WO2021114974A1 - 用户风险评估方法及装置、电子设备、存储介质 - Google Patents

用户风险评估方法及装置、电子设备、存储介质 Download PDF

Info

Publication number
WO2021114974A1
WO2021114974A1 PCT/CN2020/126961 CN2020126961W WO2021114974A1 WO 2021114974 A1 WO2021114974 A1 WO 2021114974A1 CN 2020126961 W CN2020126961 W CN 2020126961W WO 2021114974 A1 WO2021114974 A1 WO 2021114974A1
Authority
WO
WIPO (PCT)
Prior art keywords
sample data
target
target sample
teacher
label value
Prior art date
Application number
PCT/CN2020/126961
Other languages
English (en)
French (fr)
Inventor
陈岑
Original Assignee
支付宝(杭州)信息技术有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 支付宝(杭州)信息技术有限公司 filed Critical 支付宝(杭州)信息技术有限公司
Publication of WO2021114974A1 publication Critical patent/WO2021114974A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • 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
    • 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
    • G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N20/00—Machine learning

Definitions

  • One or more embodiments of this specification relate to the field of artificial intelligence technology, and in particular to a user risk assessment method and device, electronic equipment, and storage medium.
  • Risk control means that risk managers take various measures and methods to eliminate or reduce the various possibilities of risk events, or risk controllers to reduce the losses caused when risk events occur. By accurately identifying potential risks for users, companies can improve the security protection capabilities of themselves and their partners, and contribute to business growth.
  • one or more embodiments of this specification provide a user risk assessment method and device, electronic equipment, and storage medium.
  • a user risk assessment method which includes: inputting behavior information of users of a target partner into a student risk control model corresponding to the target partner;
  • the student risk control model is obtained by knowledge distillation of the target sample data based on the soft label value of the target sample data of the target partner and the risk label value originally marked as the hard label value of the target sample data.
  • the soft label value is obtained by integrating the prediction results of a plurality of teacher risk control models on the target sample data, and each teacher risk control model is obtained by training the corresponding sample data of other partners; among them, any sample data Contains behavior information marked with a risk label value; and determines the risk score of the user according to the output result of the student risk control model.
  • a knowledge transfer method based on a machine learning model which includes: obtaining the prediction results of multiple teacher networks for target sample data from the target field, and each teacher network Obtained by training the sample data of the respective source fields; integrating the obtained multiple prediction results to obtain the soft label value corresponding to the target sample data; based on the soft label value and the original target sample data
  • the marked hard label value is used to perform knowledge distillation on the target sample data to obtain a student network in the target field.
  • a knowledge transfer method based on a machine learning model which includes: inputting the received target sample data from the target field into a teacher network, and the teacher network passes It is obtained by training the sample data of the source field to which it belongs; and the prediction result output by the teacher network is returned to the provider of the target sample data, so that the provider can compare the prediction result and other teacher networks to the target.
  • the prediction results of the sample data are integrated to obtain the soft label value corresponding to the target sample data, and the target sample data is knowledge-distilled based on the soft label value and the hard label value of the target sample data originally marked Get a network of students in the target field.
  • a user risk assessment device which includes: an information input unit that inputs behavior information of a user of a target partner into student risk control corresponding to the target partner Model; the student risk control model is based on the target sample data soft label value of the target sample data and the target sample data originally marked as the hard label value of the risk label value, the target sample data Knowledge distillation is obtained, the soft label value is obtained by integrating the prediction results of multiple teacher risk control models for the target sample data, and each teacher risk control model is obtained by training the corresponding sample data of other partners; where Any sample data includes behavior information marked with a risk label value; the risk assessment unit determines the risk score of the user according to the output result of the student risk control model.
  • a knowledge transfer device based on a machine learning model
  • a prediction result obtaining unit which obtains predictions of multiple teacher networks for target sample data from a target field
  • each teacher network is obtained by training the sample data of their respective source fields
  • the integrated learning unit integrates the obtained multiple prediction results to obtain the soft label value corresponding to the target sample data
  • the student network training unit Based on the soft label value and the hard label value originally marked on the target sample data, knowledge distillation is performed on the target sample data to obtain a student network in the target field.
  • a knowledge transfer device based on a machine learning model, including: a sample data input unit, which inputs the received target sample data from the target domain into the teacher network, The teacher network is obtained by training the sample data of the source field to which it belongs; the prediction result returning unit returns the prediction result output by the teacher network to the provider of the target sample data, so that the provider can
  • the prediction result is integrated with the prediction result of the target sample data by other teacher networks to obtain the soft label value corresponding to the target sample data, and the hard label value originally marked based on the soft label value and the target sample data Perform knowledge distillation on the target sample data to obtain a student network in the target field.
  • an electronic device including: a processor; a memory for storing executable instructions of the processor; wherein the processor runs the executable instructions In order to realize the user risk assessment method as described in the above first aspect.
  • an electronic device including: a processor; a memory for storing executable instructions of the processor; wherein the processor runs the executable instructions In order to realize the knowledge transfer method based on the machine learning model as described in the above second aspect.
  • an electronic device including: a processor; a memory for storing executable instructions of the processor; wherein the processor runs the executable instructions In order to realize the knowledge transfer method based on the machine learning model as described in the above third aspect.
  • a computer-readable storage medium having computer instructions stored thereon, which, when executed by a processor, implement the steps of the user risk assessment method described in the first aspect.
  • a computer-readable storage medium having computer instructions stored thereon, and when the instructions are executed by a processor, the machine learning model-based knowledge as described in the second aspect is realized. Steps of the migration method.
  • a computer-readable storage medium having computer instructions stored thereon, and when the instructions are executed by a processor, the knowledge based on the machine learning model as described in the third aspect is realized. Steps of the migration method.
  • Fig. 1 is a schematic structural diagram of a knowledge transfer system based on a machine learning model provided by an exemplary embodiment.
  • Fig. 2 is a flowchart of a method for knowledge transfer based on a machine learning model provided by an exemplary embodiment.
  • Fig. 3 is a flowchart of another method for knowledge transfer based on a machine learning model provided by an exemplary embodiment.
  • Fig. 4 is a flowchart of a user risk assessment method provided by an exemplary embodiment.
  • Fig. 5 is an interaction diagram of a knowledge transfer method based on a machine learning model provided by an exemplary embodiment.
  • Fig. 6 is a schematic structural diagram of a device provided by an exemplary embodiment.
  • Fig. 7 is a block diagram of a user risk assessment device provided by an exemplary embodiment.
  • Fig. 8 is a schematic structural diagram of another device provided by an exemplary embodiment.
  • Fig. 9 is a block diagram of a device for knowledge transfer based on a machine learning model provided by an exemplary embodiment.
  • Fig. 10 is a schematic structural diagram of another device provided by an exemplary embodiment.
  • Fig. 11 is a block diagram of another apparatus for knowledge transfer based on a machine learning model provided by an exemplary embodiment.
  • the steps of the corresponding method may not be executed in the order shown and described in this specification.
  • the method may include more or fewer steps than described in this specification.
  • a single step described in this specification may be decomposed into multiple steps for description in other embodiments; and multiple steps described in this specification may also be combined into a single step in other embodiments. description.
  • Fig. 1 is a schematic structural diagram of a knowledge transfer system based on a machine learning model provided by an exemplary embodiment.
  • the system may include a server 11, a network 12, and several electronic devices, such as a mobile phone 13, a mobile phone 14, and a PC15-16.
  • the server 11 may be a physical server including an independent host, or the server 11 may be a virtual server carried by a host cluster. During operation, the server 11 is used as a server to interface with each partner, that is, to provide a platform for cooperation with each partner, for migrating the performance of the teacher network trained by each partner to the student network.
  • Mobile phones 13-14 and PC15-16 are just one type of electronic equipment that users can use.
  • the partners that interface with the server 11 can obviously also use electronic devices such as the following types: tablet devices, notebook computers, PDAs (Personal Digital Assistants), wearable devices (such as smart glasses, smart watches, etc.) Etc., one or more embodiments of this specification do not limit this.
  • each partner uses its own sample data to train to obtain a teacher network, which can guide the training of related student networks, and take the model parameters learned by the teacher network (also can be understood as The knowledge learned by the teacher network) is shared with the student network to improve the performance of the student network.
  • the network 12 for interaction between the mobile phone 13-14, the PC 15-16 and the server 11, it may include multiple types of wired or wireless networks.
  • the network 12 may include a Public Switched Telephone Network (PSTN) and the Internet.
  • PSTN Public Switched Telephone Network
  • Fig. 2 is a flowchart of a method for knowledge transfer based on a machine learning model provided by an exemplary embodiment. As shown in Figure 2, the method is applied to the server and may include steps 202-206.
  • Step 202 Obtain the prediction results of multiple teacher networks with respect to the target sample data from the target domain, and each teacher network is obtained by training the sample data of their respective source domains.
  • sample data labeled with label values when training a supervised machine learning model, it may be difficult to collect sample data labeled with label values. For example, sample data is less accumulated due to time issues, and the amount of data collected for sample data is relatively large, which is time-consuming. ,higher cost. Furthermore, even when the sample data is sufficient, the cost of building a model from scratch is higher and the efficiency is lower. therefore.
  • transfer learning Transfer Learning
  • transfer Learning technology can be used to learn from the trained model that is related to the field (for example, of the same type, high similarity, etc.) The acquired knowledge is transferred to the machine learning model in the field, thereby improving the efficiency of training the model.
  • the domain of the existing knowledge is called the source domain
  • the domain of the new knowledge to be learned is called the target domain.
  • the source domain usually has a large amount of label data, while the target domain often There are only a small number of label samples, and the source field and the target field are different but related to a certain extent. Knowledge transfer can be carried out by reducing the distribution difference between the source field and the target field.
  • a teacher-student network is used to guide the training of the student network by distilling the knowledge of the teacher network.
  • the teacher network is often a more complex network with very good performance and generalization ability.
  • the teacher network can be used as a soft target to guide another simpler student network to learn, making it simpler and more computationally expensive.
  • a few student models can also have performance similar to that of a teacher network.
  • the teacher network corresponds to the source domain, that is, the supervised learning model that has been trained in the source domain is used as the teacher network to guide students' network learning and learn from themselves
  • the knowledge of is transferred to the student network
  • the student network corresponds to the target field, that is, the model to be trained in the target field is used as the student network.
  • the server when a partner docking with the server has a model to be trained, the server can perform migration learning on the supervised machine learning models that have been trained by other partners related to the partner's field. To guide the learning of the model to be trained. Then, in the process of training the student network in the target field, there is no need to recollect a large amount of sample data in the target field for training, so that the efficiency of training the student network can be improved. At the same time, the student network can also inherit the better generalization ability and performance of the teacher network.
  • one or more teacher networks can be selected to guide the training of student networks.
  • a field with higher similarity to the target field can be selected as the source field.
  • the knowledge transfer scheme based on the machine learning model of this specification can be understood as the data providers of various source fields work together to complete the training of the student network, namely Multiple data providers have their own sample data and can use each other's data to train machine learning models in a unified manner. It should be noted that the sample data of each data provider belongs to its own private data, so the above-mentioned multi-party joint modeling (joint modelling) process should be carried out while ensuring the security of the data of all parties. Therefore, the data provider, as the executive body of training the teacher network, uses its own labeled sample data to train the teacher network in their respective source fields.
  • each teacher network uses its own private data as sample data for training through data providers in their respective source fields. It can be seen that, on the one hand, each data provider cooperates to train their own teacher network, which can improve the efficiency of subsequent training of the student network; on the other hand, the training process of the teacher network in each source field does not need to be out of the domain, which can ensure The privacy of the sample data in the field.
  • Step 204 Integrate the obtained multiple prediction results to obtain the soft label value corresponding to the target sample data.
