CN111325444A

CN111325444A - Risk prevention and control decision method, device, system and equipment

Info

Publication number: CN111325444A
Application number: CN202010069040.7A
Authority: CN
Inventors: 付大鹏
Original assignee: Alipay Hangzhou Information Technology Co Ltd
Current assignee: Alipay Hangzhou Information Technology Co Ltd
Priority date: 2020-01-21
Filing date: 2020-01-21
Publication date: 2020-06-23
Anticipated expiration: 2040-01-21
Also published as: CN111325444B

Abstract

The present specification provides a risk prevention and control decision method, apparatus, system and device, the method comprising: and calling a risk decision model by using a risk prevention and control system to carry out risk decision. The risk decision model is obtained by training and constructing historical risk event data serving as sample data by using a reinforcement learning algorithm. The reinforcement learning is applied to a risk prevention and control system, a decision space problem meeting multiple targets is solved by using an optimization algorithm, and the decision space problem is returned to.

Description

Risk prevention and control decision method, device, system and equipment

Technical Field

The present disclosure relates to computer technologies, and in particular, to a risk prevention and control decision method, apparatus, system, and device.

Background

With the development of computer technology and internet technology, more and more scenes of using the internet in work and life are provided, the internet is convenient for people to live and work, but the safety of internet communication or work is more and more important, and the risk prevention and control are more and more important. Such as: users who utilize the internet to carry out online transaction are gradually increased, daily flow especially in the great flow is exponential growth, and the risk form is also complicated changeable day by day, and risk prevention and control is more and more important. The core function in risk prevention and control can be regarded as risk identification and decision making, and the identification means that risks need to be identified quickly and accurately; the decision-making means how to manage and control the decision-making according to the identified risk, and the balance between the risk and the experience is needed, so that the risk is prevented and the experience is protected.

Disclosure of Invention

The embodiment of the specification aims to provide a risk prevention and control decision method, a risk prevention and control decision device, a risk prevention and control decision system and risk prevention and control decision equipment, so that the efficiency and accuracy of risk prevention and control decision are improved.

In one aspect, the present specification provides a risk prevention and control decision method, including:

acquiring risk event data to be prevented and controlled;

and performing risk prevention and control on the risk event data by using a risk decision model in a risk prevention and control system to obtain a risk decision result of the risk event data to be prevented and controlled, wherein the risk decision model is obtained by using a reinforcement learning algorithm and taking historical risk event data and a risk result corresponding to the historical risk event data as sample data to perform model training and construction.

In another aspect, the present specification provides a risk prevention and control decision device, comprising:

the data acquisition module is used for acquiring risk event data to be prevented and controlled;

and the risk decision module is used for performing risk prevention and control on the risk event data by using a risk decision model in a risk prevention and control system to obtain a risk decision result of the risk event data to be prevented and controlled, wherein the risk decision model is obtained by using a reinforcement learning algorithm and taking historical risk event data and a risk result corresponding to the historical risk event data as sample data to perform model training and construction.

In another aspect, an embodiment of the present specification provides a service right application system, including:

the risk decision model is obtained by using a reinforcement learning algorithm and taking historical risk event data and risk results corresponding to the historical risk event data as sample data and conducting model training construction;

and each risk prevention and control layer sequentially calls the risk decision model according to the order of the levels of the risk prevention and control layers from low to high to perform risk prevention and control on the risk event data to be prevented and controlled, wherein the risk prevention and control result output by the lower level risk prevention and control layer is the risk event data of a non-safety event and is used as the input risk event data of the higher level risk prevention and control layer until the target risk prevention and control layer outputs the risk prevention and control result.

In yet another aspect, the present specification provides a risk prevention and control decision processing apparatus comprising: at least one processor and a memory for storing processor-executable instructions, which when executed by the processor implement the above-described risk prevention and control decision method.

The risk prevention and control decision method, the risk prevention and control decision device, the risk prevention and control decision processing equipment and the risk prevention and control decision system apply reinforcement learning to the risk prevention and control system, a decision space problem meeting multiple targets is solved by using an optimization algorithm, and the decision space problem is returned to the decision problem. Through using reinforcement learning, the control intensity is according to environmental change, and the risk prevention and control decision-making has high antagonism, can adapt to the change of risk situation. And the risk prevention and control system is used for calling the risk decision model constructed based on the reinforcement learning algorithm to carry out risk decision, so that the adaptability and the accuracy of the risk decision are improved.

Drawings

In order to more clearly illustrate the embodiments of the present specification or the technical solutions in the prior art, the drawings needed to be used in the description of the embodiments or the prior art will be briefly introduced below, it is obvious that the drawings in the following description are only some embodiments described in the present specification, and for those skilled in the art, other drawings can be obtained according to the drawings without any creative effort.

