Risky equipment identification method and device based on cold start scenario domain adaptive learning

Abstract

The present invention discloses a method and device for identifying risky equipment based on domain adaptive learning of cold start scenarios, the method comprising: configuring an auxiliary task model associated with a risk identification model; performing domain adaptive learning on a source domain and a target domain based on the auxiliary task model; selecting training samples corresponding to the source domain and the target domain according to a cold start scenario of the target domain; inputting the selected training samples into a risk identification model for training; and using the trained risk identification model to identify equipment risks in cold start scenarios. The present invention configures an auxiliary task model associated with a risk identification model; performs domain adaptive learning on the source domain and the target domain based on the auxiliary task model, thereby migrating features of a source domain with sufficient labeled samples to a target domain with only a small number of labeled samples, and training a risk identification model that can be applied to the target domain through the selected training samples, thereby being able to identify equipment risks in cold start scenarios in the target domain.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to a method, device, electronic device and computer-readable medium for identifying risky devices based on cold start scenario domain adaptive learning. Background Art

[0002] With the development of the Internet, various Internet service platforms have emerged, such as online shopping platforms, online car-hailing platforms, sharing platforms, maps, music, etc. These platforms have brought great convenience to people's lives, but because they are closely integrated with the Internet, there will be some risks such as fraud and breach of trust.

[0003] In order to identify these risky devices, we usually collect sample data related to the device and use machine learning models to predict the risk value of the device. However, before a new service or product is put into production, there is no or only a small amount of sample data. The performance period of these sample data is not long enough, resulting in the number of labeled samples being less than the preset sample size, which is called a cold start. Because cold start will cause insufficient training samples, it is impossible to effectively train the machine learning model, which affects the development of new business. Summary of the invention

[0004] In view of this, the main purpose of the present invention is to propose a risk device identification method, device, electronic device and computer-readable medium based on cold start scenario field adaptive learning, in order to at least partially solve at least one of the above-mentioned technical problems.

[0005] In order to solve the above technical problems, the first aspect of the present invention proposes a risk device identification method based on cold start scenario domain adaptation learning, the method comprising:

[0006] configuring an auxiliary task model associated with the risk identification model;

[0007] Performing domain adaptation learning on the source domain and the target domain based on the auxiliary task model;

[0008] Select training samples corresponding to the source domain and the target domain according to the target domain cold start scenario;

[0009] Input the selected training samples into the risk identification model for training;

[0010] Use the trained risk identification model to identify equipment risks in cold start scenarios;

[0011] The cold start refers to a situation where the number of labeled samples is less than the preset number of samples due to insufficient sample performance period.

[0012] According to a preferred embodiment of the present invention, the performing domain adaptation learning on the source domain and the target domain based on the auxiliary task model includes:

[0013] Optimize the marginal probability distance or local marginal probability distance between the source domain and the target domain through the auxiliary task model;

[0014] Alternatively, the classification error between the source and target domains is maximized through a gradient reversal layer between the source and target domains.

[0015] According to a preferred embodiment of the present invention, the domain adaptation learning is:

[0016]

[0017] Among them, l is the loss function, n is the number of samples, v i is the sample weight, λ is the complexity weight, R is the domain adaptation learning regularization term, T is the feature transformation function, x i is the characteristic data of the i-th sample, y i is the label data of the i-th sample, D s represents the source domain and D t Indicates the target domain.

[0018] According to a preferred implementation manner of the present invention, the cold start scenarios include: a new business scenario with a sample size less than a threshold, a restart business scenario containing only old samples older than a predetermined period, and a new business scenario in other regions.

[0019] According to a preferred embodiment of the present invention, selecting training samples corresponding to the source domain and the target domain according to the target domain cold start scenario includes:

[0020] For new business scenarios where the sample size is less than the threshold and new business scenarios in other regions, the source domain samples are used as training samples;

[0021] For the restart business scenario that only contains old samples older than a predetermined period, the source domain and target domain samples are mixed as training samples.

