% of the first service data, replace the original data with a mask, and determine the mask feature according to the masked data.
In this specification, the mask feature is the same as the dimension of the original feature, and each feature value in the feature matrix corresponding to the mask feature indicates whether the data corresponding to the feature value is masked, in other words, each feature value in the mask feature is used to characterize whether the data of each dimension included in the first service data is masked.
For example, for each feature value in the mask feature correspondence feature matrix, the data corresponding to the feature value is not masked when the value of the feature value is 0, and the data corresponding to the feature value is masked when the value of the feature value is 1, so that each feature value in the mask feature satisfies the bernoulli distribution (0-1 distribution).
And then the server can adjust the original features through the mask features to obtain adjusted features, and further the server can extract the adjusted feature target features from the network to determine target features corresponding to the masked data according to the adjusted features, and the target features are used as target features corresponding to the first service data.
The server may determine, according to the mask feature, a target position corresponding to the masked data in the first service data, then determine a feature value corresponding to the data of the target position in the original feature, as a target feature value, and further randomly adjust the target feature value to a feature value corresponding to other data that is not in the same data line as the target position, so as to obtain an adjusted feature.
Specifically, the first service data in this specification may be represented as n×d dimension data X, and the server may set, in advance, a super parameter in the wind control model, for example, a proportion of data to be masked, and mask the X to obtain n×d masked data M.
Then, an original feature X corresponding to the data X and a mask feature M corresponding to the masked data M may be determined, where dimensions corresponding to the original feature X and the mask feature M are the same.
For each row in data X, the server may obtain replacement data of the same dimension by sampling with a put backNew data representing the random substitution of the masked data in the line data with the data of the other data line, and +.>Corresponding replacement feature->The above-mentioned adjusted features can be expressed as:
wherein, the liquid crystal display device comprises a liquid crystal display device,is a target feature. />Indicating the alternative features corresponding to the masked data, x # (1-m) indicating the original features corresponding to the unmasked data.
And then the server can input the adjusted features into a feature extraction network of the wind control model, and further extract the features through an encoder and a decoder of the feature extraction network, so as to obtain target features capable of representing semantic information corresponding to the first service data.
S106: and inputting the target features corresponding to the masked data into a feature reconstruction network of the wind control model to obtain reconstructed features.
In particular, the server may characterize the targetThe characteristic reconstruction network input into the wind control model can be added according to the target characteristic >Reconstructing the original feature, thereby outputting the reconstructed feature +.>
Meanwhile, the server can also input the target characteristics into a mask prediction network in the wind control model, and the mask prediction network can be used for predicting the target characteristics according to the target characteristicsOutput mask prediction result +.>The mask prediction structure is used to characterize the probability that data for each dimension in the first traffic data is masked.
S108: and training at least a feature extraction network in the wind control model with the aim of optimizing the deviation between the original features and the reconstructed features to be minimized.
The server may determine a first loss value of the wind control model based on a deviation between the original feature and the target feature, and determine a second loss value of the wind control model based on a deviation between the mask prediction result and the mask feature.
It should be noted that, since the mask feature indicates whether the data of each dimension in the first service data is masked, that is, obeys the 0-1 distribution, and the mask prediction result indicates the probability that the data of each dimension in the first service data is masked, the mask feature may be used as a tag, and the second loss value may be determined according to the deviation between the mask prediction result and the tag.
In this description, the loss function that determines the first loss value may be a mean square error loss function (Mean Squared Error Loss, MSELoss), and the loss function that determines the second loss value may be a Binary cross entropy loss function (Binary CrossEntropy, BCELoss).
And then the server can determine the comprehensive loss value of the wind control model according to the first loss value and the second loss value, further train the wind control model by taking the minimized comprehensive loss value as an optimization target until reaching a training target, wherein the training target can be a preset range or preset training times for the wind control model, the preset range and the preset training times can be set according to actual conditions, and the specification is not limited in particular.
