CN116340852A - Model training and business wind control method and device - Google Patents
Model training and business wind control method and device Download PDFInfo
- Publication number
- CN116340852A CN116340852A CN202310632423.4A CN202310632423A CN116340852A CN 116340852 A CN116340852 A CN 116340852A CN 202310632423 A CN202310632423 A CN 202310632423A CN 116340852 A CN116340852 A CN 116340852A
- Authority
- CN
- China
- Prior art keywords
- risk prediction
- prediction result
- target
- risk
- wind control
- 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.)
- Granted
Links
- 238000012549 training Methods 0.000 title claims abstract description 185
- 238000000034 method Methods 0.000 title claims abstract description 70
- 238000004590 computer program Methods 0.000 claims description 16
- 238000003860 storage Methods 0.000 claims description 16
- 238000000605 extraction Methods 0.000 claims description 15
- 238000005457 optimization Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 description 13
- 230000006870 function Effects 0.000 description 9
- 230000006872 improvement Effects 0.000 description 8
- 230000008569 process Effects 0.000 description 7
- 238000012545 processing Methods 0.000 description 7
- 238000005516 engineering process Methods 0.000 description 6
- 238000003062 neural network model Methods 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000006399 behavior Effects 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000012935 Averaging Methods 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 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
- 238000011217 control strategy Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000004821 distillation Methods 0.000 description 1
- 238000013140 knowledge distillation 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
- 230000004044 response Effects 0.000 description 1
- 238000012954 risk control Methods 0.000 description 1
- 239000010979 ruby Substances 0.000 description 1
- 229910001750 ruby Inorganic materials 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
-
- 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
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
Abstract
The specification discloses a method and a device for model training and business wind control, which are used for privacy protection, and a main classifier can be assisted by each auxiliary classifier arranged in a lightweight model, so that the main classifier can integrate risk prediction results of each auxiliary classifier to obtain more accurate risk prediction results.
Description
Technical Field
The present disclosure relates to the field of computer technologies, and in particular, to a method and apparatus for model training and business wind control.
Background
With the development of internet technology, a Neural Network model (NN) has been receiving attention with its excellent data processing capability.
However, the neural network model with higher accuracy of the output result is often larger in scale and inconvenient to be deployed to a server, and the neural network model with smaller scale is relatively lower in accuracy of the output result. Therefore, the existing neural network model cannot always keep a smaller scale and ensure that the output result is higher in accuracy.
Disclosure of Invention
The specification provides a method and a device for model training and business wind control, which are used for solving the problem of poor accuracy of a small-scale neural network model in the prior art.
The technical scheme adopted in the specification is as follows:
the specification provides a model training method, comprising:
obtaining a lightweight model as a model to be trained, and obtaining a training sample for training the model to be trained, wherein the model to be trained comprises: a wind control model, the training sample comprising: sample business data;
inputting the training sample into the wind control model to obtain feature data of the training sample through a feature extraction layer in the wind control model;
inputting the characteristic data into each auxiliary classifier in the wind control model to obtain a risk prediction result of the auxiliary classifier for the training sample, wherein the risk prediction result is used as a first risk prediction result corresponding to the auxiliary classifier;
inputting each first risk prediction result into a main classifier in the wind control model, so that the main classifier obtains a risk prediction result aiming at the training sample according to each first risk prediction result and takes the risk prediction result as a second risk prediction result;
and training the wind control model according to the first risk prediction result, the second risk prediction result and the actual risk label of the training sample, wherein the wind control module is used for executing business wind control.
Optionally, training the wind control model according to the first risk prediction result, the second risk prediction result, and the actual risk label of the training sample specifically includes:
determining at least one target loss according to the first risk prediction result, the second risk prediction result and the actual risk label of the training sample, wherein the target loss comprises: a first penalty for characterizing a deviation between the first risk prediction results and the actual risk label, a second penalty for characterizing a deviation between the first risk prediction results and a target risk prediction result, a third penalty for characterizing a deviation between the target risk prediction result and the second risk prediction result, a fourth penalty for characterizing a deviation between the second risk prediction result and the actual risk label, a fifth penalty for characterizing a deviation between the first risk prediction results, the target risk prediction result being selected from the first risk prediction results based on a deviation between the first risk prediction results and the actual risk label, the smaller the deviation between the first risk prediction results, the larger the fifth penalty;
And training the model to be trained by taking the minimized target loss as an optimization target.
Optionally, based on a deviation between each first risk prediction result and the actual risk label, selecting the target risk prediction result from each first risk result, including:
and taking the first risk prediction result with the smallest deviation from the actual risk label as the selected target risk prediction result.
Optionally, determining at least one target loss according to the first risk prediction result, the second risk prediction result, and the actual risk label of the training sample specifically includes:
when the target loss is the first loss, determining a first sub-loss according to the deviation between the first risk prediction result and the actual risk label of the training sample for each first risk prediction result;
and determining the first loss according to each first sub-loss.
Optionally, determining at least one target loss according to the first risk prediction result, the second risk prediction result, and the actual risk label of the training sample specifically includes:
when the target loss is the second loss, determining a second sub-loss according to the deviation between the first risk prediction result and the target risk prediction result for each first risk prediction result;
Based on each second sub-loss, a second loss is determined.