  • the trained student network in order to improve the trained student network to be a diverse (comprehensive) strong supervision model, so that the student network is stable and performs well in all aspects, instead of preference (weak supervision model, in some It performs better in terms of performance), and can perform integrated learning on the obtained prediction results of multiple teacher networks.
  • the integrated learning of the obtained multiple prediction results when a certain teacher network has an error prediction for the target sample data, the error prediction can be corrected by other teacher networks, thereby reducing bagging and bias (boosting) and improving the effect of prediction (stacking).
  • the specific implementation manner of the integrated learning can be flexibly selected according to the actual situation, and one or more embodiments of this specification do not limit this. For example, voting, weighted average, etc. can be adopted.
  • algorithms such as Bagging (bootstrap aggregating, bagging; such as random forest), Boosting, and Stacking can be used.
  • Step 206 Perform knowledge distillation on the target sample data based on the soft label value and the original hard label value of the target sample data to obtain a student network in the target field.
  • the hard label value is the label value originally marked in the target sample data.
  • the hard label value is obtained by annotating the target sample data by the provider (belonging to the target field) of the target sample data.
  • After obtaining the soft target value corresponding to the target sample data through integrated learning based on the soft label value and the hard target value originally marked on the target sample data, perform knowledge distillation on the target sample data to obtain the target Network of students in the field.
  • the hard target originally labeled from the target sample data contains a lower amount of information (information entropy); while the soft target comes from the prediction output of the large model (teacher network), which has higher entropy. Can provide more information than hard target.
  • the soft target is used to assist the hard target to train together, that is, less data and a larger learning rate are used, so that a simpler student model with fewer parameter calculations can also have performance similar to that of a teacher network (and therefore also Can be understood as a way of model compression).
  • the training of the student network contains two objective functions: one corresponds to the hard target, that is, the original objective function, which is the cross-entropy of the class probability output of the student network and the true value of the label; the other corresponds to the soft target, which is The cross entropy of the category probability output of the student network and the category probability output of the teacher network.
  • the soft target add the temperature parameter T to the softmax function:
  • q i is the probability value of the i-th class
  • the input z i is the prediction vector (logarithmic logits) of the i-th class
  • logits is the original (non-standardized) generated by the classification model, and the prediction vector is usually passed to the normalization function.
  • logits are usually used as the input of the softmax function to generate a (normalized) probability vector from the softmax function, corresponding to each possible category.
  • the softmax function calculates the logit z i of each category as a probability q i by comparing the input z i with other logits.
  • the objective function corresponding to the hard target and the objective function corresponding to the soft target can be used as the final objective function of the student network through a weighted average. For example, it can be set to have a larger weight for soft target.
  • the value of T can take an intermediate value, and the weight assigned by the soft target is T ⁇ 2, and the weight of the hard target is 1.
  • other arbitrary weights can also be set, and one or more embodiments of this specification do not limit this.
  • a student network with strong interpretability can be obtained.
  • a classifier with strong interpretability can be used for training.
  • FIG. 3 is a flowchart of another method for knowledge transfer based on a machine learning model provided by an exemplary embodiment. As shown in Fig. 3, the method is applied to the provider of the teacher network and may include steps 302-304.
  • Step 302 Input the received target sample data from the target domain into a teacher network, which is obtained by training the sample data of the source domain to which it belongs.
  • the provider of the teacher network may be a sample data provider of the training teacher network.
  • the sample data provider as the executive body of training the teacher network, uses its own labeled sample data to train the teacher network in their respective source fields. It can be seen that the training process of the teacher network in each source field does not need to be out-of-domain, and the privacy of the sample data in each source field can be guaranteed.
  • the differential privacy technology can be used to protect the privacy of decision-making (that is, to ensure the privacy of the output results of each teacher network). Therefore, differential privacy processing can be performed on the prediction results output by the teacher network, and then the prediction results subjected to differential privacy processing can be returned to the provider of the target sample data.
  • Laplacian Noise can be introduced for the prediction result, and the prediction result output by the teacher network can be processed for differential privacy through the following formula:
  • f(i) represents the probability prediction value of the i-th sample data
  • Lap(1/ ⁇ ) represents the Laplacian probability distribution centered at 0 and scaled by 1/ ⁇
  • ⁇ represents the degree of privacy protection Parameters.
  • differential privacy can be flexibly selected according to actual conditions, and one or more embodiments of this specification do not limit this.
  • Laplace mechanism Laplace distribution, exponential mechanism, etc.
  • Step 304 Return the prediction result output by the teacher network to the provider of the target sample data, so that the provider integrates the prediction result and the prediction result of other teacher networks for the target sample data to obtain a corresponding Performing knowledge distillation on the target sample data based on the soft label value of the target sample data and the hard label value originally marked on the target sample data to obtain a student network in the target field.
  • the specific content of the sample data can be flexibly set according to actual application scenarios.
  • the data type of the sample data can include image, text, voice, and so on.
  • the labeling of sample data can also be flexibly set according to actual application scenarios, as described below with examples.
  • the potential risks of users or merchants can be predicted, such as the risks of predicting loans and real-time transactions.
  • the cooperation platform has docked and cooperated with merchants, and each merchant has accumulated a large amount of sample data during the business process.
  • the sample data in text form, or other data types
  • the sample data includes the user's basic information, behavior information, transaction information, and so on.
  • merchants can label sample data in the transaction risk dimension.
  • the newly accessed merchant a can cooperate with other merchants of the same type on the cooperation platform to perform joint modeling.
  • the newly-accessed merchant a belongs to the target field, a small amount of sample data it owns is the target sample data, and the risk control model to be trained is the student network; the other merchants on the cooperation platform are in the same industry as the newly-accessed merchant (For example, the merchants 1-n belonging to the same fund, insurance company, etc.) belong to the source field, and the merchants 1-n can use the large amount of sample data they have accumulated to train the teacher network to guide the training of the student network.
  • the merchant a can input the acquired user's basic information, behavior information, transaction information and other data into the student network, thereby predicting the risk score of the current transaction with the user.
  • the potential needs of users can be predicted, such as predicting the products the user wants to buy, news of interest, books that they like to read, and so on.
  • the cooperation platform has docked and cooperated with multiple sellers, and each seller has accumulated a large number of user purchase records in the course of business.
  • the sample data (in text form, or other data types) is user information such as occupation, income, age, gender, etc.
  • the merchant can mark the sample data according to the products purchased by the user in the user purchase record.
  • the newly connected seller a can cooperate with other sellers of the same type on the cooperation platform to perform joint modeling.
  • the newly accessed merchant a belongs to the target field, a small number of user purchase records in its own hands are used as the target sample data, and the product recommendation model to be trained is the student network; other sellers on the cooperation platform are the same as the newly accessed seller Sellers 1-n in the industry (such as catering, clothing, etc.) belong to the source field, and sellers 1-n can use their accumulated large number of user purchase records to train to obtain a teacher network to guide the training of the student network.
  • seller a can enter the user information of the acquired user into the student network, thereby predicting that the user may have a purchase demand product, and then recommending the corresponding product to the user based on the prediction result commodity.
  • the cooperation platform cooperates with many companies, and each company has accumulated a large amount of dialogue data in the process of providing customer service to users.
  • the sample data can be text, image, user's voice, etc. input by the user, and the annotation for the sample data is the content of the customer service's reply to the user in the conversation data.
  • another company a newly accesses the cooperation platform and hopes to provide users with intelligent customer service, if the conversation data between the user and the customer service is limited, it can work with other companies in the cooperation platform to conduct joint modeling.
  • companies 1-n that provide customer service services such as voice assistants, chat tools, and answering questions can conduct joint modeling through their own accumulated conversation data.
  • customer service services such as voice assistants, chat tools, and answering questions
  • the newly-connected company a belongs to the target field, the small amount of dialogue data it owns is the target sample data, and the customer service model to be trained is the student network; the company 1-n belongs to the source field, and the company 1-n can use each
  • the accumulated large amount of dialogue data is trained by the teacher network to guide the training of the student network.
  • enterprise a (or enterprise 1-n) can use the student network to provide users with intelligent customer service, that is, the conversation content (text, image, voice, etc.) initiated by the user as The input of the student network, and the output result as a reply to this conversation.
  • intelligent customer service that is, the conversation content (text, image, voice, etc.) initiated by the user as The input of the student network, and the output result as a reply to this conversation.
  • FIG. 4 is a flowchart of a user risk assessment method provided by an exemplary embodiment.
  • the evaluation method may include steps 402-404.
  • Step 402 Input the behavior information of the user of the target partner into the student risk control model corresponding to the target partner; the student risk control model uses the soft label value based on the target sample data of the target partner and the
  • the target sample data is originally labeled as the hard label value of the risk label value, which is obtained by knowledge distillation of the target sample data, and the soft label value is obtained by predicting the result of the target sample data by a plurality of teacher risk control models It is obtained through integration, and each teacher's risk control model is obtained by training the corresponding sample data of other partners; among them, any sample data contains behavior information marked with a risk label value.
  • Step 404 Determine the risk score of the user according to the output result of the student risk control model.
  • the student risk control model corresponds to the student network in the above embodiment in Figure 2-3
  • the teacher risk control model corresponds to the teacher network in the above embodiment in Figure 2-3.
  • the specific content of the sample data for training each model is the user's behavior information, and the marked content is the user's risk score; in other words, the input of each model is the user's behavior information, and the output is the user's risk score (including probability distribution).
  • Multiple parties cooperate on the same platform.
  • the target partner belongs to the target field and is the provider of the target sample data.
  • the model to be trained is the student risk control model.
  • the teacher risk control model of other partners can be used to guide the training of the student risk control model.
  • the specific process of training can refer to the embodiment shown in Figs. 2-3, which will not be repeated here.
  • the student risk control model can be configured on the client side of the target partner. Then, after the target partner obtains the user's behavior information, The client can input behavior information into the student's risk control model to determine the user's risk score based on the output result, and then determine the subsequent processing method for the user. For example, when the risk score is low (indicating that the user is safer), consumer rights can be issued to the user; when the risk score is high (indicating that the user has potential risks), the user's registration request can be intercepted.
  • the student risk control model can be configured on the server side that is docked with the target partner. After obtaining the user's behavior information, the target partner can send the behavior information to the server through the client. The server uses the student risk control model to determine the user's risk score and returns to the client for display.
  • the target partner in order to improve the generalization ability and performance of the student's risk control model (that is, the generalization ability and performance of the teacher's risk control model can be better transferred to the student's risk control model), the target partner can be selected similar to the target partner Teacher risk control models of other partners with higher degrees to guide students' risk control model training.
  • it can be set that the target partner and the other partner belong to the same type of partner. For example, all belong to the catering category, and all belong to the financial category.
  • each teacher risk control model is obtained through training on its own sample data by the corresponding other partner.
  • other partners use their own labeled sample data to train the teacher's risk control model.
  • the interaction process may include the following steps:
  • step 502A the partner 1 obtains the teacher network 1 through training on the private data marked by itself.
  • step 502B the partner 2 obtains the teacher network 2 through the training of the private data marked by itself.
  • step 502C the partner n obtains the teacher network n through the private data training marked by itself.
  • steps 502A-502C are steps in parallel with each other, and there is no requirement on the time sequence.
  • “Merchant Health Score” is a risk assessment conducted by the server as a merchant cooperation platform to ISV (Independent Software Vendors) channel providers for merchants under the channel.
  • ISV Independent Software Vendors
  • Indicators through the evaluation of the "merchant health score” of the merchants under the channel, can help partners (ISV channel providers) to improve their risk control capabilities.
  • ISV channel providers modeling the models used to evaluate merchants’ health scores, due to limited merchant behavior data (ie limited sample data), merchant cooperation platforms can be used to obtain information from other partners (other ISV channel providers).