[0022] In order to solve the above technical problems, the second aspect of the present invention provides a risk device identification device based on cold start scenario domain adaptive learning, the device comprising:

[0023] A configuration module, used for configuring an auxiliary task model associated with the risk identification model;

[0024] An adaptation module, used for performing domain adaptation learning on a source domain and a target domain based on the auxiliary task model;

[0025] A selection module is used to select training samples corresponding to the source domain and the target domain according to the target domain cold start scenario;

[0026] A training module, used to input the selected training samples into the risk identification model for training;

[0027] An identification module is used to identify equipment risks in cold start scenarios using a trained risk identification model;

[0028] The cold start refers to a situation where the number of labeled samples is less than the preset number of samples due to insufficient sample performance period.

[0029] According to a preferred embodiment of the present invention, the adaptation module optimizes the edge probability distance or the local edge probability distance between the source domain and the target domain through an auxiliary task model;

[0030] Alternatively, the adaptation module maximizes the classification error between the source domain and the target domain through a gradient reversal layer between the source domain and the target domain.

[0031] According to a preferred embodiment of the present invention, the domain adaptation learning is:

[0032]

[0033] Among them, l is the loss function, n is the number of samples, v i is the sample weight, λ is the complexity weight, R is the domain adaptation learning regularization term, T is the feature transformation function, x i is the characteristic data of the i-th sample, y i is the label data of the i-th sample, D s represents the source domain and D t Indicates the target domain.

[0034] According to a preferred implementation manner of the present invention, the cold start scenarios include: a new business scenario with a sample size less than a threshold, a restart business scenario containing only old samples older than a predetermined period, and a new business scenario in other regions.

[0035] According to a preferred embodiment of the present invention, the selection module uses source domain samples as training samples for newly opened business scenarios with a sample size less than a threshold and newly opened business scenarios in other regions; and mixes source domain and target domain samples as training samples for restarted business scenarios that only contain old samples older than a predetermined period.

[0036] In order to solve the above technical problems, the third aspect of the present invention provides an electronic device, including:

[0037] Processor; and

[0038] A memory storing computer executable instructions, which when executed cause the processor to perform the above method.

[0039] To solve the above technical problem, the fourth aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores one or more programs, and when the one or more programs are executed by a processor, the above method is implemented.

[0040] The present invention configures an auxiliary task model associated with a risk identification model; performs domain adaptation learning on the source domain and the target domain based on the auxiliary task model, thereby migrating the features of the source domain with sufficient labeled samples to the target domain with only a small number of labeled samples, and then selects training samples corresponding to the source domain and the target domain according to the cold start scenario of the target domain; trains a risk identification model that can be applied to the target domain through the selected training samples, thereby being able to identify the equipment risks in the cold start scenario of the target domain. This solves the problem that model training cannot be performed due to a small number of labeled samples, which affects the development of new businesses. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to make the technical problems solved by the present invention, the technical means adopted and the technical effects achieved more clearly, the specific embodiments of the present invention will be described in detail with reference to the accompanying drawings. However, it should be noted that the drawings described below are only drawings of exemplary embodiments of the present invention, and those skilled in the art can obtain drawings of other embodiments based on these drawings without creative work.

[0042] Figure 1 It is a flowchart of a method for identifying risky devices based on cold start scenario domain adaptive learning according to an embodiment of the present invention;

[0043] Figure 2a-2b They are two schematic diagrams of domain adaptation learning for a source domain and a target domain according to an embodiment of the present invention;

[0044] Figure 3 It is a schematic diagram of the structural framework of a risk device identification device based on cold start scenario domain adaptive learning according to an embodiment of the present invention;

[0045] Figure 4 is a structural block diagram of an exemplary embodiment of an electronic device according to the present invention;

[0046] Figure 5 is a schematic diagram of an embodiment of a computer readable medium of the present invention. DETAILED DESCRIPTION

[0047] The exemplary embodiments of the present invention will now be described more fully with reference to the accompanying drawings. Although each exemplary embodiment can be implemented in a variety of specific ways, it should not be understood that the present invention is limited to the embodiments set forth herein. On the contrary, these exemplary embodiments are provided to make the content of the present invention more complete and to facilitate the comprehensive communication of the inventive concept to those skilled in the art.