Optionally, for each auxiliary classifier in the wind control model, the feature data is input into the auxiliary classifier to obtain a risk prediction result of the auxiliary classifier for the training sample, and the risk prediction result is used as a first risk prediction result corresponding to the auxiliary classifier, and specifically includes:
determining a service scene corresponding to the training sample as a target service scene;
determining all auxiliary classifiers matched with the target business scene from all auxiliary classifiers in the wind control model to be used as all first target classifiers;
inputting the characteristic data into each first target classifier to obtain a risk prediction result of the first target classifier for the training sample, wherein the risk prediction result is used as a first risk prediction result corresponding to the first target classifier;
inputting each first risk prediction result into a main classifier in the wind control model, so that the main classifier obtains a risk prediction result for the training sample according to each first risk prediction result, wherein the method specifically comprises the following steps:
determining a main classifier matched with the target service scene from all main classifiers in the wind control model as a second target classifier;
Inputting the first risk prediction results corresponding to the first target classifiers into the second target classifier, so that the second target classifier obtains the risk prediction results aiming at the training sample according to the first risk prediction results corresponding to the first target classifiers, and takes the risk prediction results as second risk prediction results.
The specification provides a method for business wind control, which comprises the following steps:
acquiring service data of a service to be executed;
inputting the business data into a wind control model to obtain a risk prediction result of the business to be executed according to the business data of the business to be executed through the wind control model, wherein the wind control model is trained by the model training method;
and executing corresponding tasks according to the risk prediction result of the service to be executed.
The present specification provides a model training apparatus comprising:
the system comprises an acquisition module, a training module and a training module, wherein the acquisition module is used for acquiring a light model as a model to be trained and acquiring a training sample for training the model to be trained, and the model to be trained comprises: a wind control model, the training sample comprising: sample business data;
the feature extraction module is used for inputting the training sample into the wind control model so as to obtain feature data of the training sample through a feature extraction layer in the wind control model;
The first classification module is used for inputting the characteristic data into each auxiliary classifier in the wind control model to obtain a risk prediction result of the auxiliary classifier for the training sample, and the risk prediction result is used as a first risk prediction result corresponding to the auxiliary classifier;
the second classification module is used for inputting each first risk prediction result into a main classifier in the wind control model so that the main classifier obtains a risk prediction result aiming at the training sample according to each first risk prediction result and takes the risk prediction result as a second risk prediction result;
the training module is used for training the wind control model according to the first risk prediction result, the second risk prediction result and the actual risk label of the training sample, and the wind control model is used for executing business wind control.
Optionally, the training module is specifically configured to determine at least one target loss according to the first risk prediction result, the second risk prediction result, and an actual risk tag of the training sample, where the target loss includes: a first penalty for characterizing a deviation between the first risk prediction results and the actual risk label, a second penalty for characterizing a deviation between the first risk prediction results and a target risk prediction result, a third penalty for characterizing a deviation between the target risk prediction result and the second risk prediction result, a fourth penalty for characterizing a deviation between the second risk prediction result and the actual risk label, a fifth penalty for characterizing a deviation between the first risk prediction results, the target risk prediction result being selected from the first risk prediction results based on a deviation between the first risk prediction results and the actual risk label, the smaller the deviation between the first risk prediction results, the larger the fifth penalty; and training the model to be trained by taking the minimized target loss as an optimization target.
Optionally, the training module is specifically configured to use a first risk prediction result with a minimum deviation from the actual risk label as the selected target risk prediction result.
Optionally, when the target loss is the first loss, the training module is specifically configured to determine, for each first risk prediction result, a first sub-loss according to a deviation between the first risk prediction result and an actual risk label of the training sample; and determining the first loss according to each first sub-loss.
Optionally, when the target loss is the second loss, the training module is specifically configured to determine, for each first risk prediction result, a second sub-loss according to a deviation between the first risk prediction result and the target risk prediction result; based on each second sub-loss, a second loss is determined.
Optionally, the first classification module is specifically configured to determine a service scenario corresponding to the training sample as a target service scenario; determining all auxiliary classifiers matched with the target business scene from all auxiliary classifiers in the wind control model to be used as all first target classifiers; inputting the characteristic data into each first target classifier to obtain a risk prediction result of the first target classifier for the training sample, wherein the risk prediction result is used as a first risk prediction result corresponding to the first target classifier;
The second classification module is specifically configured to determine, from each main classifier in the wind control model, a main classifier that matches the target service scene, as a second target classifier; inputting the first risk prediction results corresponding to the first target classifiers into the second target classifier, so that the second target classifier obtains the risk prediction results aiming at the training sample according to the first risk prediction results corresponding to the first target classifiers, and takes the risk prediction results as second risk prediction results.
The present specification provides a device for service wind control, including:
the service data acquisition module is used for acquiring service data of a service to be executed;
the wind control module is used for inputting the service data into a wind control model so as to obtain a risk prediction result of the service to be executed according to the service data of the service to be executed through the wind control model, wherein the wind control model is trained by the model training method;
and the service execution module is used for executing corresponding tasks according to the risk prediction result of the service to be executed.
The present specification provides a computer readable storage medium storing a computer program which when executed by a processor implements the method of model training, business wind control 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 the above-described method of model training, business wind control when executing the program.