FIG. 1 is a schematic flow chart of a risk prevention and control decision method in one embodiment of the present disclosure;

FIG. 2 is a block diagram of a risk decision framework in one embodiment of the present description;

FIG. 3 is a schematic illustration of reinforcement learning reward value determination in some embodiments of the present disclosure;

FIG. 4 is a schematic diagram of risk decision making in one embodiment of the present description;

FIG. 5 is a block diagram of an embodiment of a risk prevention and control decision device provided in the present specification;

FIG. 6 is a schematic block diagram of a risk prevention and control system provided in one embodiment of the present description;

fig. 7 is a block diagram of a hardware configuration of a risk prevention and control decision processing server in one embodiment of the present specification.

Detailed Description

In order to make those skilled in the art better understand the technical solutions in the present specification, the technical solutions in the embodiments of the present specification will be clearly and completely described below with reference to the drawings in the embodiments of the present specification, and it is obvious that the described embodiments are only a part of the embodiments of the present specification, and not all of the embodiments. All other embodiments obtained by a person skilled in the art based on the embodiments in the present specification without any inventive step should fall within the scope of protection of the present specification.

People are increasingly looking at computer security or internet security, risk prevention and control is becoming a key technology of internet security, and more risk prevention and control systems for risk prevention and control are used. After the data are input into the risk prevention and control system, the system can perform risk identification, and based on the identification result, a corresponding control method is decided, such as: and carrying out verification, direct failure transaction or authorized user and the like.

The risk prevention and control decision method in the specification can be applied to a client or a server, and the client can be an electronic device such as a smart phone, a tablet computer, a smart wearable device (a smart watch and the like), a smart vehicle-mounted device and the like.

Fig. 1 is a schematic flow chart of a risk prevention and control decision method in an embodiment of the present specification, and as shown in fig. 1, the risk prevention and control decision method provided in an embodiment of the present specification may include:

and 102, acquiring risk event data to be prevented and controlled.

In a specific implementation, the risk event data may represent data related to an event that needs risk control, such as: risk prevention and control are carried out on transaction information of the shopping website, and risk event data can comprise transaction related data such as: transaction amount, transaction channel, user information, merchant information, and the like. If risk prevention and control are carried out on a certain transfer transaction, the risk event data can comprise the transfer amount, a transaction channel and information of users of both transfer parties. According to different risk prevention and control scenes, the specific content of the risk event data is different, and the risk event data can be specifically set according to actual needs. Such as: the risk prevention and control can be performed on platforms needing risk prevention and control such as: and risk prevention and control systems are arranged in the payment platform and the shopping platform, and data for transaction on the platforms are acquired as risk event data to be prevented and controlled.

And 104, performing risk prevention and control on the risk event data by using a risk decision model in a risk prevention and control system to obtain a risk decision result of the risk event data to be prevented and controlled, wherein the risk decision model is obtained by using a reinforcement learning algorithm and taking historical risk event data and a risk result corresponding to the historical risk event data as sample data to perform model training and construction.

The risk prevention and control system may represent a computer processing system for performing risk identification, risk decision, and an algorithm or a machine learning model for performing risk identification, risk decision may be included in the risk prevention and control system. The specific form of the risk prevention and control system can be selected according to actual needs, and the embodiment of the specification is not particularly limited. The acquired risk event data can be input into a risk prevention and control system, risk prevention and control is performed on the risk event data by using a risk decision model in the risk prevention and control system, and the risk prevention and control system can output a risk decision result of the risk event data to be prevented and controlled based on a decision result of the risk decision model, namely, a risk decision result of the risk event data to be prevented and controlled is obtained. The risk decision model can be understood as a risk decision model constructed based on a reinforcement learning algorithm, historical risk event data and an actual risk result can be used as sample data, the reinforcement learning algorithm is used for carrying out risk decision model training, and the obtained risk decision model can output corresponding risk decisions according to input risk event data, such as: pass, verify, fail transaction, authorized user, etc.

Reinforcement learning is learning by an Agent in a trial and error manner, and a reward guidance behavior obtained by interacting with an environment aims to enable the Agent to obtain the maximum reward, and is different from supervised learning in connection insights learning and mainly represented on a reinforcement signal, wherein the reinforcement signal provided by the environment in reinforcement learning is an evaluation (generally, a scalar signal) on the quality of a generated action instead of telling a reinforcement learning system (rls) (relationship learning) how to generate a correct action. Since the information provided by the external environment is very small, the RLS must learn from its own experience. In this way, the RLS gains knowledge in the context of action-assessment, improving the action scheme to adapt to the context.

The agent can sense the state (state) of the external environment and reward (reward) for feedback and perform learning and decision making. The decision function of the agent refers to making different actions (actions) according to the state of the external environment, and the learning function refers to adjusting the strategy according to the reward of the external environment. Environment (environment) refers to all things outside the agent, and is influenced by the action of the agent to change the state of the agent, and feeds back corresponding rewards to the agent.