  • the accumulated business behavior data is jointly modeled.
  • the other partners of the joint modeling should have a certain relationship with the ISV channel provider, for example, belong to the same industry.
  • the following takes the ISV channel provider and partner 1-n joint modeling as an example for illustration.
  • the partner 1-n labels the behavior information of the merchants in the historical business process in the risk dimension, and then obtains the sample data (private data belonging to itself) used to train the teacher network, that is, the trained teacher network
  • the input of is the behavior information of the merchant
  • the output is the corresponding risk score.
  • the supervised machine learning algorithm used for training can be flexibly selected according to actual conditions, and one or more embodiments of this specification do not limit this.
  • the following takes the classifier as an example for description.
  • step 504A the cooperation platform sends target sample data to the partner 1.
  • step 504B the cooperation platform sends the target sample data to the partner 2.
  • step 504C the cooperation platform sends target sample data to the partner n.
  • the ISV channel provider may send the target sample data (ie, the merchant behavior information it owns) to the cooperation platform, so that the cooperation platform can perform joint modeling with the partner 1-n based on the target sample data.
  • the ISV channel provider can directly perform joint modeling with the partner 1-n, that is, the steps performed by the cooperation platform in this embodiment are directly performed by the ISV channel provider.
  • the cooperation platform can also send the target sample data to the partner 1, and then the partner 1 forwards the target sample data to the partner 2-n respectively.
  • step 506A the partner 1 inputs the target sample data into the teacher network 1 to obtain the prediction result 1.
  • step 506B the partner 2 inputs the target sample data into the teacher network 2 to obtain the prediction result 2.
  • step 506C the partner n inputs the target sample data into the teacher network n to obtain the prediction result n.
  • each classifier fk (teacher network) can predict A probability distribution fk(xi) is obtained, then each fk(xi) can be integrated through integrated learning technology to obtain the final score.
  • step 508A the partner 1 performs differential privacy processing on the prediction result 1.
  • step 508B the partner 2 performs differential privacy processing on the prediction result 2.
  • step 508C the partner n performs differential privacy processing on the prediction result n.
  • each partner can use differential privacy technology to protect the privacy of decision-making (that is, to ensure the privacy of the output results of each teacher's network) for the obtained prediction results. Therefore, differential privacy processing can be performed on the prediction results output by the teacher network.
  • Laplacian Noise can be introduced on the probability prediction value (ie prediction result) of each classifier, namely fk(xi)+Lap(1/ ⁇ ); where, Lap(1/ ⁇ ) Represents the Laplacian probability distribution centered at 0 and scaled by 1/ ⁇ , and ⁇ represents the parameter used to control the degree of privacy protection.
  • the specific implementation mechanism of differential privacy can be flexibly selected according to actual conditions, and one or more embodiments of this specification do not limit this. For example, Laplace mechanism, Laplace distribution, exponential mechanism, etc.
  • step 510A the partner 1 returns the prediction result 1 after differential privacy processing to the cooperation platform.
  • step 510B the partner 2 returns the prediction result 2 after the differential privacy processing to the cooperation platform.
  • step 510C the partner n returns the prediction result n after differential privacy processing to the cooperation platform.
  • this specification does not set the time sequence requirements between steps 506A-506C, steps 508A-508C, and steps 510A-510C.
  • step 512 the cooperation platform integrates the prediction results 1-n to obtain the soft label value.
  • the obtained prediction results 1-n can be integrated learning to obtain the soft label value corresponding to the target sample data.
  • the result of ensemble learning is used as the soft label value corresponding to the target sample data.
  • the specific implementation manner of the integrated learning can be flexibly selected according to the actual situation, and one or more embodiments of this specification do not limit this. For example, voting, averaging, etc. can be adopted. For another example, algorithms such as Bagging (bootstrap aggregating, bagging; such as random forest), Boosting, and Stacking can be used.
  • Bagging bootsstrap aggregating, bagging; such as random forest
  • Boosting Boosting
  • Stacking can be used.
  • step 514 the cooperation platform performs knowledge distillation on the target sample data to obtain a student network based on the soft label value and the original hard label value of the target sample data.
  • the probability distribution output of all classifiers after differential privacy processing is averaged, and the final probability output obtained by averaging is used as a soft target to guide students' network learning.
  • the student network can be configured on the client side of the ISV channel provider.
  • the ISV channel provider can input the behavior information to the student network through the client to determine according to the output result
  • the risk score of the merchant determines the subsequent processing method for the merchant. For example, when the risk score is low (indicating that the merchant is safer), consumer rights can be issued to the merchant; when the risk score is high (indicating that the merchant has potential risks), the merchant's registration request can be intercepted.
  • the student network can be configured on the cooperation platform, then the ISV channel provider can send the behavior information to the cooperation platform through the client after obtaining the behavior information of the merchant, so that the cooperation platform can use the student network to determine The risk score of the merchant is returned to the client for display.
  • this specification also provides device embodiments.
  • the embodiments of the user risk assessment device in this specification can be applied to electronic equipment.
  • the device embodiments can be implemented by software, or can be implemented by hardware or a combination of software and hardware.
  • Taking software implementation as an example as a logical device, it is formed by reading the corresponding computer program instructions in the non-volatile memory into the memory through the processor of the electronic device where it is located.
  • FIG. 6 is a schematic structural diagram of a device provided by an exemplary embodiment. Please refer to FIG. 6.
  • the device includes a processor 602, an internal bus 604, a network interface 606, a memory 608, and a non-volatile memory 610.
  • the processor 602 reads the corresponding computer program from the non-volatile memory 610 to the memory 608 and then runs it to form a user risk assessment device on a logical level.
  • one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, and so on. That is to say, the execution subject of the following processing flow is not limited to each
  • the logic unit can also be a hardware or a logic device.
  • the user risk assessment device may include: an information input unit 71, which inputs the behavior information of the user of the target partner into the student risk control model corresponding to the target partner; the student The risk control model is obtained by performing knowledge distillation on the target sample data based on the soft label value of the target sample data of the target partner and the risk label value originally marked as the hard label value of the target sample data.
  • the soft label value is obtained by integrating the prediction results of multiple teacher risk control models with respect to the target sample data, and each teacher risk control model is obtained by training the corresponding sample data of other partners; among them, any sample data Contains behavior information marked with a risk label value; the risk assessment unit 72 determines the risk score of the user according to the output result of the student risk control model.
  • each teacher's risk control model is obtained through training on its own sample data by corresponding other partners.
  • the embodiment of the knowledge transfer device based on the machine learning model of this specification can be applied to electronic equipment.
  • the device embodiments can be implemented by software, or can be implemented by hardware or a combination of software and hardware.
  • Taking software implementation as an example as a logical device, it is formed by reading the corresponding computer program instructions in the non-volatile memory into the memory through the processor of the electronic device where it is located.
  • FIG. 8 is a schematic structural diagram of a device provided by an exemplary embodiment.
  • the device includes a processor 802, an internal bus 804, a network interface 806, a memory 808, and a non-volatile memory 810.
  • the processor 802 reads the corresponding computer program from the non-volatile memory 810 to the memory 808 and then runs it to form a knowledge transfer device based on a machine learning model on a logical level.
  • one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, and so on. That is to say, the execution subject of the following processing flow is not limited to each
  • the logic unit can also be a hardware or a logic device.
  • the knowledge transfer device based on the machine learning model may include: a prediction result obtaining unit 91, which obtains prediction results of multiple teacher networks for target sample data from the target field, and each teacher network It is obtained by training the sample data of the respective source fields; the integrated learning unit 92 integrates the obtained multiple prediction results to obtain the soft label value corresponding to the target sample data; the student network training unit 93 is based on the The soft label value and the hard label value originally marked on the target sample data are subjected to knowledge distillation on the target sample data to obtain a student network in the target field.
  • each source domain and the target domain are of the same type.
  • each teacher network is obtained by training its own private data as sample data through data providers in their respective source fields.
  • the data type of the target sample data and/or the sample data of each source field includes at least one of the following: image, text, and voice.
  • the embodiment of the knowledge transfer device based on the machine learning model of this specification can be applied to electronic equipment.
  • the device embodiments can be implemented by software, or can be implemented by hardware or a combination of software and hardware.
  • Taking software implementation as an example as a logical device, it is formed by reading the corresponding computer program instructions in the non-volatile memory into the memory through the processor of the electronic device where it is located.
  • FIG. 10 is a schematic structural diagram of a device provided by an exemplary embodiment. Please refer to FIG. 10.
  • the device includes a processor 1002, an internal bus 1004, a network interface 1006, a memory 1008, and a non-volatile memory 1010.
  • the processor 1002 reads the corresponding computer program from the non-volatile memory 1010 to the memory 1008 and then runs it to form a knowledge transfer device based on the machine learning model at the logical level.
  • one or more embodiments of this specification do not exclude other implementations, such as logic devices or a combination of software and hardware, and so on. That is to say, the execution subject of the following processing flow is not limited to each
  • the logic unit can also be a hardware or a logic device.
  • the machine learning model-based knowledge transfer device may include: a sample data input unit 1101, which inputs the received target sample data from the target field into the teacher network, and the teacher network passes It is obtained by training the sample data of the source field to which it belongs; the prediction result returning unit 1102 returns the prediction result output by the teacher network to the provider of the target sample data, so that the provider can compare the prediction result and other information to the provider of the target sample data.
  • the teacher network integrates the prediction results of the target sample data to obtain the soft label value corresponding to the target sample data, and compares the target sample data with the hard label value originally marked based on the soft label value and the target sample data.
  • the sample data is subjected to knowledge distillation to obtain a network of students in the target field.
  • the prediction result returning unit 1102 is specifically configured to: perform differential privacy processing on the prediction result output by the teacher network; and return the prediction result subjected to differential privacy processing to the provider.
  • differential privacy processing is performed on the prediction result output by the teacher network through the following formula:
  • f(i) represents the probability prediction value of the i-th sample data
  • Lap(1/ ⁇ ) represents the Laplacian probability distribution centered at 0 and scaled by 1/ ⁇
  • ⁇ represents the degree of privacy protection Parameters.
  • it further includes: a privacy obtaining unit 1103, which obtains private data of the source domain to which it belongs; and a teacher network training unit 1104, which trains the private data as sample data to obtain the teacher network.
  • a privacy obtaining unit 1103 which obtains private data of the source domain to which it belongs
  • a teacher network training unit 1104 which trains the private data as sample data to obtain the teacher network.
  • the data type of the target sample data and/or the sample data of each source field includes at least one of the following: image, text, and voice.
  • a typical implementation device is a computer.
  • the specific form of the computer can be a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email receiving and sending device, and a game control A console, a tablet computer, a wearable device, or a combination of any of these devices.
  • the computer includes one or more processors (CPU), input/output interfaces, network interfaces, and memory.
  • processors CPU
  • input/output interfaces network interfaces
  • memory volatile and non-volatile memory
  • the memory may include non-permanent memory in computer readable media, random access memory (RAM) and/or non-volatile memory, such as read-only memory (ROM) or flash memory (flash RAM). Memory is an example of computer readable media.
  • RAM random access memory
  • ROM read-only memory
  • flash RAM flash memory
  • Computer-readable media include permanent and non-permanent, removable and non-removable media, and information storage can be realized by any method or technology.
  • the information can be computer-readable instructions, data structures, program modules, or other data.
  • Examples of computer storage media 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, CD-ROM, digital versatile disc (DVD) or other optical storage, Magnetic cassettes, disk storage, quantum memory, graphene-based storage media or other magnetic storage devices, or any other non-transmission media, can be used to store information that can be accessed by computing devices.
  • computer-readable media does not include transitory media, such as modulated data signals and carrier waves.
  • first, second, third, etc. may be used to describe various information in one or more embodiments of this specification, the information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other.
  • first information may also be referred to as second information, and similarly, the second information may also be referred to as first information.
  • word “if” as used herein can be interpreted as "when” or “when” or "in response to determination”.