For easy understanding, the present disclosure also provides a schematic diagram of a training process of the wind control model, as shown in fig. 2.
Fig. 2 is a schematic diagram of a training process of a wind control model provided in the present specification.
The server can determine the adjusted characteristics according to the original characteristics and the mask characteristics of the first service data, input the adjusted characteristics into a characteristic extraction network of the wind control model to obtain target characteristics, determine the reconstructed characteristics through the reconstruction network, determine a first loss value according to the reconstructed characteristics and the original characteristics, determine a mask prediction result through the mask prediction network, determine a second loss value according to the mask prediction result and the mask characteristics, further determine a comprehensive loss value according to the first loss value and the second loss value, and train the wind control model according to the comprehensive loss value.
Of course, the wind control model in the present specification may not include a mask prediction network, and the server may determine the reconstructed feature only according to the feature reconstruction network, so as to train each network layer including the feature extraction network in the wind control model with the deviation between the original feature and the reconstructed feature minimized as an optimization target.
Further, after the wind control model is trained through the first service data each time, the server can perform model migration on the main service. After the model is trained on the label-free data and converged, the parameters of the common part (the characteristic extraction network) of the wind control model are initialized to a main task, and the main task is continuously subjected to fine tuning optimization on a small amount of labeled data, however, a migration mode of continuous excitation in an actual process can often play a better effect.
Therefore, in this specification, after each training of the wind control model by the first service data, the server may perform at least one more training of the wind control model by the second service data with a label, so that the first service data (without a label) and the second service data (with a label) are used in accordance with 1: and N, training the wind control model back and forth in turn to finish the pre-training and fine-tuning of the wind control model.
In the process of training the wind control model through the second service data, the server can firstly acquire the second service data, input the second service data into a feature extraction network of the wind control model, determine target features corresponding to the second service data through the feature extraction network, then input the target features corresponding to the second service data into a classification network of the wind control model, determine a risk detection result corresponding to the second service data through the classification network, and further train at least the feature extraction network in the wind control model with the deviation between the minimum risk detection result and an actual label corresponding to the second service data as an optimization target.
For ease of understanding, the present disclosure also provides a schematic diagram of a fine tuning process of a model, as shown in fig. 3.
Fig. 3 is a schematic diagram of a fine tuning process of a wind control model provided in the present specification.
After the wind control model is trained through the label-free data each time, the server carries out N times of training on the wind control model through the label-free data, so that fine adjustment is carried out on the wind control model, and the fact that encoder parameters of the feature extraction network in the process of carrying out fine adjustment on the wind control model and the wind control model are shared in the process of training the wind control model is needed to be described.
S110: and deploying the wind control model after training, and after receiving the service request, inputting target service data corresponding to the service request into the deployed wind control model so as to execute risk detection through the wind control model.
After the wind control model is trained and fine-tuned, the server can deploy the trained wind control model, in the practical application process, the server can further acquire target service data corresponding to the service request after acquiring the service request, and input the target service data into a feature extraction network of the wind control model, so that target features corresponding to the target service data are obtained.
The server may then determine risk detection results (e.g., whether there is risk, risk type, etc.) based on the target features via the classification network of the wind-controlled model.
According to the method, in the process of training the wind control model, the labeled sample data are not needed, and the feature extraction network of the wind control model is trained according to the deviation between the original features and the reconstruction features of the first service data and the deviation between the mask data and the mask prediction result, so that the training of the feature extraction network of the wind control model can be completed even if the sample data are fewer, and therefore risk detection and risk control are accurately carried out on the service data, and privacy and property safety of a user are guaranteed.
In addition, after the wind control model is trained through the label-free training sample (first service data) each time, the wind control model can be trained for N times through the label-free training sample (second service data), so that the training effect of the wind control model is further improved.
The foregoing describes one or more methods for performing risk detection according to the present disclosure, and provides a corresponding risk detection apparatus based on the same concept, as shown in fig. 4.