The basic elements in reinforcement learning may include:

state S is a description of the environment, either discrete or continuous, with a state space of S;

the action a is the description of the behavior of the agent, and can be discrete or continuous, and the action space is A;

the policy π (a | s) is the agent that decides the next action a according to the environment state s.

The reinforcement learning can obtain a result by trying to make some behaviors in advance without any labels, and the algorithm can learn under what conditions to select what behavior to obtain the best result by adjusting the previous behavior through the feedback of whether the result is right or wrong. In the embodiment of the present description, a risk decision model may be constructed based on reinforcement learning by using historical data.

In some embodiments of the present disclosure, the principles of reinforcement learning may refer to the following formula:

in the above formula, G_tCan represent the action at time t, R_t+1May represent the reward at time (t +1), γ may represent a discount factor or attenuation factor, and k may represent time k, which may also be understood as the kth reward.

In the risk prevention and control decision method provided by the embodiment of the specification, reinforcement learning is applied to a risk prevention and control system, a decision space problem meeting multiple targets is solved by using an optimization algorithm, and the decision space problem is returned to the decision problem. Through using reinforcement learning, the control intensity is changed according to the environment, the risk prevention and control decision has high antagonism, can adapt to the change of the risk situation, and improves the accuracy of the risk decision result.

On the basis of the foregoing embodiments, in some embodiments of the present specification, the risk prevention and control system includes multiple risk prevention and control layers, and performing risk prevention and control on the risk event data by using a risk decision model in the risk prevention and control system includes:

according to the order of the grades of the risk prevention and control layers from low to high, sequentially utilizing each risk prevention and control layer in the risk prevention and control system to call the risk decision model, and performing risk prevention and control on the risk event data to be prevented and controlled, wherein the risk prevention and control result output by the lower-grade risk prevention and control layer is the risk event data of a non-safety event and is used as the input risk event data of the higher-grade risk prevention and control layer until the target risk prevention and control layer outputs the risk prevention and control result;

and taking the risk prevention and control result of the target risk prevention and control layer as a risk decision result of the risk event data to be prevented and controlled.

In a specific implementation, in some embodiments of the present description, the risk prevention and control system is a hierarchical system, and the system processing hierarchy divides the risk identification process into a plurality of risk prevention and control layers according to differences in effectiveness, timeliness, cost, and risk processing capability of the process. Different risk prevention and control layers adopt different risk prevention and control methods to carry out risk identification and risk decision, namely, the risk prevention and control system can comprise a plurality of risk prevention and control layers, and each layer of risk prevention and control system has a risk identification and risk decision algorithm of the risk prevention and control system. Fig. 2 is a schematic diagram of a risk decision framework in an embodiment of the present disclosure, and as shown in fig. 2, in some embodiments of the present disclosure, a risk prevention and control system may be divided into 5 layers, where a T0 layer is terminal/near-end upper prevention and control; the T1 layer is a rapid risk identification layer, which can directly perform credibility, blacklist and rapid calculation of user portrait through the capability of caching data, and is expected to filter 90% identification of risk events in the two layers, and complete the whole analysis within 50 ms; the T2 layer is a deep risk analysis layer, and is used for carrying out complete synchronous risk analysis on the residual 10% of risk events which are leaked down, and updating credibility, a list and the like in real time; the T3 layer is a comprehensive risk asynchronous analysis layer, comprehensive early warning is carried out on the risk situation through asynchronous analysis, meanwhile, asynchronous event accumulation and asynchronous analysis processing are divided into the layer, and the early warning capability is enhanced through the improvement of data calculation capability; the T4 layer is an off-line analysis, and the online rules, model maturity and deployment timeliness are accelerated by means of strategy simulation, model training and the like and by means of the most comprehensive data.

The risk prevention and control layers can be classified into levels according to the degree or depth of risk prevention and control, and in the 5-layer risk prevention and control layer described in the above embodiment, the T0 layer may be an end prevention and control layer, the T1 layer may be a fast recognition layer, the T2 layer may be a depth recognition layer, the T3 layer may be an asynchronous processing layer, and the T4 layer may be an offline analysis layer. Generally, the levels of the T0-T3 layers can be set to increase sequentially, wherein T0 is the lowest level risk control layer and T3 is the highest level risk control layer. T0, T1, T2, and T3 in fig. 2 represent different risk prevention and control layers, and specific meanings can refer to the descriptions of the above embodiments, and are not described herein again. As shown in fig. 2, T0, T1, T2, and T3 may perform risk decision sorting from a low level to a high level according to the level of the risk prevention and control layer, perform risk prevention and control on the risk events to be prevented and controlled in sequence according to the sorting, and perform risk identification and risk decision on each risk prevention and control layer in sequence to form a full link of risk prevention and control. And the lower-level risk prevention and control layer of the previous layer calls a risk decision model, after risk prevention and control are carried out on the input risk event data, a risk decision result is output, the risk event data belonging to the non-safety event in the lower-level risk prevention and control layer are used as the input risk event data of the higher-level risk prevention and control layer of the next layer, risk prevention and control are carried out by the higher-level risk prevention and control layer, and the like until the target risk prevention and control layer is reached, and the risk decision result output by the target prevention and control layer is obtained and used as the final risk decision result of the risk event data to be prevented and controlled.