[0048] Under the premise of conforming to the technical concept of the present invention, the structure, performance, effect or other characteristics described in a specific embodiment may be combined with one or more other embodiments in any appropriate manner.

[0049] In the process of introducing specific embodiments, the detailed description of the structure, performance, effect or other characteristics is to enable those skilled in the art to fully understand the embodiments. However, it does not exclude that those skilled in the art can implement the present invention with a technical solution that does not contain the above-mentioned structure, performance, effect or other characteristics under certain circumstances.

[0050] The flowchart in the accompanying drawings is only an exemplary process demonstration, and does not mean that the solution of the present invention must include all the contents, operations and steps in the flowchart, nor does it mean that it must be executed in the order shown in the figure. For example, some operations / steps in the flowchart can be decomposed, some operations / steps can be combined or partially combined, etc. The execution order shown in the flowchart can be changed according to the actual situation without departing from the main purpose of the present invention.

[0051] Frames in the attached figure Figure 1 Generally, it refers to functional entities, which do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software form, or in one or more hardware modules or integrated circuits, or in different networks and / or processor devices and / or microcontroller devices.

[0052] The same reference numerals in the drawings represent the same or similar elements, components or parts, and thus the repeated description of the same or similar elements, components or parts may be omitted below. It should also be understood that although the first, second, third and other attributives representing the numbers may be used herein to describe various devices, elements, components or parts, these devices, elements, components or parts should not be limited by these attributives. In other words, these attributives are only used to distinguish one from another. For example, the first device may also be called the second device, but it does not deviate from the essential technical solution of the present invention. In addition, the terms "and / or" and "and / or" refer to all combinations including any one or more of the listed items.

[0053] See also Figure 1 , Figure 1The present invention provides a method for identifying risky devices based on cold start scenario domain adaptive learning. The method is used to identify risky devices in Internet services, wherein the Internet services can be services provided by various Internet service platforms, and the Internet service platforms can be online shopping platforms, online car-hailing platforms, sharing platforms, search platforms, social platforms, etc. As long as the services provided based on the Internet are applicable, the present invention does not make specific limitations. Figure 1 As shown, the method includes:

[0054] S1. Configure the auxiliary task model associated with the risk identification model;

[0055] In this embodiment, it is necessary to configure the auxiliary task model associated with the risk identification model according to the business scenario of the risk identification model application and the identified risk category; that is, the auxiliary task model is related to the business scenario of the risk identification model application and the identified risk type. Exemplarily, the task of identifying the first risk type in the business scenario of the risk identification model application can be used as an auxiliary task, and the corresponding auxiliary task model can be configured. Among them, the first risk type is a risk type different from the risk type identified by the risk identification model. For example: in the overdue business scenario, the risk category identified by the risk identification model is overdue, and the identification of fraud can be used as an auxiliary task and the corresponding fraud model can be configured; in the credit assessment business scenario, the risk category identified by the risk identification model is dishonesty, and the identification of fraud can be used as an auxiliary task and the corresponding fraud model can be configured.

[0056] S2. Performing domain adaptation learning on the source domain and the target domain based on the auxiliary task model;

[0057] In this embodiment, the number of labeled samples in the source domain is greater than the first threshold, that is, the source domain has enough labeled samples, and the target domain is in a cold start scenario, resulting in the number of labeled samples in the target domain being less than the preset sample number, that is, the target domain has only a small number of, or no, labeled samples. The first threshold is much greater than the preset sample number.

[0058] The cold start refers to the situation where the number of labeled samples is less than the preset number of samples due to insufficient sample performance period. Exemplarily, the cold start scenarios include: a newly opened business scenario with a sample size less than a threshold (e.g., a newly opened business with only a small number of labeled samples), a restarted business scenario containing only old samples older than a predetermined period (e.g., a business was forced to stop for 3 months and then restarted, and the restarted business only contains old samples from 3 months ago), and a newly opened business scenario in other regions (e.g., conducting similar business in other regions).

[0059] The domain adaptation learning described in the present invention is described in detail below.