The above-mentioned at least one technical scheme that this specification adopted can reach following beneficial effect:
according to the model training method provided by the specification, firstly, a light model is obtained as a model to be trained, the model to be trained comprises a wind control model, a training sample for training the wind control model is obtained, the training sample is input into the wind control model, characteristic data of the training sample are obtained through a characteristic extraction layer in the wind control model, the characteristic data are input into each auxiliary classifier in the wind control model, a risk prediction result of the auxiliary classifier for the training sample is obtained, the first risk prediction result corresponding to the auxiliary classifier is used as a first risk prediction result, each first risk prediction result is input into a main classifier in the wind control model, so that the main classifier obtains a risk prediction result for the training sample according to each first risk prediction result, and the wind control model is trained according to the first risk prediction result, the second risk prediction result and an actual risk label of the training sample as a second risk prediction result.
According to the method, the main classifier can be assisted by the auxiliary classifiers arranged in the lightweight model, so that the main classifier can integrate risk prediction results of the auxiliary classifiers to obtain more accurate risk prediction results.
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. In the drawings:
FIG. 1 is a schematic flow chart of a model training method provided in the present specification;
FIG. 2 is a schematic structural view of the wind control model provided in the present specification;
FIG. 3 is a schematic flow chart of a method for traffic wind control provided in the present specification;
FIG. 4 is a schematic diagram of a model training apparatus provided in the present specification;
FIG. 5 is a schematic diagram of a device for service air control provided in the present specification;
fig. 6 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 schematic flow chart of a model training method provided in the present specification, including the following steps:
s100: obtaining a lightweight model as a model to be trained, and obtaining a training sample for training the model to be trained, wherein the model to be trained comprises: a wind control model, the training sample comprising: sample traffic data.
In business scenarios such as payment risk monitoring and anti-fraud, the business platform generally classifies business data of the business to be executed through a wind control model to determine whether the business to be executed is a risk business, and further can control the risk business according to a preset risk business control strategy.
Before the risk prediction of the service to be executed is performed through the wind control model, training is needed, and the service can be deployed in a service platform.
Based on this, when the service platform needs to train the wind control model, a light model for wind control can be obtained as a model to be trained, corresponding marked service data can be obtained, and as a training sample, the marked service data can be determined according to a service scene, and can be as follows: structured data, sequence data, graphical data, text data, etc., such as: in the payment risk monitoring scenario, the structured data determined according to the payment behavior of the user is the service data, and further, for example: in a scene of judging whether the account number is logged in by the user, the historical behavior sequence data determined according to the historical login behavior of the user is the service data.
The light model may be obtained by compressing a complex wind control model obtained by pre-training through a knowledge distillation method or the like, or may be directly created.
After the service platform acquires the service data, the service data can be marked by a marking mode such as manual marking so as to obtain an actual risk label corresponding to each service data.
In the present specification, the execution body for implementing the model training method may refer to a designated device such as a server provided on a service platform, or may refer to a designated device such as a desktop computer or a notebook computer, and for convenience of description, the model training method provided in the present specification will be described below by taking the server as an example of the execution body.
S102: and inputting the training sample into the wind control model to obtain the characteristic data of the training sample through a characteristic extraction layer in the wind control model.
S104: and inputting the characteristic data into each auxiliary classifier in the wind control model to obtain a risk prediction result of the auxiliary classifier for the training sample, wherein the risk prediction result is used as a first risk prediction result corresponding to the auxiliary classifier.
After each training sample is obtained by the server, each training sample can be input into the wind control model so as to obtain the characteristic data of the training sample through the characteristic extraction layer in the wind control model.
Further, the server may input the feature data into each auxiliary classifier in the wind control model, to obtain a risk prediction result of the auxiliary classifier for the training sample, as a first risk prediction result corresponding to the auxiliary classifier.
The first risk prediction result may be a logic value corresponding to the training sample output by the auxiliary model for the training sample, where the logic value is a logarithm of a ratio Odds of probabilities of the training sample belonging to each class, for example: for a dice, the probability of the upward number of the dice being 6 per rotation is
The probability of not being 6 is
The number of each rotation of this die upwards is 6 and the corresponding Odds value is +.>
Logit value is +.>
。
The above-mentioned auxiliary classifiers require only one parameter corresponding to one full connection layer (Fully connected layers, FC), and thus have a low influence on the calculation efficiency of the model.
S106: and inputting each first risk prediction result into a main classifier in the wind control model, so that the main classifier obtains a risk prediction result aiming at the training sample according to each first risk prediction result and takes the risk prediction result as a second risk prediction result.
Further, the server may input each first risk prediction result into the main classifier in the wind control model, so that the main classifier obtains a risk prediction result for the training sample according to each first risk prediction result, and the risk prediction result is used as a second risk prediction result, as shown in fig. 2 in detail.
Fig. 2 is a schematic structural diagram of the wind control model provided in the present specification.
As can be seen from fig. 2, the server may add each auxiliary classifier between the feature extraction layer and the output layer, so that the auxiliary classifier is matched with the main classifier for use, and the classification result is obtained through an integrated learning manner, so that the accuracy of the model may be improved.
In addition, in a practical application scenario, it is generally required to perform multiple classification predictions for one service data, for example: judging whether the service to be executed corresponding to the service data is an illegal receipt and payment service, judging whether the service to be executed corresponding to the service data is a high risk service induced by others, and the like, wherein the existence of a plurality of wind control service types can be understood, and whether the service data belongs to the wind control service type or not is generally required to be judged according to each wind control service type, namely a plurality of classification labels corresponding to the service data are required to be determined.