For example: from the layer T0, the layer T0 calls a risk decision model to carry out risk prevention and control on risk event data to be prevented and controlled, after the layer T0 carries out risk prevention and control, it can be determined that part of risk events do not have risks, and the risk events which have risks or are uncertain whether the risks exist, namely non-safety events, can be directly passed through, and can be input into the next layer of risk prevention and control layer, namely the layer T1. And the T1 layer calls a risk decision model to carry out risk prevention and control on the input risk event data, after the T1 layer carries out risk prevention and control, the fact that partial risk events do not have risks can be determined, the partial risk events directly pass through, the risk events with risks or uncertain risk or not can be input into the next risk prevention and control layer, namely the T2 layer, and the like until the target risk prevention and control layer is reached.

The target risk prevention and control layer can be the last risk prevention and control layer in the risk prevention and control system, and can also be a designated risk prevention and control layer in the risk prevention and control system. As shown in fig. 2, the embodiment of this specification mainly aims at online risk prevention and control, and the target risk prevention and control layer in the embodiment of this specification may be understood as the T3 layer in the above embodiment is an asynchronous processing layer. When each risk prevention and control layer calls a risk decision model to perform risk prevention and control on the risk event data, the risk decision model can be called to determine a corresponding risk prevention and control method after the data of the layer is used for risk identification. Such as: the T0 layer is end prevention and control, and can only acquire data of a client end under normal conditions, perform risk identification on risk event data based on the acquired data, and then call a risk decision model to perform risk decision according to a risk identification result. The T1 layer is a rapid identification risk prevention and control layer, and can acquire data of a server, perform risk identification on risk event data based on the acquired data, and then call a risk decision model to perform risk decision according to a risk identification result.

As shown in fig. 2, each risk prevention and control layer in the risk prevention and control system sequentially performs risk prevention and control until reaching a target risk prevention and control layer, i.e., a T3 layer, and a risk prevention and control result output by the T3 layer may be used as a risk decision result of corresponding risk event data, i.e., a final risk prevention and control strategy. As shown in fig. 2, the risk decision result may be a verification or a direct failure transaction or a limited-right user, where the verification method may be determined according to actual needs, such as: the method can be used for short message authentication, voice authentication, gesture password authentication, face recognition and the like, and the right-limited user can express that certain rights of the user are limited.

In the risk prevention and control decision method provided by the embodiment of the specification, reinforcement learning is applied to a risk prevention and control system, a decision space problem meeting multiple targets is solved by using an optimization algorithm, and the decision space problem is returned to the decision problem. Through using reinforcement learning, the control intensity is according to environmental change, and the risk prevention and control decision-making has high antagonism, can adapt to the change of risk situation. The risk decision model is called in sequence by each risk prevention and control layer in the risk prevention and control system to carry out risk decision, the next risk prevention and control layer further carries out risk decision on the basis of the previous risk prevention and control layer, the global optimal decision problem is realized by uniformly considering and deciding the whole link, the decision is prevented from being split in the layered control of the wind control system, the local optimal decision problem is obtained, and the accuracy and the safety of the risk prevention and control are improved.

On the basis of the foregoing embodiments, in some embodiments of the present specification, the risk decision model includes risk decision constraints, and the risk decision constraints include:

setting different disturbance rates, case rates and transaction failure rates based on a service scene, user grouping and a transaction channel;

and setting different risk core methods based on the transaction channel and the risk type.

In a specific implementation process, a risk decision constraint condition may be preset, and the risk decision constraint condition is used as a constraint of an output risk decision result when a risk decision model is called to perform risk decision. In some embodiments of the present description, the risk decision constraints may include disturbance rate, case rate, transaction failure rate, and different core methods. The disturbance rate can be understood as the probability of needing the user to check the identity, the case rate can be understood as the ratio of risk events needing risk prevention and control (such as risk checking, failed transaction and right-limited user) to all risk events, and the transaction failure rate can be understood as the ratio of risk events directly failing to all risk events. Different disturbance rates, case rates and transaction failure rates can be set as constraint conditions of the risk decision according to different business scenes, user groups and transaction channels, and different risk core methods can be set as constraint conditions of the risk decision based on the transaction channels and the risk types.