Fig. 4 is a schematic diagram of an apparatus for risk detection provided in the present specification, including:
an obtaining module 400, configured to obtain first service data;
a masking module 402, configured to determine an original feature corresponding to the first service data, and perform masking processing on data of a part of dimensions in the first service data to obtain masked data;
a processing module 404, configured to process the masked data through a feature extraction network of a preset wind control model, so as to determine a target feature corresponding to the masked data;
a reconstruction module 406, configured to input the target feature corresponding to the masked data into a feature reconstruction network of the wind control model, to obtain a reconstructed feature;
A training module 408, configured to train at least a feature extraction network in the wind control model with a goal of optimizing to minimize a deviation between the original feature and the reconstructed feature;
the detection module 410 is configured to deploy the wind control model after training is completed, and after receiving a service request, input target service data corresponding to the service request into the deployed wind control model, so as to execute risk detection through the wind control model.
Optionally, the processing module 404 is specifically configured to determine, according to the masked data, a mask feature corresponding to the masked data, where each feature value in the mask feature is used to represent whether data of each dimension included in the first service data is masked; adjusting the original features through the mask features to obtain adjusted features; and inputting the adjusted features into the feature extraction network to determine target features corresponding to the masked data according to the adjusted features.
Optionally, the processing module 404 is specifically configured to determine, according to the mask feature, a target location corresponding to the masked data in the first service data; determining a corresponding characteristic value of the data of the target position in the original characteristic as a target characteristic value; and adjusting the target characteristic value to obtain the adjusted characteristic.
Optionally, the processing module 404 is specifically configured to randomly adjust the target feature value to a feature value corresponding to other data that is not in the same data line with the target position, so as to obtain the adjusted feature.
Optionally, the training module 408 is specifically configured to input the target feature into a mask prediction network of the wind control model to obtain a mask prediction result, where the mask prediction result is used to characterize a probability that data of each dimension in the first service data is masked; determining a first loss value of the wind control model according to the deviation between the original characteristic and the reconstruction characteristic, and determining a second loss value of the wind control model according to the deviation between the mask prediction result and the mask characteristic; determining a comprehensive loss value of the wind control model according to the first loss value and the second loss value; and training at least the feature extraction network in the wind control model by taking the minimum comprehensive loss value as an optimization target.
Optionally, before deploying the trained wind control model, the training module 408 is further configured to obtain second service data; inputting the second service data into the feature extraction network as sample data to determine target features corresponding to the second service data through the feature extraction network; inputting target features corresponding to the second service data into a classification network of the wind control model to determine risk detection results corresponding to the second service data through the classification network; and training at least the feature extraction network in the wind control model by taking the deviation between the minimum risk detection result and the actual label corresponding to the second service data as an optimization target.
The present specification also provides a computer readable storage medium storing a computer program operable to perform a method of risk detection as provided in figure 1 above.
The present specification also provides a schematic structural diagram of an electronic device corresponding to fig. 1 shown in fig. 5. At the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile storage, as illustrated in fig. 5, although other hardware required by other services may be included. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs to implement the risk detection method described above with respect to fig. 1. Of course, other implementations, such as logic devices or combinations of hardware and software, are not excluded from the present description, that is, the execution subject of the following processing flows is not limited to each logic unit, but may be hardware or logic devices.