Based on this, the wind control model may include: each auxiliary classifier group and each main classifier group comprise an auxiliary classifier or a main classifier for predicting whether the service data belongs to one type of classification label, for example: it is assumed that one auxiliary classifier set comprises three auxiliary classifiers, wherein a first auxiliary classifier is used for predicting whether the service data belongs to a type risk service, and a first auxiliary classifier is used for predicting whether the service data belongs to a type risk service.
Therefore, when the server inputs the feature data into the auxiliary classifier and inputs each first risk prediction result into the main classifier, the server can determine a service scene corresponding to the training sample, as a target service scene, from all the auxiliary classifiers in the model to be trained, determine each auxiliary classifier matched with the target service scene, as each first target classifier, input the feature data into the first target classifier for each first target classifier, obtain a risk prediction result of the first target classifier for the training sample, as a first risk prediction result corresponding to the first target classifier, and determine the main classifier matched with the target service scene from all the main classifiers in the wind control model, as a second target classifier, and input the first risk prediction result corresponding to each first target classifier into the second target classifier, so that the second target classifier obtains a risk prediction result for the training sample according to the first risk prediction result corresponding to each first target classifier, as a second risk prediction result.
It should be noted that, the wind control model obtained through training by the method can output the service data in multiple labels through each auxiliary classifier group and the main classifier group.
S108: and training the wind control model according to the first risk prediction result, the second risk prediction result and the actual risk label of the training sample.
After obtaining each first risk prediction result and each second risk prediction result through each auxiliary classifier and each main classifier, the server can determine at least one target loss according to each first risk prediction result, each second risk prediction result and an actual risk label of a training sample, wherein the target loss comprises: a first penalty for characterizing the deviation between each first risk prediction result and the actual risk tag, a second penalty for characterizing the deviation between each first risk prediction result and the target risk prediction result, a third penalty for characterizing the deviation between the target risk prediction result and the second risk prediction result, a fourth penalty for characterizing the deviation between the second risk prediction result and the actual risk tag, and a fifth penalty for characterizing the deviation between each first risk prediction result.
Specifically, when the first loss needs to be determined, the server may determine, for each first risk prediction result, a first sub-loss according to a deviation between the first risk prediction result and an actual risk label of the training sample, and determine, according to each first sub-loss, a first loss, where, according to each first sub-loss, a first loss may be determined by summing, averaging, and so on.
When the second loss needs to be determined, the server may determine, for each first risk prediction result, a second sub-loss according to a deviation between the first risk prediction result and the target risk prediction result, and determine a second loss according to each second sub-loss.
The target risk prediction result may be selected from the first risk prediction results based on a deviation between the first risk prediction results and the actual risk label.
Specifically, the server may use the first risk prediction result with the smallest deviation from the actual risk label as the selected target risk prediction result, where in the process of performing multiple rounds of iterative training on the wind control model, the selected auxiliary classifiers in each round of iterative training may be different, that is, the learned content of each auxiliary classifier in each round of iterative training is different, so each auxiliary classifier may be used as the target classifier, so that other auxiliary classifiers learn the auxiliary classifier.
It should be noted that, through the second and third losses, the auxiliary classifiers and the main classifier can be distilled and learned mutually in a mutual learning and distillation mode, so that the training effect on the wind control model is improved.
In the practical application scene, because when the wind control model is trained through the third loss, mutual learning is performed between each auxiliary classifier, so that the convergence problem (that is, the parameters of each auxiliary classifier are similar) occurs to each auxiliary classifier, and further, the training effect is reduced, therefore, in order to avoid the problem, certain difference between the first risk prediction results output by each auxiliary classifier can be ensured through the fifth loss, that is, through the negative learning mode, that is, the effect of 'sum but difference' is realized.
Specifically, when it is necessary to determine the fifth loss, the server may determine, for each first risk prediction result, a deviation value between the first risk prediction result and each other risk prediction result except for the first risk prediction result, and determine the fifth loss according to the respective deviation values.
The method for determining the fifth loss by the server according to the deviation value between the first risk prediction result and each other risk prediction result except the first risk prediction result may be two, and the two methods are respectively described in detail below.
The first way may be to determine the fifth loss according to the variance of each first risk prediction result, specifically, the following formula may be referred to:
in the above-mentioned formula(s),
namely fifth loss, < >>
I.e. i first risk prediction result,/->
I.e. the average value of the first risk prediction results.
The second method may be that the fifth loss is determined according to the difference between the first risk prediction results, and the following formula may be specifically referred to:
in the above-mentioned formula(s),
namely fifth loss, < >>
I.e. i first risk prediction result,/->
The k first risk prediction result is obtained.
According to the two methods, the mode that the deviation between the first risk prediction results is smaller and the fifth loss is larger can be adopted to enable the auxiliary trainers to keep certain difference in the process of training the wind control model, so that the training effect of the wind control model can be further improved.
It should be noted that, among the five losses, the first loss and the fourth loss are that the wind control model is trained by using the hard tag, the second loss and the third loss are that the wind control model is trained by using the soft tag, where the hard tag and the soft tag are two types of tags for model training, the hard tag is discrete data, generally is 0 or 1 in the two classification problems, that is, whether the training sample belongs to risk service, the actual risk tag of the training sample is the hard tag, and the soft tag is; and continuous data, namely whether the training sample belongs to the probability value of the risk service or not, wherein the target risk prediction result is a soft label.
Further, after determining the at least one target loss, the server may train the wind control model with the minimized target loss as an optimization target.