The service scenes can comprise various different service scenes such as online shopping transaction, transfer transaction, meal ordering transaction and the like, and different service scenes can correspond to different levels of disturbance rate, case rate and transaction failure rate. Such as: the online shopping transaction and the transfer transaction have more possible risks, higher disturbance rate, case rate and transaction failure rate can be set, the probability of the risk of the online meal ordering transaction is usually lower, and the disturbance rate, the case rate and the transaction failure rate can be set lower. In practical applications, users can be divided into different user groups according to their historical transaction information, credit information, occupation, education level, social status, etc., such as: the method can be divided into high-risk user groups, namely user groups with higher risk probability, high-value user groups, namely user groups with higher credit, larger transaction amount, more stable occupation, higher income and higher education level, and can also be used according to the social influence degree of users: the network is large V, etc., and a large V user group is divided, and the method for dividing the user group may be selected according to actual situations, and the embodiments of this specification are not particularly limited. Different user groups can set different levels of disturbance rate, case rate, transaction failure rate, for example: for high-risk user groups, higher disturbance rate, case rate and transaction failure rate can be set, and for high-value people, lower disturbance rate, case rate and transaction failure rate can be set. Similarly, different transaction channels can set different levels of disturbance rate, case rate and transaction failure rate, such as: when the client application program APP (application) is adopted to carry out transaction and a PC (personal computer) end is used to carry out transaction, different levels of disturbance rate, case rate and transaction failure rate can be set.

In practical application, the service scene of the risk event to be prevented and controlled, the user group where the user is located and the transaction channel can be comprehensively considered, and the disturbance rate, the case rate and the transaction failure rate are set. Such as: based on an expert decision algorithm or a machine learning model and the like, a service scene, user grouping and a transaction channel are used as input data, and corresponding disturbance rate, case rate and transaction failure rate are comprehensively determined, so that the user experience is not influenced under the condition of ensuring the transaction safety. For example: for the risk event that the business scene is online shopping, the transaction user belongs to a general user group, and the transaction channel is client APP shopping, the possible risk probability is higher, and higher-level disturbance rate, case rate and transaction failure rate can be set. For the risk event that the business scene is online shopping, the transaction user belongs to a high-value user group, and the transaction channel is client APP shopping, the lower disturbance rate case rate and the lower transaction failure rate can be set.

In addition, for risk events of different transaction channels and risk types, different risk verification methods can be set, the risk types can include account stealing, card stealing and the like, and can be specifically set according to actual needs, and the embodiment of the specification is not specifically limited. For example: for the risk event that the transaction channel is client APP transaction and the risk type is card stealing, the risk verification mode cannot be the verification product related to the output mobile phone, such as: short message verification code checking, voice callback checking and the like. For the risk event that the transaction channel is a PC end and the risk type is a stolen account or a stolen card, the risk verification mode cannot be face identification.

In practical application, the control cost, the system efficiency and the like can be used as risk decision constraint conditions to constrain corresponding risk prevention and control operation, so that the benefit maximization of risk prevention and control is realized.

In the embodiment of the specification, the disturbance rate, the transaction failure rate and the like are used as constraint conditions together, the multi-objective constraint optimization problem is realized, the user experience is not influenced under the condition that the transaction safety is ensured, and a proper risk prevention and control result can be determined more quickly and accurately.

On the basis of the above embodiments, in some embodiments of the present specification, a method for each layer of risk prevention and control layer to call the risk decision model to perform risk prevention and control on risk event data includes:

acquiring a business scene, user information, a transaction channel and a risk type corresponding to the risk event data;

and calling the risk decision model, and determining a corresponding risk prevention and control result by combining the risk decision constraint condition with the acquired business scene, the user information, the transaction channel and the risk type.

In a specific implementation process, when each risk prevention and control layer calls a risk decision model to perform risk prevention and control on risk event data, the risk level, the risk type and the like corresponding to a risk event can be identified based on a model or algorithm of each risk prevention and control layer, and then the risk decision model is called to determine a risk prevention and control strategy. When risk decision is carried out, a business scene, user information, a transaction channel and a risk type corresponding to a risk event can be obtained, a risk decision model is called, a preset risk decision constraint condition is used as a constraint condition of a risk prevention and control strategy, and a risk prevention and control strategy, namely a risk prevention and control result corresponding to each risk event is determined based on the obtained business scene, the user information, the transaction channel and the risk type.

The embodiment of the specification combines the risk decision model and the risk decision constraint condition, constrains the risk decision result, avoids inconvenience brought to a user due to high disturbance rate and transaction failure rate, reduces physical examination of the user, and improves accuracy of the risk decision while determining user experience. The corresponding risk decision result is output quickly and accurately, and the efficiency and accuracy of risk prevention and control are improved.

Based on the above embodiments, in some embodiments of the present specification, the process of constructing the risk decision model by using a reinforcement learning algorithm includes:

setting the environmental parameters, risk prevention and control action parameters and reward parameters of the risk decision model, wherein:

the environmental parameters include: the method comprises the following steps of (1) service scene, transaction channel, user grouping, risk type, transaction amount, risk value, daily transaction number and management and control times;

the risk prevention and control action parameters comprise: release, risk checking method, direct failure transaction, direct right-limited user;

the reward parameters include reward values of different sizes;

inputting historical risk event data into the risk decision model, training the risk decision model based on the environmental parameters, and returning a corresponding reward value according to a risk prevention and control result output by the risk decision model and a risk result corresponding to the historical risk event data until the model precision reaches a preset threshold value.