In the 90 s of the 20 th century, improvements to one technology could clearly be distinguished as improvements in hardware (e.g., improvements to circuit structures such as diodes, transistors, switches, etc.) or software (improvements to the process flow). However, with the development of technology, many improvements of the current method flows can be regarded as direct improvements of hardware circuit structures. Designers almost always obtain corresponding hardware circuit structures by programming improved method flows into hardware circuits. Therefore, an improvement of a method flow cannot be said to be realized by a hardware entity module. For example, a programmable logic device (Programmable Logic Device, PLD) (e.g., field programmable gate array (Field Programmable Gate Array, FPGA)) is an integrated circuit whose logic function is determined by the programming of the device by a user. A designer programs to "integrate" a digital system onto a PLD without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Moreover, nowadays, instead of manually manufacturing integrated circuit chips, such programming is mostly implemented by using "logic compiler" software, which is similar to the software compiler used in program development and writing, and the original code before the compiling is also written in a specific programming language, which is called hardware description language (Hardware Description Language, HDL), but not just one of the hdds, but a plurality of kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), lava, lola, myHDL, PALASM, RHDL (Ruby Hardware Description Language), etc., VHDL (Very-High-Speed Integrated Circuit Hardware Description Language) and Verilog are currently most commonly used. It will also be apparent to those skilled in the art that a hardware circuit implementing the logic method flow can be readily obtained by merely slightly programming the method flow into an integrated circuit using several of the hardware description languages described above.
The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer readable medium storing computer readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, application specific integrated circuits (Application Specific Integrated Circuit, ASIC), programmable logic controllers, and embedded microcontrollers, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, atmel AT91SAM, microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic of the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller in a pure computer readable program code, it is well possible to implement the same functionality by logically programming the method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers, etc. Such a controller may thus be regarded as a kind of hardware component, and means for performing various functions included therein may also be regarded as structures within the hardware component. Or even means for achieving the various functions may be regarded as either software modules implementing the methods or structures within hardware components.
The system, apparatus, module or unit set forth in the above embodiments may be implemented in particular by a computer chip or entity, or by a product having a certain function. One typical implementation is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
For convenience of description, the above devices are described as being functionally divided into various units, respectively. Of course, the functions of each element may be implemented in one or more software and/or hardware elements when implemented in the present specification.
It will be appreciated by those skilled in the art that embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the present specification may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
The present description is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the specification. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
In one typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.
The memory may include volatile memory in a computer-readable medium, random Access Memory (RAM) and/or nonvolatile memory, such as Read Only Memory (ROM) or flash memory (flash RAM). Memory is an example of computer-readable media.
Computer readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of storage media for a computer include, but are not limited to, phase change memory (PRAM), static Random Access Memory (SRAM), dynamic Random Access Memory (DRAM), other types of Random Access Memory (RAM), read Only Memory (ROM), electrically Erasable Programmable Read Only Memory (EEPROM), flash memory or other memory technology, compact disc read only memory (CD-ROM), digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transmission medium, which can be used to store information that can be accessed by a computing device. Computer-readable media, as defined herein, does not include transitory computer-readable media (transmission media), such as modulated data signals and carrier waves.
It should also be noted that the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising one … …" does not exclude the presence of other like elements in a process, method, article or apparatus that comprises the element.
It will be appreciated by those skilled in the art that embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, the present specification may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
The description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. The specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
In this specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are all referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, as relevant to see a section of the description of method embodiments.
The foregoing is merely exemplary of the present disclosure and is not intended to limit the disclosure. Various modifications and alterations to this specification will become apparent to those skilled in the art. Any modifications, equivalent substitutions, improvements, or the like, which are within the spirit and principles of the present description, are intended to be included within the scope of the claims of the present description.
Claims (14)
1. A method of risk detection, comprising:
acquiring first service data;
determining original characteristics corresponding to the first service data, and performing mask processing on data of partial dimensions in the first service data to obtain masked data;
processing the masked data through a characteristic extraction network of a preset wind control model to determine target characteristics corresponding to the masked data;
Inputting target features corresponding to the masked data into a feature reconstruction network of the wind control model to obtain reconstructed features;
training at least a feature extraction network in the wind control model with a minimum deviation between the original features and the reconstructed features as an optimization objective;
and deploying the wind control model after training, and after receiving the service request, inputting target service data corresponding to the service request into the deployed wind control model so as to execute risk detection through the wind control model.