From the above, it can be seen that, not only can the risk prediction result be obtained by matching the auxiliary classifier with the main classifier, so that the accuracy of the risk prediction result is improved, but also each target loss can be determined according to the risk prediction result of each auxiliary classifier and the main classifier, so that each auxiliary classifier and the main classifier learn preset labels, and the auxiliary classifier and the main classifier can learn each other, so that the model training effect can be improved.
In order to further explain the above, the model training method will be described in detail below with respect to an example in which the model to be trained is a wind control model and the training sample is sample service data.
Specifically, the server may input the sample service data into the wind control model, so as to obtain feature data of the sample service data through a feature extraction layer in the wind control model, input the feature data into each auxiliary classifier in the wind control model, obtain a risk prediction result of the auxiliary classifier for the sample service data, as a first risk prediction result corresponding to the auxiliary classifier, input each first risk prediction result into a main classifier in the model to be trained, so that the main classifier obtains a risk prediction result for the sample service data according to each first risk prediction result, and trains the wind control model according to the first risk prediction result, the second risk prediction result and an actual risk tag of the sample data as a second risk prediction result.
The sample service data may be service data corresponding to a history service request initiated by a user and processed by the server.
The present disclosure also provides a method for performing service wind control by using a wind control model obtained by training the model training method, as shown in fig. 3.
Fig. 3 is a flow chart of a method for business wind control provided in the present specification, which includes the following steps:
s300: and acquiring service data of the service to be executed.
S302: and inputting the business data into a wind control model to obtain a risk prediction result of the business to be executed according to the business data of the business to be executed through the wind control model, wherein the wind control model is trained by the model training method.
S304: and executing corresponding tasks according to the risk prediction result of the service to be executed.
Before service execution is performed in response to a service request initiated by a user, the server can acquire service data of the service to be executed and input the service data into the wind control model so as to obtain feature data of the service to be executed through a feature extraction layer in the wind control model.
Further, through a feature extraction layer in the wind control model, feature data of the to-be-executed service are input into each auxiliary classifier of the wind control model to obtain a risk prediction result of each auxiliary classifier for the to-be-executed service, the risk prediction result of each auxiliary classifier for the to-be-executed service is input into a main classifier in the wind control model to obtain a risk prediction result of the to-be-executed service through the main classifier, and corresponding tasks can be executed according to the risk prediction result of the to-be-executed service.
The corresponding task may be that if the service to be executed is determined to be a risk service by the main classifier, risk management and control are performed on the service to be executed according to a preset risk management and control policy.
From the above, it can be seen that the risk control is performed on the service to be executed by the wind control model obtained through training by the model training method.
The above model training method provided for one or more embodiments of the present disclosure, based on the same thought, further provides a corresponding model training device, and a device for business wind control are shown in fig. 4 and fig. 5.
Fig. 4 is a schematic diagram of a model training apparatus provided in the present specification, the apparatus includes:
an obtaining module 401, configured to obtain a lightweight model as a model to be trained, and obtain a training sample for training the model to be trained, where the model to be trained includes: a wind control model, the training sample comprising: sample business data;
the feature extraction module 402 is configured to input the training sample into the wind control model, so as to obtain feature data of the training sample through a feature extraction layer in the wind control model;
the first classification module 403 is configured to input, for each auxiliary classifier in the wind control model, the feature data into the auxiliary classifier, to obtain a risk prediction result of the auxiliary classifier for the training sample, as a first risk prediction result corresponding to the auxiliary classifier;
The second classification module 404 is configured to input each first risk prediction result into a main classifier in the wind control model, so that the main classifier obtains a risk prediction result for the training sample according to each first risk prediction result, and uses the risk prediction result as a second risk prediction result;
the training module 405 is configured to train the wind control model according to the first risk prediction result, the second risk prediction result, and an actual risk tag of the training sample, where the wind control model is used to execute business wind control.
Optionally, the training module 405 is specifically configured to determine at least one target loss according to the first risk prediction result, the second risk prediction result, and an actual risk tag of the training sample, where the target loss includes: a first penalty for characterizing a deviation between the first risk prediction results and the actual risk label, a second penalty for characterizing a deviation between the first risk prediction results and a target risk prediction result, a third penalty for characterizing a deviation between the target risk prediction result and the second risk prediction result, a fourth penalty for characterizing a deviation between the second risk prediction result and the actual risk label, a fifth penalty for characterizing a deviation between the first risk prediction results, the target risk prediction result being selected from the first risk prediction results based on a deviation between the first risk prediction results and the actual risk label, the smaller the deviation between the first risk prediction results, the larger the fifth penalty; and training the model to be trained by taking the minimized target loss as an optimization target.
Optionally, the training module 405 is specifically configured to use a first risk prediction result with a minimum deviation from the actual risk label as the selected target risk prediction result.
Optionally, when the target loss is the first loss, the training module 405 is specifically configured to determine, for each first risk prediction result, a first sub-loss according to a deviation between the first risk prediction result and an actual risk label of the training sample; and determining the first loss according to each first sub-loss.
Optionally, when the target loss is the second loss, the training module 405 is specifically configured to determine, for each first risk prediction result, a second sub-loss according to a deviation between the first risk prediction result and the target risk prediction result; based on each second sub-loss, a second loss is determined.