In a specific implementation process, in some embodiments of the present description, when a risk decision model is trained and constructed based on a reinforcement learning algorithm, reinforcement learning elements in a wind control scene may be preset: environmental parameters (i.e., environmental status state), risk prevention action parameters (i.e., action), reward parameters (i.e., reward). The environmental parameters may include a service scenario, a transaction channel, a user group, a risk type, a transaction amount, a risk score, a daily transaction number, and a management and control frequency, and of course, the number and specific content of the environmental parameters may be adjusted according to actual needs, and the embodiments of the present specification are not limited specifically. The risk score in the environmental parameters can be understood as the risk score output by each risk prevention and control layer when performing risk identification, that is, each risk prevention and control layer can determine the risk score of the risk event data by using the algorithm or model of each layer when performing risk identification, and the risk score can be used as the environmental parameters in the risk decision model for determining the risk policy. The daily transaction count may be understood as the daily transaction count of the user in the risk event data (i.e. the day of risk decision), and the control times may be understood as the daily transaction count of the user in the risk event data (which may be the day) is controlled as follows: the number of times that the verification or transaction is failed or the user is entitled is required. The multiple transactions and the control conditions of the user on the same day are taken into consideration of the current risk decision, the model is subjected to reinforcement learning training, the multiple-sequence decision is realized, the problem that the decision result is inaccurate due to the fact that the historical data is used for controlling the unknown risk in the future is avoided, the accuracy of the risk decision result is improved, and resource loss and disturbance to the user are reduced.

The set risk prevention and control action parameters of the risk decision model, namely the adopted risk prevention and control strategy, can comprise: release, risk verification method, direct failure transaction, direct right-limited user.

Inputting historical risk event data into a risk decision model, training the risk decision model based on set environmental parameters, risk prevention and control action parameters and reward parameters, comparing the risk prevention and control results (namely release, risk core, direct failure transaction or direct right-limited user and the like) output by the risk decision model with the actual risk results of the historical risk event data, returning corresponding reward values according to the setting rules of the reward parameters, and obtaining the risk decision model until the model precision reaches a preset threshold or the training times reaches a specified number.

In the embodiment of the description, the reinforcement learning algorithm is creatively applied to the risk prevention and control scene, the environmental parameters of the reinforcement learning algorithm in the risk prevention and control scene are provided, and the risk decision model is trained based on the set environmental parameter risk prevention and control action parameters and the set reward parameters, so that the risk decision model is more adaptive to the risk prevention and control scene, and the accuracy of risk decision is improved.

Fig. 3 is a schematic diagram of reinforcement learning Reward value determination in some embodiments of the present specification, as shown in fig. 3, where the meaning of Reward may refer to the meaning of the reference value in the embodiments of the present specification, and the white user in fig. 3 is a secure user. A black user is a risky user. As shown in fig. 3, the model may be rewarded as follows:

and if the risk result corresponding to the historical risk event data is a safe user, the risk prevention and control result output by the risk decision model is management and control, the core passes the risk prevention and control result, and a positive first reward value is returned. That is, if the user in the risk data is a white user, when making a risk decision, the risk control is adopted, such as a risk core, and the risk core passes through, a forward first reward value is returned. The white user may be regarded as a user without risk or a user with a low risk level, i.e., a safe user.

And if the risk result corresponding to the historical risk event data is a safe user, outputting a risk prevention and control result which is controlled and passes the control without being verified, and returning a negative second reward value, wherein the second reward value is larger than the first reward value. That is, if the user in the risk data is a white user, when the risk decision is made, the risk control is adopted as a risk core, and the risk core fails, it can be considered that an inaccurate risk policy is adopted for the white user, and a negative second reward value is returned, which can be considered as a penalty.

And if the risk result corresponding to the historical risk event data is a safe user and the output risk prevention and control result is no control, returning a forward third reward value, wherein the third reward value is larger than the second reward value. That is, if the user in the risk data is a white user, when the risk decision is made, no risk control is performed, and the right permission is directly released, because the white user does not have any risk, it can be considered that an accurate risk control strategy is adopted, and a larger positive third reward value is returned.

And if the risk result corresponding to the historical risk event data is a risk user, outputting a risk prevention and control result which is control and passes through the core, and returning a negative second reward value. Namely, if the user in the risk data is a black user, when the risk decision is made, risk control is adopted, such as risk verification, and the risk verification passes through, because the black user is a risk user, the risk verification can have a relatively large risk, a certain loss is possibly brought, and if the risk control strategy is considered to be inaccurate, a negative second reward value is returned. Among them, the black user may be considered as a user with risk or a user with a higher risk level, i.e., a risk user.