[0060] For any sample (x i ,yi ),x i ∈X,y i ∈Y, D={X,Y,P(x,y)} is called a domain containing the sample. Among them, X is the feature space and Y is the label space. For structured data, the feature space is all the predictive variables, and the label space is the classification category and regression interval. i is the characteristic data of the i-th sample, y i is the label data of the i-th sample, P(x,y) is the probability distribution that the sample obeys, and conventional machine learning models need to satisfy the independent and identically distributed assumption.

[0061] Domain adaptation learning can be applied to a given source domain D when any of the following conditions exist: s and the target domain D t In: (1) The feature space is different, X s ≠X t ; (2) Different label spaces, Y s ≠Y t ; (3) The sample probability distribution space is different, P s (x,y)≠P t (x,y).

[0062] On the other hand, the supervised machine learning model for conventional structural risk minimization can be defined as follows

[0063]

[0064] Among them, l represents the loss function, n represents the number of samples, R(f) is the model complexity, λ is the complexity weight, and x i is the characteristic data of the i-th sample, y i is the predicted value of the ith sample.

[0065] Then the domain adaptation learning is:

[0066]

[0067] Among them, l is the loss function, n is the number of samples, v i is the sample weight, λ is the complexity weight, R is the domain adaptation learning regularization term, T is the feature transformation function, x i is the characteristic data of the i-th sample, y i is the label data of the i-th sample.

[0068] Based on the above formula (1), domain adaptation learning can be divided into the following three types:

[0069] (1) Sample weight transfer: by adjusting the sample weight, the samples that are less likely to be correctly classified in the source domain are given smaller weights, and the samples that are less likely to be correctly classified in the target domain are given larger weights; for example, the Tradaboost algorithm.

[0070] (2) Feature transformation migration: By learning the feature transformation function T, we can reduce the R term so that the marginal distribution P(T(X s )) for P(T(X t )), conditional distribution P(Y s |T(X s )) for P(Y t |T(X t )) are as close as possible; for example, the joint distribution adaptation JDA algorithm.

[0071] (3) Pre-training model migration, let R(T(D s ),T(D t ))=R(f s ,T(D t ), that is, the model f learned in the source domain s Fine-tuning the data in the target domain is common in the fields of computer vision (CV) and natural language processing (NLP).

[0072] In this embodiment, the domain adaptation learning can be defined as follows:

[0073] The source domain is D s ,D t ; The target domain is D t , the feature space of the source domain and the target domain is the same: X s =X t ; The label space of the source domain and the target domain is the same: Y s =Y t ; The sample probability distribution spaces of the source domain and the target domain are different: P s (x,y)≠P t (x,y); the target domain can be labeled or completely unlabeled.

[0074] In one example, domain adaptation learning of the source domain and the target domain can be completed by optimizing the marginal probability distance between the source domain and the target domain through the auxiliary task model, or domain adaptation learning of the source domain and the target domain can be completed by optimizing the local marginal probability distance between the source domain and the target domain through the auxiliary task model. Figure 2a As shown in the figure, the source domain samples and target domain samples of the auxiliary task are first input into the shared layer for processing. The auxiliary task model optimizes R(T(D s ),T(D t )) to learn the feature transformation function T, R(T(Ds ),T(D t )) can be defined as the marginal probability distance MMD (Maximum Mean Discrepancy) between the source domain and the target domain; or, R(T(D s ),T(D t )) can be defined as the local marginal probability distance LMMD (Local Maximum Mean Discrepancy) between the source domain and the target domain to complete domain adaptation learning for the source domain and the target domain.

[0075] In another example, if Figure 2b The source domain samples and target domain samples of the auxiliary task are first input into the shared layer for processing, and then the classification error between the source domain and the target domain is maximized through the gradient reversal layer between the source domain and the target domain, eliminating the recognition of the domain while identifying the correct category, thereby completing the domain adaptation learning of the source domain and the target domain.