2. The method of claim 1, wherein the masked data is processed through a feature extraction network of a preset wind control model to determine target features corresponding to the masked data, and specifically comprises:
determining mask features corresponding to the masked data according to the masked data, wherein each feature value in the mask features is used for representing whether the data of each dimension contained in the first business data is masked or not;
adjusting the original features through the mask features to obtain adjusted features;
and inputting the adjusted features into the feature extraction network to determine target features corresponding to the masked data according to the adjusted features.
3. The method according to claim 2, wherein the original feature is adjusted by the mask feature to obtain an adjusted feature, and the method specifically comprises:
determining a target position corresponding to the masked data in the first service data according to the mask characteristics;
determining a corresponding characteristic value of the data of the target position in the original characteristic as a target characteristic value;
and adjusting the target characteristic value to obtain the adjusted characteristic.
4. The method of claim 3, wherein the adjusting the target feature value to obtain the adjusted feature specifically includes:
and randomly adjusting the target characteristic value to the characteristic value corresponding to other data which is not in the same data line with the target position, and obtaining the adjusted characteristic.
5. The method of claim 1, training at least a feature extraction network in the wind control model with a view to minimizing the deviation between the original features and the reconstructed features, comprising in particular:
inputting the target features into a mask prediction network of the wind control model to obtain a mask prediction result, wherein the mask prediction result is used for representing the probability of masking data of each dimension in the first business data;
Determining a first loss value of the wind control model according to the deviation between the original characteristic and the reconstruction characteristic, and determining a second loss value of the wind control model according to the deviation between the mask prediction result and the mask characteristic;
determining a comprehensive loss value of the wind control model according to the first loss value and the second loss value;
and training at least the feature extraction network in the wind control model by taking the minimum comprehensive loss value as an optimization target.
6. The method of claim 1, prior to deploying the trained wind control model, the method further comprising:
acquiring second service data;
inputting the second service data into the feature extraction network as sample data to determine target features corresponding to the second service data through the feature extraction network;
inputting target features corresponding to the second service data into a classification network of the wind control model to determine risk detection results corresponding to the second service data through the classification network;
and training at least the feature extraction network in the wind control model by taking the deviation between the minimum risk detection result and the actual label corresponding to the second service data as an optimization target.
7. An apparatus for risk detection, comprising:
the acquisition module acquires first service data;
the mask module is used for determining original characteristics corresponding to the first service data, and masking data of partial dimensions in the first service data to obtain masked data;
the processing module is used for processing the masked data through a characteristic extraction network of a preset wind control model so as to determine target characteristics corresponding to the masked data;
the reconstruction module inputs the target features corresponding to the masked data into a feature reconstruction network of the wind control model to obtain reconstruction features;
the training module is used for training at least a feature extraction network in the wind control model by taking the deviation between the minimized original features and the reconstructed features as an optimization target;
the detection module deploys the wind control model after the training is completed, and inputs target service data corresponding to the service request into the deployed wind control model after the service request is received, so as to execute risk detection through the wind control model.
8. The apparatus of claim 7, the processing module is specifically configured to determine, according to the masked data, a mask feature corresponding to the masked data, where each feature value in the mask feature is used to characterize whether data of each dimension included in the first service data is masked; adjusting the original features through the mask features to obtain adjusted features; and inputting the adjusted features into the feature extraction network to determine target features corresponding to the masked data according to the adjusted features.
9. The apparatus of claim 8, wherein the processing module is specifically configured to determine, according to the mask feature, a target location corresponding to the masked data in the first service data; determining a corresponding characteristic value of the data of the target position in the original characteristic as a target characteristic value; and adjusting the target characteristic value to obtain the adjusted characteristic.
10. The apparatus of claim 9, wherein the processing module is specifically configured to randomly adjust the target feature value to a feature value corresponding to other data that is not in the same data line as the target location, and obtain the adjusted feature.