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Economics (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Software Systems (AREA)
  • Strategic Management (AREA)
  • Computing Systems (AREA)
  • Game Theory and Decision Science (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Development Economics (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Educational Administration (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Medical Informatics (AREA)
  • Probability & Statistics with Applications (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

本说明书提供一种用户风险评估方法，该方法包括：将目标合作方的用户的行为信息输入对应于该目标合作方的学生风控模型；该学生风控模型通过基于该目标合作方的目标样本数据的软标签值和该目标样本数据原本被标注的被作为硬标签值的风险标签值，对该目标样本数据进行知识蒸馏得到，该软标签值通过对多个教师风控模型针对该目标样本数据的预测结果进行集成得到，各个教师风控模型通过对相应的其他合作方的样本数据进行训练得到；其中，任一样本数据包含被标注有风险标签值的行为信息；根据该学生风控模型的输出结果确定该用户的风险评分。该方法可在保证各个合作方隐私的情况下，使得各个合作方协同训练学生风控模型，以用于进行风险评估。

Description

用户风险评估方法及装置、电子设备、存储介质 技术领域
本说明书一个或多个实施例涉及人工智能技术领域，尤其涉及一种用户风险评估方法及装置、电子设备、存储介质。
背景技术
风险控制是指风险管理者采取各种措施和方法，消灭或减少风险事件发生的各种可能性，或风险控制者减少风险事件发生时造成的损失。企业通过对用户潜在的风险进行精准识别，可以提升自身以及合作伙伴的安全防护能力，有助于业务增长。
发明内容
有鉴于此，本说明书一个或多个实施例提供一种用户风险评估方法及装置、电子设备、存储介质。
为实现上述目的，本说明书一个或多个实施例提供技术方案如下。
根据本说明书一个或多个实施例的第一方面，提出了一种用户风险评估方法，包括：将目标合作方的用户的行为信息输入对应于所述目标合作方的学生风控模型；所述学生风控模型通过基于所述目标合作方的目标样本数据的软标签值和所述目标样本数据原本被标注的被作为硬标签值的风险标签值，对所述目标样本数据进行知识蒸馏得到，所述软标签值通过多个教师风控模型针对所述目标样本数据的预测结果进行集成得到，各个教师风控模型通过对相应的其他合作方的样本数据进行训练得到；其中，任一样本数据包含被标注有风险标签值的行为信息；根据所述学生风控模型的输出结果确定所述用户的风险评分。
根据本说明书一个或多个实施例的第二方面，提出了一种基于机器学习模型的知识迁移方法，包括：获取多个教师网络针对来自于目标领域的目标样本数据的预测结果，各个教师网络通过对各自源领域的样本数据进行训练得到；对获取到的多个预测结果进行集成，得到对应于所述目标样本数据的软标签值；基于所述软标签值和所述目标样本数据原本被标注的硬标签值，对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
根据本说明书一个或多个实施例的第三方面，提出了一种基于机器学习模型的知识迁移方法，包括：将接收到的来自于目标领域的目标样本数据输入教师网络，所述教师网络通过自身对所属源领域的样本数据进行训练得到；向所述目标样本数据的提供方返回所述教师网络输出的预测结果，以使得所述提供方对所述预测结果和其他教师网络针对所述目标样本数据的预测结果进行集成得到对应于所述目标样本数据的软标签值，以及基于所述软标签值和所述目标样本数据原本被标注的硬标签值对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
根据本说明书一个或多个实施例的第四方面，提出了一种用户风险评估装置，包括：信息输入单元，将目标合作方的用户的行为信息输入对应于所述目标合作方的学生风控模型；所述学生风控模型通过基于所述目标合作方的目标样本数据的软标签值和所述目标样本数据原本被标注的被作为硬标签值的风险标签值，对所述目标样本数据进行知识蒸馏得到，所述软标签值通过对多个教师风控模型针对所述目标样本数据的预测结果进行集成得到，各个教师风控模型通过对相应的其他合作方的样本数据进行训练得到；其中，任一样本数据包含被标注有风险标签值的行为信息；风险评估单元，根据所述学生风控模型的输出结果确定所述用户的风险评分。
根据本说明书一个或多个实施例的第五方面，提出了一种基于机器学习模型的知识迁移装置，包括：预测结果获取单元，获取多个教师网络针对来自于目标领域的目标样本数据的预测结果，各个教师网络通过对各自源领域的样本数据进行训练得到；集成学习单元，对获取到的多个预测结果进行集成，得到对应于所述目标样本数据的软标签值；学生网络训练单元，基于所述软标签值和所述目标样本数据原本被标注的硬标签值，对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
根据本说明书一个或多个实施例的第六方面，提出了一种基于机器学习模型的知识迁移装置，包括：样本数据输入单元，将接收到的来自于目标领域的目标样本数据输入教师网络，所述教师网络通过自身对所属源领域的样本数据进行训练得到；预测结果返回单元，向所述目标样本数据的提供方返回所述教师网络输出的预测结果，以使得所述提供方对所述预测结果和其他教师网络针对所述目标样本数据的预测结果进行集成得到对应于所述目标样本数据的软标签值，以及基于所述软标签值和所述目标样本数据原本被标注的硬标签值对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
根据本说明书一个或多个实施例的第七方面，提出了一种电子设备，包括：处理 器；用于存储处理器可执行指令的存储器；其中，所述处理器通过运行所述可执行指令以实现如上述第一方面中所述的用户风险评估方法。
根据本说明书一个或多个实施例的第八方面，提出了一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器通过运行所述可执行指令以实现如上述第二方面中所述的基于机器学习模型的知识迁移方法。
根据本说明书一个或多个实施例的第九方面，提出了一种电子设备，包括：处理器；用于存储处理器可执行指令的存储器；其中，所述处理器通过运行所述可执行指令以实现如上述第三方面中所述的基于机器学习模型的知识迁移方法。
根据本公开实施例的第十方面，提供一种计算机可读存储介质，其上存储有计算机指令，该指令被处理器执行时实现如上述第一方面中所述的用户风险评估方法的步骤。
根据本公开实施例的第十一方面，提供一种计算机可读存储介质，其上存储有计算机指令，该指令被处理器执行时实现如上述第二方面中所述的基于机器学习模型的知识迁移方法的步骤。
根据本公开实施例的第十二方面，提供一种计算机可读存储介质，其上存储有计算机指令，该指令被处理器执行时实现如上述第三方面中所述的基于机器学习模型的知识迁移方法的步骤。
附图说明
图1是一示例性实施例提供的一种基于机器学习模型的知识迁移系统的架构示意图。
图2是一示例性实施例提供的一种基于机器学习模型的知识迁移方法的流程图。
图3是一示例性实施例提供的另一种基于机器学习模型的知识迁移方法的流程图。
图4是一示例性实施例提供的一种用户风险评估方法的流程图。
图5是一示例性实施例提供的一种基于机器学习模型的知识迁移方法的交互图。
图6是一示例性实施例提供的一种设备的结构示意图。
图7是一示例性实施例提供的一种用户风险评估装置的框图。
图8是一示例性实施例提供的另一种设备的结构示意图。
图9是一示例性实施例提供的一种基于机器学习模型的知识迁移装置的框图。
图10是一示例性实施例提供的另一种设备的结构示意图。
图11是一示例性实施例提供的另一种基于机器学习模型的知识迁移装置的框图。
具体实施方式
这里将详细地对示例性实施例进行说明，其示例表示在附图中。下面的描述涉及附图时，除非另有表示，不同附图中的相同数字表示相同或相似的要素。以下示例性实施例中所描述的实施方式并不代表与本说明书一个或多个实施例相一致的所有实施方式。相反，它们仅是与如所附权利要求书中所详述的、本说明书一个或多个实施例的一些方面相一致的装置和方法的例子。
需要说明的是：在其他实施例中并不一定按照本说明书示出和描述的顺序来执行相应方法的步骤。在一些其他实施例中，其方法所包括的步骤可以比本说明书所描述的更多或更少。此外，本说明书中所描述的单个步骤，在其他实施例中可能被分解为多个步骤进行描述；而本说明书中所描述的多个步骤，在其他实施例中也可能被合并为单个步骤进行描述。
图1是一示例性实施例提供的一种基于机器学习模型的知识迁移系统的架构示意图。如图1所示，该系统可以包括服务器11、网络12、若干电子设备，比如手机13、手机14和PC15-16等。
服务器11可以为包含一独立主机的物理服务器，或者该服务器11可以为主机集群承载的虚拟服务器。在运行过程中，服务器11作为服务端与各个合作方对接，也即向各个合作方提供合作的平台，用于将与之对接的各个合作方训练的教师网络的性能迁移到学生网络中。
手机13-14、PC15-16只是用户可以使用的一种类型的电子设备。实际上，与服务器11对接的合作方显然还可以使用诸如下述类型的电子设备：平板设备、笔记本电脑、掌上电脑(PDAs，Personal Digital Assistants)、可穿戴设备(如智能眼镜、智能手表等)等，本说明书一个或多个实施例并不对此进行限制。在本说明书一个或多个实施例的技术方案中，各个合作方利用自身的样本数据训练得到教师网络，从而可指导相关的学生网络的训练，将教师网络学习到的模型参数(也可理解为教师网络学到的知识)分享给学生网络从而提升学生网络的性能。
而对于手机13-14、PC15-16与服务器11之间进行交互的网络12，可以包括多种类型的有线或无线网络。在一实施例中，该网络12可以包括公共交换电话网络(Public Switched Telephone Network，PSTN)和因特网。
图2是一示例性实施例提供的一种基于机器学习模型的知识迁移方法的流程图。如图2所示，该方法应用于服务端，可以包括步骤202～206。
步骤202，获取多个教师网络针对来自于目标领域的目标样本数据的预测结果，各个教师网络通过对各自源领域的样本数据进行训练得到。
在本实施例中，在训练监督式机器学习模型时，收集标注有标签值的样本数据可能存在一定困难，例如，样本数据因时间问题积累较少，收集样本数据的数据量较大，耗时，成本较高。进一步的，即便在样本数据充足的情况下，从头开始构建模型的成本较高，效率较低。因此。当存在训练某一领域的监督式机器学习模型的需求时，可利用迁移学习(Transfer Learning)技术，将与该领域相关(比如，属于同一类型，相似度较高等)的已经训练好的模型学习到的知识，迁移至该领域的机器学习模型中，从而提高训练模型的效率。换言之，利用已有的知识来学习新的知识，已有的知识和新的知识之间存在相似性。在迁移学习中，将已有知识所属领域称为源领域(source domain)，待学习的新知识所属领域称为目标领域(target domain)；其中，源领域通常有大量标签数据，而目标领域往往只有少量标签样本，源领域和目标领域不同但有一定关联，可通过减小源领域和目标领域的分布差异，进而进行知识迁移。
进一步的，在迁移过程中，引入知识蒸馏(Knowledge Distillation)技术来提高待训练模型的泛化能力和性能。具体而言，采用教师-学生网络(teacher-student network)，通过对教师网络进行知识蒸馏以指导训练学生网络。其中，教师网络往往是一个更加复杂的网络，具有非常好的性能和泛化能力，可以将教师网络作为一个soft target来指导另外一个更加简单的学生网络进行学习，使得更加简单、参数运算量更少的学生模型也能够具有和教师网络相近的性能。
在本说明书一个或多个实施例的技术方案中，教师网络与源领域相对应，即由源领域已经训练好的监督式学习模型作为教师网络，用于指导学生网络的学习，将自身学习到的知识迁移至学生网络，而学生网络与目标领域相对应，即由目标领域的待训练模型作为学生网络。
在本实施例中，当与服务端对接的某一合作方存在待训练模型时，服务端可通过 对其他与该合作方所属领域相关的合作方已经训练好的监督式机器学习模型进行迁移学习，以指导该待训练模型的学习。那么，在训练目标领域的学生网络的过程中，无需重新收集大量目标领域的样本数据以进行训练，从而可提高训练学生网络的效率。同时，学生网络还可继承教师网络较好的泛化能力和性能。
在本实施例中，可以选取一个或多个教师网络来指导学生网络的训练。其中，源领域与教师网络一一对应。为了提高学生网络的泛化能力和性能(即能够将教师网络的泛化能力和性能较好地迁移至学生网络)，可选取与目标领域相似度较高的领域作为源领域。作为一示例性实施例，可设定为各个源领域与目标领域属于同一类型。例如，在图像识别领域，均用于识别车辆、均用于识别猫科动物、均用于人脸识别等。
在本实施例中，在选取多个教师网络的情况下，本说明书的基于机器学习模型的知识迁移方案，可理解为各个源领域的数据提供方共同协同合作来完成对学生网络的训练，即多个数据提供方拥有自己的样本数据，可共同使用彼此的数据来统一训练机器学习模型。需要注意的是，各个数据提供方的样本数据属于自身的隐私数据，因此上述多方联合建模(joint modelling)的过程应在保证各方数据安全的情况下进行。因此，数据提供方作为训练教师网络的执行主体，分别在各自的源领域利用自身标注的样本数据来训练得到教师网络。换言之，各个教师网络通过各自源领域的数据提供方将自身的隐私数据作为样本数据进行训练得到。由此可见，一方面，各个数据提供方协同合作训练各自的教师网络，可提高后续训练学生网络的效率；另一方面，各个源领域的教师网络的训练过程都不用出域，可以保证各个源领域的样本数据的隐私。