[0076] S3, selecting training samples corresponding to the source domain and the target domain according to the target domain cold start scenario;

[0077] Take the target domain cold start scenario including: a new business scenario with a sample size less than a threshold, a restarted business scenario containing only old samples earlier than a predetermined period, and a new business scenario in other regions as examples: for the new business scenario with a sample size less than a threshold and the new business scenario in other regions, the source domain samples are used as training samples; for the restarted business scenario containing only old samples earlier than a predetermined period, the source domain and target domain samples are mixed as training samples.

[0078] S4, inputting the selected training samples into the risk identification model for training;

[0079] In this embodiment, when there are no labels in the target domain cold start scenario, semi-supervised learning is used for training, that is, only source domain samples are sent to the main task learning, which is applicable to scenarios where new businesses are launched and samples have not yet entered the performance period; when the target domain has labels, supervised learning is used for training, that is, source domain and target domain samples are mixed together and sent to the main task learning, which is applicable to scenarios where the sample size needs to be expanded but the assumption of the same distribution of long-term and short-term samples is invalid.

[0080] S5. Use the trained risk identification model to identify equipment risks in cold start scenarios;

[0081] The risk identification model trained through the above-mentioned domain adaptation learning can identify the equipment risks of the cold start scenario in the target domain. The equipment data of the cold start scenario in the target domain is input into the trained risk identification model, and the equipment risks can be identified based on the output results.

[0082] Figure 3The present invention is a risk device identification device based on cold start scenario domain adaptive learning, such as Figure 3 As shown, the device comprises:

[0083] A configuration module 31, used to configure an auxiliary task model associated with the risk identification model;

[0084] An adaptation module 32, configured to perform domain adaptation learning on a source domain and a target domain based on the auxiliary task model;

[0085] A selection module 33 is used to select training samples corresponding to the source domain and the target domain according to the target domain cold start scenario;

[0086] The training module 34 is used to input the selected training samples into the risk identification model for training;

[0087] An identification module 35, used to identify equipment risks in a cold start scenario using a trained risk identification model;

[0088] The cold start refers to a situation where the number of labeled samples is less than the preset number of samples due to insufficient sample performance period.

[0089] In one embodiment, the adaptation module 32 optimizes the marginal probability distance or the local marginal probability distance between the source domain and the target domain through the auxiliary task model;

[0090] Alternatively, the adaptation module 32 maximizes the classification error between the source domain and the target domain through a gradient reversal layer between the source domain and the target domain.

[0091] The domain adaptation learning is:

[0092]

[0093] Among them, l is the loss function, n is the number of samples, v i is the sample weight, λ is the complexity weight, R is the domain adaptation learning regularization term, and T is the feature transformation function.

[0094] In one example, the cold start scenario includes: a newly started business scenario with a sample size less than a threshold, a restarted business scenario containing only old samples older than a predetermined period, and a newly started business scenario in other regions.

[0095] The selection module 33 uses source domain samples as training samples for new business scenarios with sample sizes less than a threshold and new business scenarios in other regions; and mixes source domain and target domain samples as training samples for restarted business scenarios that only include old samples older than a predetermined period.

[0096] Those skilled in the art will appreciate that the modules in the above device embodiments may be distributed in the device as described, or may be changed accordingly and distributed in one or more devices different from the above embodiments. The modules in the above embodiments may be combined into one module, or may be further split into multiple submodules.

[0097] The electronic device embodiment of the present invention is described below, and the electronic device can be regarded as a physical implementation of the method and device embodiments of the present invention described above. The details described in the electronic device embodiment of the present invention should be regarded as a supplement to the above method or device embodiments; details not disclosed in the electronic device embodiment of the present invention can be implemented with reference to the above method or device embodiments.

[0098] Figure 4 is a structural block diagram of an exemplary embodiment of an electronic device according to the present invention. Figure 4 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0099] like Figure 4 As shown, the electronic device 400 of this exemplary embodiment is in the form of a general data processing device. The components of the electronic device 400 may include, but are not limited to: at least one processing unit 410, at least one storage unit 420, a bus 430 connecting different electronic device components (including the storage unit 420 and the processing unit 410), a display unit 440, etc.