11. The apparatus of claim 7, the training module being specifically configured to input the target feature into a mask prediction network of the wind control model to obtain a mask prediction result, where the mask prediction result is used to characterize a probability that data of each dimension in the first service data is masked; determining a first loss value of the wind control model according to the deviation between the original characteristic and the reconstruction characteristic, and determining a second loss value of the wind control model according to the deviation between the mask prediction result and the mask characteristic; determining a comprehensive loss value of the wind control model according to the first loss value and the second loss value; and training at least the feature extraction network in the wind control model by taking the minimum comprehensive loss value as an optimization target.
12. The apparatus of claim 7, wherein the training module is further configured to obtain second business data before deploying the trained wind control model; inputting the second service data into the feature extraction network as sample data to determine target features corresponding to the second service data through the feature extraction network; inputting target features corresponding to the second service data into a classification network of the wind control model to determine risk detection results corresponding to the second service data through the classification network; and training at least the feature extraction network in the wind control model by taking the deviation between the minimum risk detection result and the actual label corresponding to the second service data as an optimization target.
13. A computer readable storage medium storing a computer program which, when executed by a processor, implements the method of any of the preceding claims 1-6.
14. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the method of any of the preceding claims 1-6 when the program is executed.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310401375.8A CN116776114A (en) | 2023-04-13 | 2023-04-13 | Risk detection method and device, storage medium and electronic equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310401375.8A CN116776114A (en) | 2023-04-13 | 2023-04-13 | Risk detection method and device, storage medium and electronic equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116776114A true CN116776114A (en) | 2023-09-19 |
Family
ID=87995245
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310401375.8A Pending CN116776114A (en) | 2023-04-13 | 2023-04-13 | Risk detection method and device, storage medium and electronic equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116776114A (en) |
-
2023
- 2023-04-13 CN CN202310401375.8A patent/CN116776114A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110443618B (en) | Method and device for generating wind control strategy | |
| CN115618748B (en) | Model optimization method, device, equipment and storage medium | |
| CN115828162B (en) | Classification model training method and device, storage medium and electronic equipment | |
| CN115238826B (en) | Model training method and device, storage medium and electronic equipment | |
| CN117197781B (en) | Traffic sign recognition method and device, storage medium and electronic equipment | |
| CN116403034A (en) | Graph data model training method and device, storage medium and electronic equipment | |
| CN116824331A (en) | Model training and image recognition method, device, equipment and storage medium | |
| CN116578877A (en) | Method and device for model training and risk identification of secondary optimization marking | |
| CN116776114A (en) | Risk detection method and device, storage medium and electronic equipment | |
| CN111242195B (en) | Model, insurance wind control model training method and device and electronic equipment | |
| CN116109008B (en) | Method and device for executing service, storage medium and electronic equipment | |
| CN115862675B (en) | Emotion recognition method, device, equipment and storage medium | |
| CN116028820B (en) | Model training method and device, storage medium and electronic equipment | |
| CN117786417B (en) | Model training method, transient source identification method and device and electronic equipment | |
| CN117746193B (en) | Label optimization method and device, storage medium and electronic equipment | |
| CN115545938B (en) | Method, device, storage medium and equipment for executing risk identification service | |
| CN115017915B (en) | Model training and task execution method and device | |
| CN117875522A (en) | Method, device, storage medium and equipment for predicting event number | |
| CN117743824A (en) | Model training and service execution method and device, storage medium and equipment | |
| CN116150685A (en) | Data classification method, device, equipment and storage medium | |
| CN116861976A (en) | Training method, device, equipment and storage medium of anomaly detection model | |
| CN116843466A (en) | Risk identification model training method and device, storage medium and electronic equipment | |
| CN117079274A (en) | Training method and device for recognition model, storage medium and electronic equipment | |
| CN118154947A (en) | Risk image interception method and device, storage medium and electronic equipment | |
| CN117876114A (en) | Method and device for service execution and model training |
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 |