步骤204，对获取到的多个预测结果进行集成，得到对应于所述目标样本数据的软标签值。
在本实施例中，为了提高训练出的学生网络为多样性(全面性)的强监督模型，使得学生网络稳定且在各个方面表现都较好，而非存在偏好(弱监督模型，在某些方面表现的比较好)，可对获取到的多个教师网络的预测结果进行集成学习。通过对获取到的多个预测结果进行集成学习，可在某一教师网络针对目标样本数据存在错误预测的情况下，通过其他的教师网络将该错误预测纠正，从而减小方差(bagging)、偏差(boosting)和改进预测(stacking)的效果。其中，集成学习的具体实现方式可根据实际情况灵活选取，本说明书一个或多个实施例并不对此进行限制。例如，可采取投票、加权平均等方式。又如，可采用Bagging(bootstrap aggregating，装袋；例如随机森林)、Boosting和Stacking等算法。
步骤206，基于所述软标签值和所述目标样本数据原本被标注的硬标签值，对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
在本实施例中，硬标签值为目标样本数据中原本被标注的标签值。例如，硬标签值由目标样本数据的提供方(属于目标领域)对目标样本数据进行标注得到。在通过集成学习得到对应于目标样本数据的软标签值(soft target)后，基于软标签值和目标样本数据原本被标注的硬标签值(hard target)，对目标样本数据进行知识蒸馏以得到目标领域的学生网络。源自目标样本数据(数据量较小)原本被标注的hard target，包含的信息量(信息熵)较低；而soft target来自于大模型(教师网络)的预测输出，具有更高的熵，能比hard target提供更加多的信息。因此，通过soft target来辅助hard target一起训练，也即使用较少的数据以及较大的学习率，使得更加简单、参数运算量更少的学生模型也能够具有和教师网络相近的性能(因此也可理解为一种模型压缩的方式)。换言之，学生网络的训练含有两个目标函数：一个与hard target对应，即原始的目标函数，为学生网络的类别概率输出与标签(label)真值的交叉熵；另一个与soft target对应，为学生网络的类别概率输出与教师网络的类别概率输出的交叉熵。在soft target中，在softmax函数中增加温度参数T：
Figure PCTCN2020126961-appb-000001
其中，q i是第i类的概率值大小，输入z i是第i类的预测向量(对数logits)；logits是分类模型生成的原始(非标准化)，预测向量通常会传递给标准化函数。当模型要解决多类别分类问题时，则logits通常作为softmax函数的输入，以由softmax函数生成一个(标准化)概率向量，对应于每个可能的类别。softmax函数通过将输入z i与其他logits进行比较，将每个类别的logit z i计算为概率q i。
进一步的，Loss值为：L＝αL (soft)+(1-α)L (hard)。其中soft loss指的是对student model(学生网络)中softmax(T＝20)的输出与teacher model(教师网络)的softmax(T＝20)的输出求loss1；hard loss指的是对softmax(T＝1)的输出与原始label求loss2。
比如，可将与hard target对应的目标函数和与soft target对应的目标函数通过加权平均来作为学生网络的最终目标函数。例如，可以设定为soft target所占的权重更大一些。又如，T值可取一个中间值，而soft target所分配的权重为T^2，hard target的权重为1。当然，还可为其他任意权重设定，本说明书一个或多个实施例并不对此进行限制。
同时，由于针对目标领域的学生网络的训练过程无任何限制，因此可得到解释性 强的学生网络。以分类器为例，由于对分类器没有限制，则可采用解释性强的分类器进行训练。
相应的，图3是一示例性实施例提供的另一种基于机器学习模型的知识迁移方法的流程图。如图3所示，该方法应用于教师网络的提供方，可以包括步骤302～304。
步骤302，将接收到的来自于目标领域的目标样本数据输入教师网络，所述教师网络通过自身对所属源领域的样本数据进行训练得到。
在本实施例中，教师网络的提供方可以是训练教师网络的样本数据提供方。为了保证样本数据提供方的隐私，样本数据应不被向外泄露。因此，样本数据提供方作为训练教师网络的执行主体，分别在各自的源领域利用自身标注的样本数据来训练得到教师网络。可见，各个源领域的教师网络的训练过程都不用出域，可以保证各个源领域的样本数据的隐私。
在本实施例中，对于将来自目标领域的目标样本数据输入教师网络得到的预测结果，可利用差分隐私技术来保护决策隐私(即保证各个教师网络输出结果的隐私)。因此，可对教师网络输出的预测结果进行差分隐私处理，再向目标样本数据的提供方返回被进行差分隐私处理的预测结果。具体而言，可针对预测结果引入拉普拉斯噪声(Laplacian Noises)，通过以下公式对教师网络输出的预测结果进行差分隐私处理：
f(i)+Lap(1/ε)；
其中，f(i)表示第i个样本数据的概率预测数值；Lap(1/ε)表示以0为中心并按1/ε缩放的拉普拉斯概率分布，ε表示用于控制隐私保护程度的参数。
当然，差分隐私具体的实现机制可根据实际情况灵活选取，本说明书一个或多个实施例并不对此进行限制。例如，Laplace机制、Laplace分布、指数机制等。
步骤304，向所述目标样本数据的提供方返回所述教师网络输出的预测结果，以使得所述提供方对所述预测结果和其他教师网络针对所述目标样本数据的预测结果进行集成得到对应于所述目标样本数据的软标签值，以及基于所述软标签值和所述目标样本数据原本被标注的硬标签值对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
在本说明书一个或多个实施例的技术方案中，样本数据的具体内容可根据实际应用场景灵活设定。比如，样本数据的数据类型可以包含图像、文本、语音等。同样的，对样本数据的标注也可以根据实际应用场景灵活设定，下面举例进行说明。
在对实体对象进行风控的场景中，可对用户或商户潜在的风险进行预测，比如预测借贷、实时交易的风险。以实时交易为例，合作平台与商户对接合作，各个商户在营业过程中已积累有大量的样本数据。其中，样本数据(以文本形式，或者为其他数据类型)包括用户的基本信息、行为信息、交易信息等。并且，商户可在交易风险维度上对样本数据进行标注。当合作平台新接入一家新开业的商户a时，由于自身掌握的样本数据有限，导致无法训练得到较为准确全面的风控模型。那么，该新接入的商户a可联合合作平台上其他同类型的商户进行联合建模。在该情况下，新接入的商户a属于目标领域，自身掌握的少量样本数据为目标样本数据，待训练的风控模型为学生网络；合作平台上其他与该新接入的商户为同一行业(比如同属于基金、保险公司等)的商户1-n属于源领域，商户1-n可利用各自积累的大量样本数据训练得到教师网络以指导学生网络的训练。而在完成对学生网络的联合建模后，商户a便可将获取到的用户的基本信息、行为信息、交易信息等数据输入该学生网络，从而预测当前与该用户进行的交易的风险评分。
在智能推荐的场景中，可对用户潜在的需求进行预测，比如预测用户想买的商品、感兴趣的新闻、喜欢看的书籍等。以卖家向用户推荐商品为例，合作平台与多个卖家对接合作，各个卖家在营业过程中已积累有大量的用户购买记录。其中，样本数据(以文本形式，或者为其他数据类型)为职业、收入、年龄、性别等用户信息，商户可根据用户购买记录中用户购买的商品对样本数据进行标注。当合作平台新接入一卖家a时，由于自身的历史用户有限，导致无法向用户推荐商品。那么，该新接入的卖家a可联合合作平台上其他同类型的卖家进行联合建模。在该情况下，新接入的商户a属于目标领域，自身掌握的少量用户购买记录作为目标样本数据，待训练的商品推荐模型为学生网络；合作平台上其他与该新接入的卖家为同一行业(比如同属于餐饮、服装等)的卖家1-n属于源领域，卖家1-n可利用各自积累的大量用户购买记录训练得到教师网络以指导学生网络的训练。而在完成对学生网络的联合建模后，卖家a便可将获取到的用户的用户信息输入该学生网络，从而预测该用户可能存在购买需求的商品，进而根据预测结果向该用户推荐相应的商品。
在智能客服的场景中，可实时与用户进行语音对话，解答用户疑问或者与用户聊天。例如，合作平台与多家企业合作，各个企业在向用户提供客服服务的过程中已积累有大量的对话数据。其中，样本数据可以为用户输入的文本、图像、用户的语音等，针对样本数据的标注为对话数据中客服向用户回复的内容。当另外一家企业a新接入合作平台，并希望向用户提供智能客服的服务时，若自身掌握的用户与客服之间的对话数据 有限，则可联合合作平台中其他企业进行联合建模。比如，可由提供语音助手、聊天工具、解答疑问等客服服务的企业1-n通过各自积累的对话数据进行联合建模。其中，企业1-n的客服与用户的对话场景存在一定的相似度。在该情况下，新接入的企业a属于目标领域，自身掌握的少量对话数据为目标样本数据，待训练的客服模型为学生网络；企业1-n属于源领域，企业1-n可利用各自积累的大量对话数据训练得到教师网络以指导学生网络的训练。而在完成对学生网络的联合建模后，企业a(或者企业1-n)便可利用该学生网络向用户提供智能客服的服务，即将用户发起的对话内容(文本、图像、语音等)作为该学生网络的输入，从而将输出结果作为本次对话的回复。
下面以风控的应用场景为例，对上述实施例训练得到的学生网络的应用过程进行说明。请参见图4，图4是一示例性实施例提供的一种用户风险评估方法的流程图。如图4所示，该评估方法可以包括步骤402～404。
步骤402，将目标合作方的用户的行为信息输入对应于所述目标合作方的学生风控模型；所述学生风控模型通过基于所述目标合作方的目标样本数据的软标签值和所述目标样本数据原本被标注的被作为硬标签值的风险标签值，对所述目标样本数据进行知识蒸馏得到，所述软标签值通过对多个教师风控模型针对所述目标样本数据的预测结果进行集成得到，各个教师风控模型通过对相应的其他合作方的样本数据进行训练得到；其中，任一样本数据包含被标注有风险标签值的行为信息。
步骤404，根据所述学生风控模型的输出结果确定所述用户的风险评分。
在本实施例中，在风控的应用场景下，学生风控模型与上述图2-3实施例中的学生网络相对应，而教师风控模型与上述图2-3实施例中的教师网络相对应。训练各个模型的样本数据的具体内容为用户的行为信息，标注内容为用户的风险评分；换言之，各个模型的输入是用户的行为信息，输出为用户的风险评分(包括概率分布)。多方在同一平台合作，目标合作方属于目标领域，为目标样本数据的提供方，待训练模型为学生风控模型，那么可通过其他合作方的教师风控模型来指导学生风控模型的训练。其中，训练的具体过程可参考上述图2-3所示的实施例，在此不再赘述。
而在训练得到对应于目标合作方的学生风控模型后，在一种情况下，可在目标合作方的客户端侧配置该学生风控模型，那么目标合作方在获取用户的行为信息后，可通过客户端向学生风控模型输入行为信息，以根据输出结果确定该用户的风险评分，进而决定后续针对该用户的处理方式。例如，当风险评分较低时(说明该用户较为安全)，可向该用户发放消费权益；当风险评分较高时(说明该用户存在潜在风险)，可拦截该 用户的注册请求。在另一种情况下，可将学生风控模型配置于与目标合作方对接的服务端侧，那么目标合作方在获取用户的行为信息后，可通过客户端向服务端发送该行为信息，以由服务端利用学生风控模型来确定该用户的风险评分并返回至客户端进行展示。
在本实施例中，为了提高学生风控模型的泛化能力和性能(即能够将教师风控模型的泛化能力和性能较好地迁移至学生风控模型)，可选取与目标合作方相似度较高的其他合作方的教师风控模型来指导学生风控模型的训练。作为一示例性实施例，可设定为目标合作方和该其他合作方属于同一类型的合作方。例如，均属于餐饮类，均属于金融类等。
在本实施例中，为了保护各个其他合作方的隐私安全，各个教师风控模型通过相应的其他合作方对自身的样本数据进行训练得到。换言之，其他合作方作为训练教师风控模型的执行主体，分别利用自身标注的样本数据来训练得到教师风控模型。由此可见，一方面，各个合作方协同合作训练各自的教师风控模型，可提高后续训练学生风控模型的效率；另一方面，各个教师风控模型的训练过程都不用出域，可以保证各个源领域的样本数据的隐私。
为了便于理解，下面结合应用场景和举例对本说明书的基于机器学习模型的知识迁移方案的交互过程进行详细说明。如图5所示，该交互过程可以包括以下步骤：
步骤502A，合作方1通过自身标注的隐私数据训练得到教师网络1。
步骤502B，合作方2通过自身标注的隐私数据训练得到教师网络2。
步骤502C，合作方n通过自身标注的隐私数据训练得到教师网络n。
需要说明的是，步骤502A-502C之间为互相并列的步骤，在时间上的先后顺序并无要求。
在本实施例中，以风控场景为例，“商户健康分”是服务端作为商家合作平台向ISV(Independent Software Vendors，独立软件开发商)渠道商针对渠道商下的商家一种风险评估的指标，通过对渠道商下的商家的“商户健康分”进行评估，可帮助合作伙伴(ISV渠道商)提升风控能力。在ISV渠道商对用于评估商户健康分的模型进行建模的过程中，由于掌握的商户行为数据有限(即样本数据有限)，可借助于商家合作平台从其他合作方(其他ISV渠道商)积累的商户行为数据进行联合建模。其中，联合建模的其他合作方应与该ISV渠道商存在一定的关联，例如属于同一行业。以下以ISV渠道商与合作方1-n联合建模为例进行说明。
其中，合作方1-n对在历史营业过程中商户的行为信息进行在风险维度上的标注，进而得到用于训练教师网络的样本数据(属于自身的隐私数据)，也即训练得到的教师网络的输入为商户的行为信息，输出为相应的风险评分。而针对训练所采用的监督式机器学习算法，可根据实际情况灵活选取，本说明书一个或多个实施例并不对此进行限制。以下以分类器为例进行说明。
步骤504A，合作平台向合作方1发送目标样本数据。