[0100] The storage unit 420 stores a computer-readable program, which may be a source program or a code of a read-only program. The program may be executed by the processing unit 410, so that the processing unit 410 performs the steps of various embodiments of the present invention. For example, the processing unit 410 may perform the following steps: Figure 1 Steps shown.

[0101] The storage unit 420 may include a readable medium in the form of a volatile storage unit, such as a random access storage unit (RAM) 4201 and / or a cache storage unit 4202, and may further include a read-only storage unit (ROM) 4203. The storage unit 420 may also include a program / utility 4204 having a set (at least one) of program modules 4205, such program modules 4205 including but not limited to: operating the electronic device, one or more application programs, other program modules, and program data, each of which or some combination may include the implementation of a network environment.

[0102] Bus 430 may represent one or more of several types of bus structures, including a memory unit bus or memory unit controller, a peripheral bus, an accelerated graphics port, a processing unit, or a local bus using any of a variety of bus architectures.

[0103] The electronic device 400 may also communicate with one or more external devices 100 (e.g., keyboards, displays, network devices, Bluetooth devices, etc.) so that a user can interact with the electronic device 400 via these external devices 100, and / or the electronic device 400 can communicate with one or more other data processing devices (e.g., routers, modems, etc.). Such communication may be performed through an input / output (I / O) interface 450, and may also be performed through a network adapter 460 with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet). The network adapter 460 may communicate with other modules of the electronic device 400 via the bus 430. It should be understood that although Figure 4 Not shown, other hardware and / or software modules may be used in the electronic device 400, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID electronic devices, tape drives, and data backup storage electronic devices, etc.

[0104] Figure 5 Schematic diagram of a computer readable medium embodiment of the present invention. Figure 5 As shown, the computer program can be stored on one or more computer-readable media. The computer-readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electronic device, device or device of electricity, magnetism, light, electromagnetic, infrared, or semiconductor, or any combination thereof. More specific examples (non-exhaustive list) of readable storage media include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. When the computer program is executed by one or more data processing devices, the computer-readable medium is enabled to implement the above method of the present invention, namely: configuring an auxiliary task model associated with a risk identification model; performing domain adaptation learning on the source domain and the target domain based on the auxiliary task model; selecting training samples corresponding to the source domain and the target domain according to the target domain cold start scenario; inputting the selected training samples into the risk identification model for training; and using the trained risk identification model to identify the equipment risk of the cold start scenario.

[0105] Through the description of the above implementation modes, it is easy for those skilled in the art to understand that the exemplary embodiments described in the present invention can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation mode of the present invention can be embodied in the form of a software product, which can be stored in a computer-readable storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, and includes several instructions to enable a data processing device (which can be a personal computer, a server, or a network device, etc.) to perform the above method according to the present invention.

[0106] The computer readable storage medium may include a data signal propagated in a baseband or as part of a carrier wave, wherein a readable program code is carried. This propagated data signal may take a variety of forms, including but not limited to an electromagnetic signal, an optical signal, or any suitable combination of the above. The readable storage medium may also be any readable medium other than a readable storage medium, which may send, propagate, or transmit a program for use by or in combination with an instruction-executing electronic device, device, or device. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, etc., or any suitable combination of the above.

[0107] Program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java, C++, etc., as well as conventional procedural programming languages ​​such as "C" language or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user device, as a separate software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. In the case of a remote computing device, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0108] In summary, the present invention can be implemented by a method, apparatus, electronic device or computer-readable medium that executes a computer program. In practice, a general data processing device such as a microprocessor or a digital signal processor (DSP) can be used to implement some or all functions of the present invention.