If the risk result corresponding to the historical risk event data is a risk user, outputting a risk prevention and control result which is controlled and passes without being verified, and returning a positive third reward value; namely, if the user in the risk data is a black user, when the risk decision is made, risk management and control are adopted, such as risk core, and the risk core passes through, because the black user is a risk user, the core fails to avoid economic loss, the risk prevention and control strategy can be considered to be accurate, and a larger forward third reward value is returned.

And if the risk result corresponding to the historical risk event data is a risk user, the output risk prevention and control result is a negative third reward value which is not controlled. That is, if the user in the risk data is a black user, no management and control is adopted when the risk decision is made, because the black user is a risk user, no matter the economic loss is caused when the control is directly released, the risk prevention and control strategy can be considered to be inaccurate, and a negative third reward value is returned.

The black users and the white users can be determined from a black list and a white list in the database, or determined based on the risk condition of the user corresponding to the risk event data in other manners. A black user may be understood as a user with a higher risk level and a white user may be understood as a user with no risk or a lower risk level. And returning different reward values based on different environments and adopted risk prevention and control actions. Wherein the bonus value may be a specific value such as: 1. 2, 3, 4, etc., positive reward values may represent positive numbers, negative reward values may represent negative numbers, and the model is motivated to learn in the exact direction by the magnitude of the reward values.

Fig. 4 is a schematic diagram illustrating the principle of risk decision in an embodiment of the present specification, and as shown in fig. 4, the embodiment of the present specification innovatively applies reinforcement learning to risk decision, and defines three parameters of reinforcement learning in a wind control scenario, that is: state, action, reward. And adopting corresponding risk decision-making actions based on the defined environmental parameters in the wind control scene, obtaining corresponding reward values, and comprehensively considering risk decision-making constraint conditions when adopting the corresponding risk decision-making actions so as to avoid bringing bad experience to users.

In some embodiments of the present description, the method further comprises:

and returning a corresponding reward value according to the risk decision result of the risk event data to be prevented and controlled and the reward parameter.

As shown in fig. 4, when a risk decision is performed by a risk decision model in an actual application risk prevention and control system, after a risk decision result of a risk event is determined, the risk decision result of each risk event may be stored. User information (which may be obtained after a risk decision based on feedback information or manual verification data of the user, etc.) of the user in the risk event may be based on: the user is a white user or a black user, and the reward parameter returns a corresponding reward value. Namely, the data used in the risk decision can be used as the training sample data of the risk decision model, and the risk decision model is continuously optimized. For example: and performing risk decision on the risk event A by using a risk decision model, wherein an output risk decision strategy is a risk core body, the core body passes through, after risk prevention and control, whether the user has a risk or not can be determined according to complaint information and the like of the user, and if the user does not have the risk and belongs to a white user, a forward first reward value can be returned according to reward parameters.

In the present specification, each embodiment of the method is described in a progressive manner, and the same and similar parts among the embodiments are referred to each other, and each embodiment focuses on the differences from the other embodiments. The relevant points can be obtained by referring to the partial description of the method embodiment.

Based on the risk prevention and control decision method, one or more embodiments of the present specification further provide a risk prevention and control decision device. The apparatus may include systems (including distributed systems), software (applications), modules, components, servers, clients, etc. that use the methods described in the embodiments of the present specification in conjunction with any necessary apparatus to implement the hardware. Based on the same innovative conception, embodiments of the present specification provide an apparatus as described in the following embodiments. Since the implementation scheme of the apparatus for solving the problem is similar to that of the method, the specific apparatus implementation in the embodiment of the present specification may refer to the implementation of the foregoing method, and repeated details are not repeated. As used hereinafter, the term "unit" or "module" may be a combination of software and/or hardware that implements a predetermined function. Although the means described in the embodiments below are preferably implemented in software, an implementation in hardware, or a combination of software and hardware is also possible and contemplated.

Specifically, fig. 5 is a schematic block diagram of an embodiment of a risk prevention and control decision apparatus provided in this specification, and as shown in fig. 5, the risk prevention and control decision apparatus provided in this specification may include: a data acquisition module 51 and a risk decision module 52, wherein:

a data obtaining module 51, which may be configured to obtain risk event data to be prevented and controlled;

the risk decision module 52 may be configured to perform risk control on the risk event data by using a risk decision model in a risk control system, and obtain a risk decision result of the risk event data to be controlled, where the risk decision model is obtained by using a reinforcement learning algorithm and performing model training and construction by using historical risk event data and a risk result corresponding to the historical risk event data as sample data.

The risk prevention and control decision device provided in the embodiment of the present specification applies reinforcement learning to a risk prevention and control system, and finds a decision space problem satisfying multiple objectives by using an optimization algorithm, and returns to the decision problem itself. Through using reinforcement learning, the control intensity is according to environmental change, and the risk prevention and control decision-making has high antagonism, can adapt to the change of risk situation.