Optionally, the first classification module 403 is specifically configured to determine a service scenario corresponding to the training sample as a target service scenario; determining all auxiliary classifiers matched with the target business scene from all auxiliary classifiers in the wind control model to be used as all first target classifiers; inputting the characteristic data into each first target classifier to obtain a risk prediction result of the first target classifier for the training sample, wherein the risk prediction result is used as a first risk prediction result corresponding to the first target classifier;
The second classification module 404 is specifically configured to determine, from each main classifier in the wind control model, a main classifier that matches the target service scene as a second target classifier; inputting the first risk prediction results corresponding to the first target classifiers into the second target classifier, so that the second target classifier obtains the risk prediction results aiming at the training sample according to the first risk prediction results corresponding to the first target classifiers, and takes the risk prediction results as second risk prediction results.
Fig. 5 is a schematic diagram of a device for service wind control provided in the present specification, where the device includes:
a service data obtaining module 501, configured to obtain service data of a service to be executed;
the wind control module 502 is configured to input the service data into a wind control model, so as to obtain a risk prediction result of the service to be executed according to the service data of the service to be executed through the wind control model, where the wind control model is obtained by training through the model training method;
and the service execution module 503 is configured to execute a corresponding task according to the risk prediction result of the service to be executed.
The present specification also provides a computer readable storage medium storing a computer program operable to perform a method of model training, business wind control as provided in fig. 1 above.
The present specification also provides a schematic structural diagram of an electronic device corresponding to fig. 1 shown in fig. 6. At the hardware level, as in fig. 6, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile storage, although it may include hardware required for other services. The processor reads the corresponding computer program from the nonvolatile memory into the memory and then runs the computer program to realize the model training and business wind control method of the figure 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 an 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 an example of the present specification and is not intended to limit the present specification. 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 (16)
1. A model training method, comprising:
obtaining a lightweight model as a model to be trained, and obtaining a training sample for training the model to be trained, wherein the model to be trained comprises: a wind control model, the training sample comprising: sample business data;
inputting the training sample into the wind control model to obtain feature data of the training sample through a feature extraction layer in the wind control model;
Inputting the characteristic data into each auxiliary classifier in the wind control model to obtain a risk prediction result of the auxiliary classifier for the training sample, wherein the risk prediction result is used as a first risk prediction result corresponding to the auxiliary classifier;
inputting each first risk prediction result into a main classifier in the wind control model, so that the main classifier obtains a risk prediction result aiming at the training sample according to each first risk prediction result and takes the risk prediction result as a second risk prediction result;
and training the wind control model according to the first risk prediction result, the second risk prediction result and the actual risk label of the training sample, wherein the wind control module is used for executing business wind control.
2. The method of claim 1, training the pneumatic control model according to the first risk prediction result, the second risk prediction result, and an actual risk tag of the training sample, specifically comprising:
determining at least one target loss according to the first risk prediction result, the second risk prediction result and the actual risk label of the training sample, wherein the target loss comprises: a first penalty for characterizing a deviation between the first risk prediction results and the actual risk label, a second penalty for characterizing a deviation between the first risk prediction results and a target risk prediction result, a third penalty for characterizing a deviation between the target risk prediction result and the second risk prediction result, a fourth penalty for characterizing a deviation between the second risk prediction result and the actual risk label, a fifth penalty for characterizing a deviation between the first risk prediction results, the target risk prediction result being selected from the first risk prediction results based on a deviation between the first risk prediction results and the actual risk label, the smaller the deviation between the first risk prediction results, the larger the fifth penalty;
And training the model to be trained by taking the minimized target loss as an optimization target.
3. The method of claim 2, selecting the target risk prediction result from the first risk results based on a deviation between the first risk prediction results and the actual risk label, specifically comprising:
and taking the first risk prediction result with the smallest deviation from the actual risk label as the selected target risk prediction result.
4. The method of claim 2, determining at least one target loss from the first risk prediction result, the second risk prediction result, and an actual risk tag of the training sample, comprising:
when the target loss is the first loss, determining a first sub-loss according to the deviation between the first risk prediction result and the actual risk label of the training sample for each first risk prediction result;
and determining the first loss according to each first sub-loss.
5. The method of claim 4, determining at least one target loss from the first risk prediction result, the second risk prediction result, and an actual risk signature of the training sample, comprising:
When the target loss is the second loss, determining a second sub-loss according to the deviation between the first risk prediction result and the target risk prediction result for each first risk prediction result;
based on each second sub-loss, a second loss is determined.
6. The method of claim 1, wherein for each auxiliary classifier in the wind control model, the feature data is input into the auxiliary classifier to obtain a risk prediction result of the auxiliary classifier for the training sample, and the risk prediction result is used as a first risk prediction result corresponding to the auxiliary classifier, and specifically includes:
determining a service scene corresponding to the training sample as a target service scene;
determining all auxiliary classifiers matched with the target business scene from all auxiliary classifiers in the wind control model to be used as all first target classifiers;
inputting the characteristic data into each first target classifier to obtain a risk prediction result of the first target classifier for the training sample, wherein the risk prediction result is used as a first risk prediction result corresponding to the first target classifier;
inputting each first risk prediction result into a main classifier in the wind control model, so that the main classifier obtains a risk prediction result for the training sample according to each first risk prediction result, wherein the method specifically comprises the following steps:
Determining a main classifier matched with the target service scene from all main classifiers in the wind control model as a second target classifier;
inputting the first risk prediction results corresponding to the first target classifiers into the second target classifier, so that the second target classifier obtains the risk prediction results aiming at the training sample according to the first risk prediction results corresponding to the first target classifiers, and takes the risk prediction results as second risk prediction results.