步骤504B，合作平台向合作方2发送目标样本数据。
步骤504C，合作平台向合作方n发送目标样本数据。
在本实施例中，可由ISV渠道商向合作平台发送目标样本数据(即自身掌握的商户行为信息)，以由合作平台基于目标样本数据与合作方1-n进行联合建模。当然，也可由ISV渠道商直接与合作方1-n进行联合建模，即本实施例中合作平台执行的步骤由ISV渠道商直接执行。
需要说明的是，向合作方1-n分享目标样本数据的方式存在多种可能，可根据实际情况灵活设定，上述步骤504A-504C仅作为一示例性举例，本说明书一个或多个实施例并不对此进行限制。比如，还可由合作平台将目标样本数据发送至合作方1，再由合作方1分别向合作方2-n转发目标样本数据。
步骤506A，合作方1将目标样本数据输入教师网络1得到预测结果1。
步骤506B，合作方2将目标样本数据输入教师网络2得到预测结果2。
步骤506C，合作方n将目标样本数据输入教师网络n得到预测结果n。
以分类器为例进行说明，假设教师网络和学生网络解决的是一个有M个类别(classes)的多分类问题，给定一个目标样本数据xi，每个分类器fk(教师网络)都能预测出一个概率分布fk(xi)，那么可以通过集成学习技术来对每个fk(xi)进行集成以得到最终分数。
步骤508A，合作方1对预测结果1进行差分隐私处理。
步骤508B，合作方2对预测结果2进行差分隐私处理。
步骤508C，合作方n对预测结果n进行差分隐私处理。
承接于上述举例，各个合作方对于得到的预测结果，可利用差分隐私技术来保护决策隐私(即保证各个教师网络输出结果的隐私)。因此，可对教师网络输出的预测结 果进行差分隐私处理。比如，可在每个分类器的概率预测数值(即预测结果)上引入拉普拉斯噪声(Laplacian Noises)，即fk(xi)+Lap(1/ε)；其中，Lap(1/ε)表示以0为中心并按1/ε缩放的拉普拉斯概率分布，ε表示用于控制隐私保护程度的参数。当然，差分隐私具体的实现机制可根据实际情况灵活选取，本说明书一个或多个实施例并不对此进行限制。例如，Laplace机制、Laplace分布、指数机制等。
步骤510A，合作方1向合作平台返回差分隐私处理后的预测结果1。
步骤510B，合作方2向合作平台返回差分隐私处理后的预测结果2。
步骤510C，合作方n向合作平台返回差分隐私处理后的预测结果n。
类似的，本说明书不对步骤506A-506C、步骤508A-508C、步骤510A-510C中并列的步骤之间设定时间先后顺序的要求。
步骤512，合作平台对预测结果1-n进行集成得到软标签值。
在本实施例中，为了提高训练出的学生网络为多样性(全面性)的强监督模型，使得学生网络稳定且在各个方面表现都较好，而非存在偏好(弱监督模型，在某些方面表现的比较好)，可对获取到的预测结果1-n进行集成学习从而得到对应于目标样本数据的软标签值。例如，将集成学习的结果作为对应于目标样本数据的软标签值。通过对获取到的多个预测结果进行集成学习，可在某一教师网络针对目标样本数据存在错误预测的情况下，通过其他的教师网络将该错误预测纠正，从而减小方差(bagging)、偏差(boosting)和改进预测(stacking)的效果。其中，集成学习的具体实现方式可根据实际情况灵活选取，本说明书一个或多个实施例并不对此进行限制。例如，可采取投票、求平均等方式。又如，可采用Bagging(bootstrap aggregating，装袋；例如随机森林)、Boosting和Stacking等算法。
步骤514，合作平台基于软标签值和目标样本数据原本被标注的硬标签值，对目标样本数据进行知识蒸馏得到学生网络。
以采用求平均的方式进行集成学习为例，针对所有分类器进行差分隐私处理后的概率分布输出取平均，并将取平均得到的最终概率输出作为一个soft target来指导学生网络学习。而目标样本数据原本被标注(比如，由目标域的ISV渠道商对自身积累的商户行为信息进行标注)的标签值定义为hard target(硬标签值)，那么最终的标签值Target＝a*hard target+b*soft target(a+b＝1)，Target则作为训练学生网络的最终标签值。其中，参数a，b是用于控制标签融合权重，比如，a＝0.1，b＝0.9。
通过上述训练的过程，可以得到一输入为商户的行为信息，输出为相应风险评分的学生网络。在一种情况下，可在ISV渠道商的客户端侧配置该学生网络，那么该ISV渠道商在获取到商户的行为信息后，可通过客户端向学生网络输入行为信息，以根据输出结果确定该商户的风险评分，进而决定后续针对该商户的处理方式。例如，当风险评分较低时(说明该商户较为安全)，可向该商户发放消费权益；当风险评分较高时(说明该商户存在潜在风险)，可拦截该商户的注册请求。在另一种情况下，可将学生网络配置于合作平台，那么ISV渠道商在获取到商户的行为信息后，可通过客户端向合作平台发送该行为信息，以由合作平台利用学生网络来确定该商户的风险评分并返回至客户端进行展示。
与上述方法实施例相对应，本说明书还提供了装置实施例。
本说明书的用户风险评估装置的实施例可以应用在电子设备上。装置实施例可以通过软件实现，也可以通过硬件或者软硬件结合的方式实现。以软件实现为例，作为一个逻辑意义上的装置，是通过其所在电子设备的处理器将非易失性存储器中对应的计算机程序指令读取到内存中运行形成的。
从硬件层面而言，图6是一示例性实施例提供的一种设备的示意结构图。请参考图6，在硬件层面，该设备包括处理器602、内部总线604、网络接口606、内存608以及非易失性存储器610，当然还可能包括其他业务所需要的硬件。处理器602从非易失性存储器610中读取对应的计算机程序到内存608中然后运行，在逻辑层面上形成用户风险评估装置。当然，除了软件实现方式之外，本说明书一个或多个实施例并不排除其他实现方式，比如逻辑器件抑或软硬件结合的方式等等，也就是说以下处理流程的执行主体并不限定于各个逻辑单元，也可以是硬件或逻辑器件。
请参考图7，在软件实施方式中，该用户风险评估装置可以包括：信息输入单元71，将目标合作方的用户的行为信息输入对应于所述目标合作方的学生风控模型；所述学生风控模型通过基于所述目标合作方的目标样本数据的软标签值和所述目标样本数据原本被标注的被作为硬标签值的风险标签值，对所述目标样本数据进行知识蒸馏得到，所述软标签值通过对多个教师风控模型针对所述目标样本数据的预测结果进行集成得到，各个教师风控模型通过对相应的其他合作方的样本数据进行训练得到；其中，任一样本数据包含被标注有风险标签值的行为信息；风险评估单元72，根据所述学生风控模型的输出结果确定所述用户的风险评分。
可选的，各个教师风控模型通过相应的其他合作方对自身的样本数据进行训练得 到。
本说明书的基于机器学习模型的知识迁移装置的实施例可以应用在电子设备上。装置实施例可以通过软件实现，也可以通过硬件或者软硬件结合的方式实现。以软件实现为例，作为一个逻辑意义上的装置，是通过其所在电子设备的处理器将非易失性存储器中对应的计算机程序指令读取到内存中运行形成的。
从硬件层面而言，图8是一示例性实施例提供的一种设备的示意结构图。请参考图8，在硬件层面，该设备包括处理器802、内部总线804、网络接口806、内存808以及非易失性存储器810，当然还可能包括其他业务所需要的硬件。处理器802从非易失性存储器810中读取对应的计算机程序到内存808中然后运行，在逻辑层面上形成基于机器学习模型的知识迁移装置。当然，除了软件实现方式之外，本说明书一个或多个实施例并不排除其他实现方式，比如逻辑器件抑或软硬件结合的方式等等，也就是说以下处理流程的执行主体并不限定于各个逻辑单元，也可以是硬件或逻辑器件。
请参考图9，在软件实施方式中，该基于机器学习模型的知识迁移装置可以包括：预测结果获取单元91，获取多个教师网络针对来自于目标领域的目标样本数据的预测结果，各个教师网络通过对各自源领域的样本数据进行训练得到；集成学习单元92，对获取到的多个预测结果进行集成，得到对应于所述目标样本数据的软标签值；学生网络训练单元93，基于所述软标签值和所述目标样本数据原本被标注的硬标签值，对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
可选的，各个源领域与所述目标领域属于同一类型。
可选的，各个教师网络通过各自源领域的数据提供方将自身的隐私数据作为样本数据进行训练得到。
可选的，所述目标样本数据和/或各个源领域的样本数据的数据类型包含以下至少之一：图像、文本、语音。
本说明书的基于机器学习模型的知识迁移装置的实施例可以应用在电子设备上。装置实施例可以通过软件实现，也可以通过硬件或者软硬件结合的方式实现。以软件实现为例，作为一个逻辑意义上的装置，是通过其所在电子设备的处理器将非易失性存储器中对应的计算机程序指令读取到内存中运行形成的。
从硬件层面而言，图10是一示例性实施例提供的一种设备的示意结构图。请参考图10，在硬件层面，该设备包括处理器1002、内部总线1004、网络接口1006、内存1008 以及非易失性存储器1010，当然还可能包括其他业务所需要的硬件。处理器1002从非易失性存储器1010中读取对应的计算机程序到内存1008中然后运行，在逻辑层面上形成基于机器学习模型的知识迁移装置。当然，除了软件实现方式之外，本说明书一个或多个实施例并不排除其他实现方式，比如逻辑器件抑或软硬件结合的方式等等，也就是说以下处理流程的执行主体并不限定于各个逻辑单元，也可以是硬件或逻辑器件。
请参考图11，在软件实施方式中，该基于机器学习模型的知识迁移装置可以包括：样本数据输入单元1101，将接收到的来自于目标领域的目标样本数据输入教师网络，所述教师网络通过自身对所属源领域的样本数据进行训练得到；预测结果返回单元1102，向所述目标样本数据的提供方返回所述教师网络输出的预测结果，以使得所述提供方对所述预测结果和其他教师网络针对所述目标样本数据的预测结果进行集成得到对应于所述目标样本数据的软标签值，以及基于所述软标签值和所述目标样本数据原本被标注的硬标签值对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
可选的，所述预测结果返回单元1102具体用于：对所述教师网络输出的预测结果进行差分隐私处理；向所述提供方返回被进行差分隐私处理的预测结果。
可选的，通过以下公式对所述教师网络输出的预测结果进行差分隐私处理：
f(i)+Lap(1/ε)；
其中，f(i)表示第i个样本数据的概率预测数值；Lap(1/ε)表示以0为中心并按1/ε缩放的拉普拉斯概率分布，ε表示用于控制隐私保护程度的参数。
可选的，还包括：隐私获取单元1103，获取自身所属源领域的隐私数据；教师网络训练单元1104，将所述隐私数据作为样本数据进行训练以得到所述教师网络。
可选的，所述目标样本数据和/或各个源领域的样本数据的数据类型包含以下至少之一：图像、文本、语音。
上述实施例阐明的系统、装置、模块或单元，具体可以由计算机芯片或实体实现，或者由具有某种功能的产品来实现。一种典型的实现设备为计算机，计算机的具体形式可以是个人计算机、膝上型计算机、蜂窝电话、相机电话、智能电话、个人数字助理、媒体播放器、导航设备、电子邮件收发设备、游戏控制台、平板计算机、可穿戴设备或者这些设备中的任意几种设备的组合。
在一个典型的配置中，计算机包括一个或多个处理器(CPU)、输入/输出接口、网络接口和内存。
内存可能包括计算机可读介质中的非永久性存储器，随机存取存储器(RAM)和/或非易失性内存等形式，如只读存储器(ROM)或闪存(flash RAM)。内存是计算机可读介质的示例。
计算机可读介质包括永久性和非永久性、可移动和非可移动媒体可以由任何方法或技术来实现信息存储。信息可以是计算机可读指令、数据结构、程序的模块或其他数据。计算机的存储介质的例子包括，但不限于相变内存(PRAM)、静态随机存取存储器(SRAM)、动态随机存取存储器(DRAM)、其他类型的随机存取存储器(RAM)、只读存储器(ROM)、电可擦除可编程只读存储器(EEPROM)、快闪记忆体或其他内存技术、只读光盘只读存储器(CD-ROM)、数字多功能光盘(DVD)或其他光学存储、磁盒式磁带、磁盘存储、量子存储器、基于石墨烯的存储介质或其他磁性存储设备或任何其他非传输介质，可用于存储可以被计算设备访问的信息。按照本文中的界定，计算机可读介质不包括暂存电脑可读媒体(transitory media)，如调制的数据信号和载波。
还需要说明的是，术语“包括”、“包含”或者其任何其他变体意在涵盖非排他性的包含，从而使得包括一系列要素的过程、方法、商品或者设备不仅包括那些要素，而且还包括没有明确列出的其他要素，或者是还包括为这种过程、方法、商品或者设备所固有的要素。在没有更多限制的情况下，由语句“包括一个……”限定的要素，并不排除在包括所述要素的过程、方法、商品或者设备中还存在另外的相同要素。
上述对本说明书特定实施例进行了描述。其它实施例在所附权利要求书的范围内。在一些情况下，在权利要求书中记载的动作或步骤可以按照不同于实施例中的顺序来执行并且仍然可以实现期望的结果。另外，在附图中描绘的过程不一定要求示出的特定顺序或者连续顺序才能实现期望的结果。在某些实施方式中，多任务处理和并行处理也是可以的或者可能是有利的。
在本说明书一个或多个实施例使用的术语是仅仅出于描述特定实施例的目的，而非旨在限制本说明书一个或多个实施例。在本说明书一个或多个实施例和所附权利要求书中所使用的单数形式的“一种”、“所述”和“该”也旨在包括多数形式，除非上下文清楚地表示其他含义。还应当理解，本文中使用的术语“和/或”是指并包含一个或多个相关联的列出项目的任何或所有可能组合。
应当理解，尽管在本说明书一个或多个实施例可能采用术语第一、第二、第三等来描述各种信息，但这些信息不应限于这些术语。这些术语仅用来将同一类型的信息彼 此区分开。例如，在不脱离本说明书一个或多个实施例范围的情况下，第一信息也可以被称为第二信息，类似地，第二信息也可以被称为第一信息。取决于语境，如在此所使用的词语“如果”可以被解释成为“在……时”或“当……时”或“响应于确定”。
以上所述仅为本说明书一个或多个实施例的较佳实施例而已，并不用以限制本说明书一个或多个实施例，凡在本说明书一个或多个实施例的精神和原则之内，所做的任何修改、等同替换、改进等，均应包含在本说明书一个或多个实施例保护的范围之内。