[0109] The specific embodiments described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the present invention is not inherently related to any specific computer, virtual device or electronic device, and various general devices can also implement the present invention. The above description is only a specific embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for identifying risky devices based on cold start scenario domain adaptation learning, characterized in that: The method comprises: Taking a task of identifying a first risk type in a business scenario in which the risk identification model is applied as an auxiliary task, and configuring an auxiliary task model associated with the risk identification model; wherein the first risk type is a risk type different from a risk type identified by the risk identification model; The source domain samples and target domain samples of the auxiliary task are first input into the shared layer for processing, and the auxiliary task model learns the feature transformation function T by optimizing the edge probability distance between the source domain and the target domain or the local edge probability distance between the source domain and the target domain to complete the domain adaptation learning of the source domain and the target domain; or, the source domain samples and target domain samples of the auxiliary task are first input into the shared layer for processing, and then the classification error between the source domain and the target domain is maximized through the gradient reversal layer between the source domain and the target domain, while identifying the correct category, eliminating the recognition of the domain, thereby completing the domain adaptation learning of the source domain and the target domain; the domain adaptation learning is: ； Among them, l is the loss function, n is the number of samples, is the sample weight, λ is the complexity weight, R is the domain adaptation learning regularization term, T is the feature transformation function, x i is the characteristic data of the i-th sample, y i is the label data of the i-th sample, Represents the source domain and represents the target domain; Select training samples corresponding to the source domain and the target domain according to the target domain cold start scenario; Input the selected training samples into the risk identification model for training; Use the trained risk identification model to identify equipment risks in cold start scenarios; The cold start refers to a situation where the number of labeled samples is less than the preset number of samples due to insufficient sample performance period.

2. The method according to claim 1, characterized in that The cold start scenarios include: a newly opened business scenario with a sample size less than a threshold, a restarted business scenario containing only old samples older than a predetermined period, and a newly opened business scenario in other regions.

3. The method according to claim 2, characterized in that: The selecting of training samples corresponding to the source domain and the target domain according to the target domain cold start scenario includes: For new business scenarios where the sample size is less than the threshold and new business scenarios in other regions, the source domain samples are used as training samples; For the restart business scenario that only contains old samples older than a predetermined period, the source domain and target domain samples are mixed as training samples.

4. A risk equipment identification device based on cold start scenario domain adaptive learning, characterized in that: The device comprises: A configuration module, configured to use a task of identifying a first risk type in a business scenario in which the risk identification model is applied as an auxiliary task, and configure an auxiliary task model associated with the risk identification model; wherein the first risk type is a risk type different from a risk type identified by the risk identification model; The adaptation module is used to input the source domain samples and the target domain samples of the auxiliary task into the shared layer for processing first, and the auxiliary task model learns the feature transformation function T by optimizing the edge probability distance between the source domain and the target domain or the local edge probability distance between the source domain and the target domain to complete the domain adaptation learning of the source domain and the target domain; or, after the source domain samples and the target domain samples of the auxiliary task are input into the shared layer for processing first, the classification error between the source domain and the target domain is maximized through the gradient reversal layer between the source domain and the target domain, and the recognition of the domain is eliminated while the correct category is identified, thereby completing the domain adaptation learning of the source domain and the target domain; the domain adaptation learning is: ； Among them, l is the loss function, n is the number of samples, is the sample weight, λ is the complexity weight, R is the domain adaptation learning regularization term, T is the feature transformation function, x i is the characteristic data of the i-th sample, y i is the label data of the i-th sample, Represents the source domain and represents the target domain; A selection module is used to select training samples corresponding to the source domain and the target domain according to the target domain cold start scenario; A training module, used to input the selected training samples into the risk identification model for training; An identification module is used to identify equipment risks in cold start scenarios using a trained risk identification model; The cold start refers to a situation where the number of labeled samples is less than the preset number of samples due to insufficient sample performance period.

5. The device according to claim 4, characterized in that The cold start scenarios include: a newly opened business scenario with a sample size less than a threshold, a restarted business scenario containing only old samples older than a predetermined period, and a newly opened business scenario in other regions.

6. The device according to claim 5, characterized in that: The selection module uses source domain samples as training samples for new business scenarios with a sample size less than a threshold and new business scenarios in other regions; For the restart business scenario that only contains old samples older than a predetermined period, the source domain and target domain samples are mixed as training samples.

7. An electronic device comprising: processor; as well as A memory storing computer executable instructions which, when executed, cause the processor to perform the method according to any one of claims 1-3.

8. A computer-readable storage medium, wherein: The computer-readable storage medium stores one or more programs, and when the one or more programs are executed by a processor, the method according to any one of claims 1 to 3 is implemented.

Images