7. A method of business wind control, comprising:
acquiring service data of a service to be executed;
inputting the service data into a wind control model to obtain a risk prediction result of the service to be executed according to the service data of the service to be executed through the wind control model, wherein the wind control model is obtained through training by the method of any one of claims 1-6;
and executing corresponding tasks according to the risk prediction result of the service to be executed.
8. A model training apparatus comprising:
the system comprises an acquisition module, a training module and a training module, wherein the acquisition module is used for acquiring a light model as a model to be trained and acquiring a training sample for training the model to be trained, and the model to be trained comprises: a wind control model, the training sample comprising: sample business data;
The feature extraction module is used for inputting the training sample into the wind control model so as to obtain feature data of the training sample through a feature extraction layer in the wind control model;
the first classification module is used for inputting the characteristic data into each auxiliary classifier in the wind control model to obtain a risk prediction result of the auxiliary classifier for the training sample, and the risk prediction result is used as a first risk prediction result corresponding to the auxiliary classifier;
the second classification module is used for inputting each first risk prediction result into a main classifier in the wind control model so that the main classifier obtains a risk prediction result aiming at the training sample according to each first risk prediction result and takes the risk prediction result as a second risk prediction result;
the training module is used for training the wind control model according to the first risk prediction result, the second risk prediction result and the actual risk label of the training sample, and the wind control model is used for executing business wind control.
9. The apparatus of claim 8, the training module being specifically configured to determine at least one target loss based on the first risk prediction result, the second risk prediction result, and an actual risk tag of the training sample, the target loss comprising: a first penalty for characterizing a deviation between the first risk prediction results and the actual risk label, a second penalty for characterizing a deviation between the first risk prediction results and a target risk prediction result, a third penalty for characterizing a deviation between the target risk prediction result and the second risk prediction result, a fourth penalty for characterizing a deviation between the second risk prediction result and the actual risk label, a fifth penalty for characterizing a deviation between the first risk prediction results, the target risk prediction result being selected from the first risk prediction results based on a deviation between the first risk prediction results and the actual risk label, the smaller the deviation between the first risk prediction results, the larger the fifth penalty; and training the model to be trained by taking the minimized target loss as an optimization target.
10. The apparatus of claim 9, the training module being specifically configured to use a first risk prediction result with a minimum deviation from the actual risk tag as the selected target risk prediction result.
11. The apparatus of claim 9, the training module being specifically configured to, when the target loss is the first loss, determine, for each first risk prediction result, a first sub-loss based on a deviation between the first risk prediction result and an actual risk tag of the training sample; and determining the first loss according to each first sub-loss.
12. The apparatus of claim 11, the training module being specifically configured to determine, for each first risk prediction result, a second sub-loss based on a deviation between the first risk prediction result and the target risk prediction result when the target loss is the second loss; based on each second sub-loss, a second loss is determined.
13. The apparatus of claim 8, wherein the first classification module is specifically configured to determine a service scenario corresponding to the training sample as a target service scenario; determining all auxiliary classifiers matched with the target business scene from all auxiliary classifiers in the wind control model to be used as all first target classifiers; inputting the characteristic data into each first target classifier to obtain a risk prediction result of the first target classifier for the training sample, wherein the risk prediction result is used as a first risk prediction result corresponding to the first target classifier;
The second classification module is specifically configured to determine, from each main classifier in the wind control model, a main classifier that matches the target service scene, as a second target classifier; inputting the first risk prediction results corresponding to the first target classifiers into the second target classifier, so that the second target classifier obtains the risk prediction results aiming at the training sample according to the first risk prediction results corresponding to the first target classifiers, and takes the risk prediction results as second risk prediction results.
14. A business air control device, comprising:
the service data acquisition module is used for acquiring service data of a service to be executed;
the wind control module is used for inputting the service data into a wind control model so as to obtain a risk prediction result of the service to be executed according to the service data of the service to be executed through the wind control model, wherein the wind control model is obtained through training by the method of any one of claims 1-6;
and the service execution module is used for executing corresponding tasks according to the risk prediction result of the service to be executed.
15. A computer readable storage medium storing a computer program which, when executed by a processor, implements the method of any of the preceding claims 1-7.