Claims (26)

  1. 一种用户风险评估方法，包括：
    将目标合作方的用户的行为信息输入对应于所述目标合作方的学生风控模型；所述学生风控模型通过基于所述目标合作方的目标样本数据的软标签值和所述目标样本数据原本被标注的被作为硬标签值的风险标签值，对所述目标样本数据进行知识蒸馏得到，所述软标签值通过对多个教师风控模型针对所述目标样本数据的预测结果进行集成得到，各个教师风控模型通过对相应的其他合作方的样本数据进行训练得到；其中，任一样本数据包含被标注有风险标签值的行为信息；
    根据所述学生风控模型的输出结果确定所述用户的风险评分。
  2. 根据权利要求1所述的方法，各个教师风控模型通过相应的其他合作方对自身的样本数据进行训练得到。
  3. 一种基于机器学习模型的知识迁移方法，包括：
    获取多个教师网络针对来自于目标领域的目标样本数据的预测结果，各个教师网络通过对各自源领域的样本数据进行训练得到；
    对获取到的多个预测结果进行集成，得到对应于所述目标样本数据的软标签值；
    基于所述软标签值和所述目标样本数据原本被标注的硬标签值，对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
  4. 根据权利要求3所述的方法，各个教师网络通过各自源领域的数据提供方将自身的隐私数据作为样本数据进行训练得到。
  5. 根据权利要求3所述的方法，所述目标样本数据和各个源领域的样本数据的数据类型包含以下至少之一：图像、文本、语音。
  6. 一种基于机器学习模型的知识迁移方法，包括：
    将接收到的来自于目标领域的目标样本数据输入教师网络，所述教师网络通过自身对所属源领域的样本数据进行训练得到；
    向所述目标样本数据的提供方返回所述教师网络输出的预测结果，以使得所述提供方对所述预测结果和其他教师网络针对所述目标样本数据的预测结果进行集成得到对应于所述目标样本数据的软标签值，以及基于所述软标签值和所述目标样本数据原本被标注的硬标签值对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
  7. 根据权利要求6所述的方法，所述向所述目标样本数据的提供方返回所述教师网络输出的预测结果，包括：
    对所述教师网络输出的预测结果进行差分隐私处理；
    向所述提供方返回被进行差分隐私处理的预测结果。
  8. 根据权利要求7所述的方法，通过以下公式对所述教师网络输出的预测结果进行差分隐私处理：
    f(i)+Lap(1/ε)；
    其中，f(i)表示第i个样本数据的概率预测数值；
    Lap(1/ε)表示以0为中心并按1/ε缩放的拉普拉斯概率分布，ε表示用于控制隐私保护程度的参数。
  9. 根据权利要求6所述的方法，还包括：
    获取自身所属源领域的隐私数据；
    将所述隐私数据作为样本数据进行训练以得到所述教师网络。
  10. 根据权利要求6所述的方法，所述目标样本数据和各个源领域的样本数据的数据类型包含以下至少之一：图像、文本、语音。
  11. 一种用户风险评估装置，包括：
    信息输入单元，将目标合作方的用户的行为信息输入对应于所述目标合作方的学生风控模型；所述学生风控模型通过基于所述目标合作方的目标样本数据的软标签值和所述目标样本数据原本被标注的被作为硬标签值的风险标签值，对所述目标样本数据进行知识蒸馏得到，所述软标签值通过对多个教师风控模型针对所述目标样本数据的预测结果进行集成得到，各个教师风控模型通过对相应的其他合作方的样本数据进行训练得到；其中，任一样本数据包含被标注有风险标签值的行为信息；
    风险评估单元，根据所述学生风控模型的输出结果确定所述用户的风险评分。
  12. 根据权利要求11所述的装置，各个教师风控模型通过相应的其他合作方对自身的样本数据进行训练得到。
  13. 一种基于机器学习模型的知识迁移装置，包括：
    预测结果获取单元，获取多个教师网络针对来自于目标领域的目标样本数据的预测结果，各个教师网络通过对各自源领域的样本数据进行训练得到；
    集成学习单元，对获取到的多个预测结果进行集成，得到对应于所述目标样本数据的软标签值；
    学生网络训练单元，基于所述软标签值和所述目标样本数据原本被标注的硬标签值，对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
  14. 根据权利要求13所述的装置，各个教师网络通过各自源领域的数据提供方将自身的隐私数据作为样本数据进行训练得到。
  15. 根据权利要求13所述的装置，所述目标样本数据和各个源领域的样本数据的数据类型包含以下至少之一：图像、文本、语音。
  16. 一种基于机器学习模型的知识迁移装置，包括：
    样本数据输入单元，将接收到的来自于目标领域的目标样本数据输入教师网络，所述教师网络通过自身对所属源领域的样本数据进行训练得到；
    预测结果返回单元，向所述目标样本数据的提供方返回所述教师网络输出的预测结果，以使得所述提供方对所述预测结果和其他教师网络针对所述目标样本数据的预测结果进行集成得到对应于所述目标样本数据的软标签值，以及基于所述软标签值和所述目标样本数据原本被标注的硬标签值对所述目标样本数据进行知识蒸馏以得到所述目标领域的学生网络。
  17. 根据权利要求16所述的装置，所述预测结果返回单元具体用于：
    对所述教师网络输出的预测结果进行差分隐私处理；
    向所述提供方返回被进行差分隐私处理的预测结果。
  18. 根据权利要求17所述的装置，通过以下公式对所述教师网络输出的预测结果进行差分隐私处理：
    f(i)+Lap(1/ε)；
    其中，f(i)表示第i个样本数据的概率预测数值；
    Lap(1/ε)表示以0为中心并按1/ε缩放的拉普拉斯概率分布，ε表示用于控制隐私保护程度的参数。
  19. 根据权利要求16所述的装置，还包括：
    隐私获取单元，获取自身所属源领域的隐私数据；
    教师网络训练单元，将所述隐私数据作为样本数据进行训练以得到所述教师网络。
  20. 根据权利要求16所述的装置，所述目标样本数据和各个源领域的样本数据的数据类型包含以下至少之一：图像、文本、语音。
  21. 一种电子设备，包括：
    处理器；
    用于存储处理器可执行指令的存储器；
    其中，所述处理器通过运行所述可执行指令以实现如权利要求1或2所述的方法。
  22. 一种电子设备，包括：
    处理器；
    用于存储处理器可执行指令的存储器；
    其中，所述处理器通过运行所述可执行指令以实现如权利要求3-5中任一项所述的方法。
  23. 一种电子设备，包括：
    处理器；
    用于存储处理器可执行指令的存储器；
    其中，所述处理器通过运行所述可执行指令以实现如权利要求6-10中任一项所述的方法。
  24. 一种计算机可读存储介质，其上存储有计算机指令，该指令被处理器执行时实现如权利要求1或2所述方法的步骤。
  25. 一种计算机可读存储介质，其上存储有计算机指令，该指令被处理器执行时实现如权利要求3-5中任一项所述方法的步骤。
  26. 一种计算机可读存储介质，其上存储有计算机指令，该指令被处理器执行时实现如权利要求6-10中任一项所述方法的步骤。
PCT/CN2020/126961 2019-12-14 2020-11-06 用户风险评估方法及装置、电子设备、存储介质 WO2021114974A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN201911287616.0A CN111080123A (zh) 2019-12-14 2019-12-14 用户风险评估方法及装置、电子设备、存储介质
CN201911287616.0 2019-12-14