16. 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 1-7 when executing the program.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310632423.4A CN116340852B (en) | 2023-05-30 | 2023-05-30 | Model training and business wind control method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310632423.4A CN116340852B (en) | 2023-05-30 | 2023-05-30 | Model training and business wind control method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116340852A true CN116340852A (en) | 2023-06-27 |
| CN116340852B CN116340852B (en) | 2023-09-15 |
Family
ID=86880821
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310632423.4A Active CN116340852B (en) | 2023-05-30 | 2023-05-30 | Model training and business wind control method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116340852B (en) |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110059854A (en) * | 2019-03-13 | 2019-07-26 | 阿里巴巴集团控股有限公司 | Method and device for risk identification |
| CN111027870A (en) * | 2019-12-14 | 2020-04-17 | 支付宝(杭州)信息技术有限公司 | User risk assessment method and device, electronic equipment and storage medium |
| CN112330035A (en) * | 2020-11-10 | 2021-02-05 | 支付宝(杭州)信息技术有限公司 | Training method and device of risk prediction model |
| CN112926663A (en) * | 2021-02-25 | 2021-06-08 | 中国平安人寿保险股份有限公司 | Method and device for training classification model, computer equipment and storage medium |
| CN114358197A (en) * | 2022-01-10 | 2022-04-15 | 上海亿保健康管理有限公司 | Method and device for training classification model, electronic equipment and storage medium |
| US20220138495A1 (en) * | 2020-11-05 | 2022-05-05 | University Of Electronic Science And Technology Of China | Model and method for multi-source domain adaptation by aligning partial features |
| US11605465B1 (en) * | 2018-08-16 | 2023-03-14 | Clarify Health Solutions, Inc. | Computer network architecture with machine learning and artificial intelligence and patient risk scoring |
| CN115908924A (en) * | 2022-11-23 | 2023-04-04 | 中国地质大学（武汉） | Multi-classifier-based small sample hyperspectral image semantic segmentation method and system |
-
2023
- 2023-05-30 CN CN202310632423.4A patent/CN116340852B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11605465B1 (en) * | 2018-08-16 | 2023-03-14 | Clarify Health Solutions, Inc. | Computer network architecture with machine learning and artificial intelligence and patient risk scoring |
| CN110059854A (en) * | 2019-03-13 | 2019-07-26 | 阿里巴巴集团控股有限公司 | Method and device for risk identification |
| CN111027870A (en) * | 2019-12-14 | 2020-04-17 | 支付宝(杭州)信息技术有限公司 | User risk assessment method and device, electronic equipment and storage medium |
| US20220138495A1 (en) * | 2020-11-05 | 2022-05-05 | University Of Electronic Science And Technology Of China | Model and method for multi-source domain adaptation by aligning partial features |
| CN112330035A (en) * | 2020-11-10 | 2021-02-05 | 支付宝(杭州)信息技术有限公司 | Training method and device of risk prediction model |
| CN112926663A (en) * | 2021-02-25 | 2021-06-08 | 中国平安人寿保险股份有限公司 | Method and device for training classification model, computer equipment and storage medium |
| CN114358197A (en) * | 2022-01-10 | 2022-04-15 | 上海亿保健康管理有限公司 | Method and device for training classification model, electronic equipment and storage medium |
| CN115908924A (en) * | 2022-11-23 | 2023-04-04 | 中国地质大学（武汉） | Multi-classifier-based small sample hyperspectral image semantic segmentation method and system |
Non-Patent Citations (4)
| Title |
|---|
| GIAN ANTONIO SUSTO等: "Machine Learning for Predictive Maintenance: A Multiple Classifier Approach", 《IEEE TRANSACTIONS ON INDUSTRIAL INFORMATICS》, vol. 11, no. 3, pages 812 - 820, XP011583481, DOI: 10.1109/TII.2014.2349359 * |
| NAKUL AGARWAL等: "Improving multiclass classification by deep networks using DAGSVM and Triplet Loss☆", 《PATTERN RECOGNITION LETTERS》, vol. 112, pages 184 - 190, XP085469302, DOI: 10.1016/j.patrec.2018.06.034 * |
| 普子恒等: "基于大数据分类与SVDD模型的多场景500kV劣化绝缘子判定研究", 《电瓷避雷器》, no. 2, pages 189 - 196 * |
| 马娜等: "具有类间差异约束的多对抗深度域适应模型", 《计算机科学与探索》, vol. 17, no. 5, pages 1168 - 1179 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116340852B (en) | 2023-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110008991B (en) | Risk event identification method, risk identification model generation method, risk event identification device, risk identification equipment and risk identification medium | |
| CN115828162B (en) | Classification model training method and device, storage medium and electronic equipment | |
| CN115712866A (en) | Data processing method, device and equipment | |
| CN116304720A (en) | Cost model training method and device, storage medium and electronic equipment | |
| CN116757278B (en) | Training method and device of prediction model, storage medium and electronic equipment | |
| CN116308738B (en) | Model training method, business wind control method and device | |
| CN117093862A (en) | Model training method and device, electronic equipment and storage medium | |
| CN117313141A (en) | Abnormality detection method, abnormality detection device, abnormality detection equipment and readable storage medium | |
| CN115618964B (en) | Model training method and device, storage medium and electronic equipment | |
| CN116340852B (en) | Model training and business wind control method and device | |
| CN116152933A (en) | Training method, device, equipment and storage medium of anomaly detection model | |
| CN116246276A (en) | Information identification method, device, equipment and readable storage medium | |
| CN115545938B (en) | Method, device, storage medium and equipment for executing risk identification service | |
| CN116434787B (en) | Voice emotion recognition method and device, storage medium and electronic equipment | |
| CN116843466A (en) | Risk identification model training method and device, storage medium and electronic equipment | |
| CN116578877B (en) | Method and device for model training and risk identification of secondary optimization marking | |
| CN112115952B (en) | Image classification method, device and medium based on full convolution neural network | |
| CN116957330A (en) | Model training method, service wind control method, device, equipment and storage medium | |
| CN116363418A (en) | Method and device for training classification model, storage medium and electronic equipment | |
| CN117575611A (en) | Risk identification method and device, storage medium and electronic equipment | |
| CN117973867A (en) | Risk control method, apparatus, storage medium and device | |
| CN116401541A (en) | Model training method and device, storage medium and electronic equipment | |
| CN117592102A (en) | Service execution method, device, equipment and storage medium | |
| CN117313739A (en) | Training method, device, equipment and storage medium of language model | |
| CN117520850A (en) | Model training method and device, storage medium and electronic equipment |
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 |