Publications (1)

Publication Number Publication Date
WO2021114974A1 true WO2021114974A1 (zh) 2021-06-17

Family

ID=70314579

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2020/126961 WO2021114974A1 (zh) 2019-12-14 2020-11-06 用户风险评估方法及装置、电子设备、存储介质

Country Status (2)

Country Link
CN (1) CN111080123A (zh)
WO (1) WO2021114974A1 (zh)

Cited By (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113705362A (zh) * 2021-08-03 2021-11-26 北京百度网讯科技有限公司 图像检测模型的训练方法、装置、电子设备及存储介质
CN113762343A (zh) * 2021-08-04 2021-12-07 德邦证券股份有限公司 处理舆情信息和训练分类模型的方法、装置以及存储介质
CN115423540A (zh) * 2022-11-04 2022-12-02 中邮消费金融有限公司 一种基于强化学习的金融模型知识蒸馏方法及装置
WO2023284321A1 (zh) * 2021-07-15 2023-01-19 华为云计算技术有限公司 预测生存风险率的方法及装置
CN116151627A (zh) * 2023-04-04 2023-05-23 支付宝(杭州)信息技术有限公司 一种业务风控的方法、装置、存储介质及电子设备
CN116306868A (zh) * 2023-03-01 2023-06-23 支付宝(杭州)信息技术有限公司 一种模型的处理方法、装置及设备
CN117275220A (zh) * 2023-08-31 2023-12-22 云南云岭高速公路交通科技有限公司 基于非完备数据的山区高速公路实时事故风险预测方法

Families Citing this family (16)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111080123A (zh) * 2019-12-14 2020-04-28 支付宝(杭州)信息技术有限公司 用户风险评估方法及装置、电子设备、存储介质
CN111523649B (zh) * 2020-05-09 2022-06-10 支付宝(杭州)信息技术有限公司 针对业务模型进行数据预处理的方法及装置
CN111832291B (zh) * 2020-06-02 2024-01-09 北京百度网讯科技有限公司 实体识别模型的生成方法、装置、电子设备及存储介质
CN111832435A (zh) * 2020-06-24 2020-10-27 五邑大学 基于迁移与弱监督的美丽预测方法、装置及存储介质
CN112101545A (zh) * 2020-08-28 2020-12-18 北京百度网讯科技有限公司 蒸馏系统的训练方法、装置、设备及存储介质
CN112085093A (zh) * 2020-09-08 2020-12-15 第四范式（北京）技术有限公司 协同滤波模型的训练方法及装置、可读介质及系统
CN112149404A (zh) * 2020-09-18 2020-12-29 支付宝(杭州)信息技术有限公司 一种用户隐私数据的风险内容识别方法、装置及系统
CN112149179B (zh) * 2020-09-18 2022-09-02 支付宝(杭州)信息技术有限公司 基于隐私保护的风险识别方法及装置
CN113392864A (zh) * 2020-10-13 2021-09-14 腾讯科技（深圳）有限公司 模型生成方法及视频筛选方法、相关装置、存储介质
CN112528109B (zh) * 2020-12-01 2023-10-27 科大讯飞(北京)有限公司 一种数据分类方法、装置、设备及存储介质
CN112734046A (zh) * 2021-01-07 2021-04-30 支付宝(杭州)信息技术有限公司 模型训练及数据检测方法、装置、设备及介质
CN112686234B (zh) * 2021-03-22 2021-06-22 杭州魔点科技有限公司 人脸图像质量评估方法、电子设备和存储介质
CN113313314B (zh) * 2021-06-11 2024-05-24 北京沃东天骏信息技术有限公司 模型训练方法、装置、设备及存储介质
CN113487208A (zh) * 2021-07-16 2021-10-08 支付宝(杭州)信息技术有限公司 风险评估方法及装置
CN114897099A (zh) * 2022-06-06 2022-08-12 上海淇玥信息技术有限公司 基于客群偏差平滑优化的用户分类方法、装置及电子设备
CN115099988A (zh) * 2022-06-28 2022-09-23 腾讯科技（深圳）有限公司 模型训练方法、数据处理方法、设备及计算机介质

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109308418A (zh) * 2017-07-28 2019-02-05 阿里巴巴集团控股有限公司 一种基于共享数据的模型训练方法及装置
CN110059854A (zh) * 2019-03-13 2019-07-26 阿里巴巴集团控股有限公司 用于风险识别的方法及装置
CN110415099A (zh) * 2018-04-28 2019-11-05 深圳先进技术研究院 一种金融信贷诈骗识别方法、系统及电子设备
CA3056098A1 (en) * 2019-06-07 2019-11-22 Tata Consultancy Services Limited Sparsity constraints and knowledge distillation based learning of sparser and compressed neural networks
CN110555148A (zh) * 2018-05-14 2019-12-10 腾讯科技（深圳）有限公司 一种用户行为评估方法、计算设备及存储介质
CN111080123A (zh) * 2019-12-14 2020-04-28 支付宝(杭州)信息技术有限公司 用户风险评估方法及装置、电子设备、存储介质

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107689950B (zh) * 2017-06-23 2019-01-29 平安科技（深圳）有限公司 数据发布方法、装置、服务器和存储介质
CN110223164A (zh) * 2019-06-10 2019-09-10 卓尔智联(武汉)研究院有限公司 基于迁移学习的风控方法及系统、计算机装置、存储介质

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109308418A (zh) * 2017-07-28 2019-02-05 阿里巴巴集团控股有限公司 一种基于共享数据的模型训练方法及装置
CN110415099A (zh) * 2018-04-28 2019-11-05 深圳先进技术研究院 一种金融信贷诈骗识别方法、系统及电子设备
CN110555148A (zh) * 2018-05-14 2019-12-10 腾讯科技（深圳）有限公司 一种用户行为评估方法、计算设备及存储介质
CN110059854A (zh) * 2019-03-13 2019-07-26 阿里巴巴集团控股有限公司 用于风险识别的方法及装置
CA3056098A1 (en) * 2019-06-07 2019-11-22 Tata Consultancy Services Limited Sparsity constraints and knowledge distillation based learning of sparser and compressed neural networks
CN111080123A (zh) * 2019-12-14 2020-04-28 支付宝(杭州)信息技术有限公司 用户风险评估方法及装置、电子设备、存储介质

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
TAIYANG625: "Deep Learning - Knowledge Distillation for Network Compression Training Method. Repost", 14 August 2018 (2018-08-14), Retrieved from the Internet <URL:https://blog.csdn.net/Taiyang625/article/details/81672717> *

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2023284321A1 (zh) * 2021-07-15 2023-01-19 华为云计算技术有限公司 预测生存风险率的方法及装置
CN113705362A (zh) * 2021-08-03 2021-11-26 北京百度网讯科技有限公司 图像检测模型的训练方法、装置、电子设备及存储介质
CN113705362B (zh) * 2021-08-03 2023-10-20 北京百度网讯科技有限公司 图像检测模型的训练方法、装置、电子设备及存储介质
CN113762343A (zh) * 2021-08-04 2021-12-07 德邦证券股份有限公司 处理舆情信息和训练分类模型的方法、装置以及存储介质
CN113762343B (zh) * 2021-08-04 2024-03-15 德邦证券股份有限公司 处理舆情信息和训练分类模型的方法、装置以及存储介质
CN115423540A (zh) * 2022-11-04 2022-12-02 中邮消费金融有限公司 一种基于强化学习的金融模型知识蒸馏方法及装置
CN115423540B (zh) * 2022-11-04 2023-02-03 中邮消费金融有限公司 一种基于强化学习的金融模型知识蒸馏方法及装置
CN116306868A (zh) * 2023-03-01 2023-06-23 支付宝(杭州)信息技术有限公司 一种模型的处理方法、装置及设备
CN116306868B (zh) * 2023-03-01 2024-01-05 支付宝(杭州)信息技术有限公司 一种模型的处理方法、装置及设备
CN116151627A (zh) * 2023-04-04 2023-05-23 支付宝(杭州)信息技术有限公司 一种业务风控的方法、装置、存储介质及电子设备
CN116151627B (zh) * 2023-04-04 2023-09-01 支付宝(杭州)信息技术有限公司 一种业务风控的方法、装置、存储介质及电子设备
CN117275220A (zh) * 2023-08-31 2023-12-22 云南云岭高速公路交通科技有限公司 基于非完备数据的山区高速公路实时事故风险预测方法

Also Published As

Publication number Publication date
CN111080123A (zh) 2020-04-28

Similar Documents

Publication Publication Date Title
WO2021114974A1 (zh) 用户风险评估方法及装置、电子设备、存储介质
WO2021114911A1 (zh) 用户风险评估方法及装置、电子设备、存储介质
WO2019196546A1 (zh) 确定业务请求事件的风险概率的方法及装置
Zheng et al. Capturing the essence of word-of-mouth for social commerce: Assessing the quality of online e-commerce reviews by a semi-supervised approach
CN111724083A (zh) 金融风险识别模型的训练方法、装置、计算机设备及介质
US20210303970A1 (en) Processing data using multiple neural networks
CN112528110A (zh) 确定实体业务属性的方法及装置
Poole et al. On statistic alignment for domain adaptation in structural health monitoring
Sina Mirabdolbaghi et al. Model optimization analysis of customer churn prediction using machine learning algorithms with focus on feature reductions
Li et al. Sentiment based multi-index integrated scoring method to improve the accuracy of recommender system
Cui et al. P2P lending analysis using the most relevant graph-based features
Wang et al. Leveraging Multisource Heterogeneous Data for Financial Risk Prediction: A Novel Hybrid-Strategy-Based Self-Adaptive Method.
Heinen et al. Spatial dependence in subprime mortgage defaults
Tasgetiren et al. On the distributed software architecture of a data analysis workflow: A case study
US20210357699A1 (en) Data quality assessment for data analytics
JP2019160089A (ja) 情報処理装置、情報処理方法及び情報処理プログラム
Hain et al. The promises of Machine Learning and Big Data in entrepreneurship research
Shilpa et al. Default risk prediction based on support vector machine and logit support vector machine
Biswas et al. A Proposed q-Rung Orthopair Fuzzy-Based Decision Support System for Comparing Marketing Automation Modules for Higher Education Admission
Latha et al. Product recommendation using enhanced convolutional neural network for e-commerce platform
Preetham et al. A Stacked Model for Approving Bank Loans
Mohammed et al. Location-aware deep learning-based framework for optimizing cloud consumer quality of service-based service composition
Ljung Synthetic Data Generation for the Financial Industry Using Generative Adversarial Networks
US20230419344A1 (en) Attribute selection for matchmaking
US20240161117A1 (en) Trigger-Based Electronic Fund Transfers

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 20897881

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 20897881

Country of ref document: EP

Kind code of ref document: A1