On the basis of the foregoing embodiments, in some embodiments of the present specification, the risk prevention and control system includes a plurality of risk prevention and control layers, and the risk decision module is specifically configured to:

The risk prevention and control decision device provided in the embodiment of the present specification applies reinforcement learning to a risk prevention and control system, and finds a decision space problem satisfying multiple objectives by using an optimization algorithm, and returns to the decision problem itself. Through using reinforcement learning, the control intensity is according to environmental change, and the risk prevention and control decision-making has high antagonism, can adapt to the change of risk situation. By uniformly considering and deciding the whole link, the global optimal decision problem is realized, and the problem that the decision is split in the layered control of the wind control system and the local optimal decision problem is obtained is avoided.

On the basis of the foregoing embodiments, in some embodiments of the present specification, the risk decision model includes a risk decision constraint condition, and the setting method of the risk decision constraint condition includes:

In the embodiment of the specification, the disturbance rate, the transaction failure rate and the like are used as constraint conditions together, the multi-objective constraint optimization problem is realized, and a proper risk prevention and control result can be determined more quickly and accurately.

On the basis of the foregoing embodiments, in some embodiments of the present specification, the risk decision module is specifically configured to:

In the embodiment of the specification, the risk decision model and the risk decision constraint condition are combined to constrain the risk decision result, so that inconvenience brought to a user due to high disturbance rate and transaction failure rate is avoided, physical examination of the user is reduced, and the accuracy of the risk decision is improved while user experience is determined. The corresponding risk decision result is output quickly and accurately, and the efficiency and accuracy of risk prevention and control are improved.

On the basis of the foregoing embodiments, in some embodiments of the present specification, the apparatus further includes a model building module, configured to build the risk decision model by using the following method:

the reward parameters include reward values of different sizes;

inputting historical risk event data into the risk decision model, training the risk decision model based on the environmental parameters, the risk prevention and control action parameters and the reward parameters, and returning a corresponding reward value according to a risk prevention and control result output by the risk decision model and a risk result corresponding to the historical risk event data until the model precision reaches a preset threshold value.

In the embodiment of the present specification, reinforcement learning is innovatively applied to risk decision, and three parameters of reinforcement learning in a wind control scene are defined, that is: state, action, reward. And adopting corresponding risk decision-making actions based on the defined environmental parameters in the wind control scene, obtaining corresponding reward values, and comprehensively considering risk decision-making constraint conditions when adopting the corresponding risk decision-making actions so as to avoid bringing bad experience to users.

On the basis of the foregoing embodiments, in some embodiments of the present specification, the model building module is specifically configured to:

for the safe user, if the output risk prevention and control result is management and control and the core passes, returning a forward first reward value;

if the safe user, if the output risk prevention and control result is management and control and the user does not check that the user passes the management and control, returning a negative second reward value, wherein the second reward value is larger than the first reward value;

if the risk result corresponding to the historical risk event data is a safe user, outputting a risk prevention and control result which is management and control and passes the core, and returning a positive first reward value;

if the risk result corresponding to the historical risk event data is a safe user, outputting a risk prevention and control result which is controlled and passes the control without being verified, and returning a negative second reward value, wherein the second reward value is larger than the first reward value;

if the risk result corresponding to the historical risk event data is a safe user, the output risk prevention and control result is a positive third reward value which is larger than the second reward value and is not controlled;

if the risk result corresponding to the historical risk event data is a risk user, outputting a risk prevention and control result which is control and passed by the user, and returning a negative second reward value;

if the risk result corresponding to the historical risk event data is a risk user, outputting a risk prevention and control result which is controlled and passes without being verified, and returning a positive third reward value;

and if the risk result corresponding to the historical risk event data is a risk user, the output risk prevention and control result is a negative third reward value which is not controlled.

In the embodiment of the present specification, reinforcement learning is innovatively applied to risk decision, and three parameters of reinforcement learning in a wind control scene are defined, that is: state, action, reward. And (3) adopting a corresponding risk decision action based on the environment parameters in the defined wind control scene, obtaining a corresponding reward value based on the reward rule of the reward value, and training a risk decision model, so that the trained risk decision model has good adaptability. Meanwhile, when corresponding risk decision actions are adopted, risk decision constraint conditions are comprehensively considered, so that bad experience brought to users is avoided.

On the basis of the foregoing embodiments, in some embodiments of the present specification, the risk decision module is further configured to:

In the embodiment of the present description, data used in a risk decision may be used as training sample data of a risk decision model, and the risk decision model is continuously optimized.

It should be noted that the above-described apparatus may also include other embodiments according to the description of the method embodiment. The specific implementation manner may refer to the description of the above corresponding method embodiment, and is not described in detail herein.

An embodiment of the present specification further provides a risk prevention and control decision processing apparatus, including: at least one processor and a memory for storing processor-executable instructions, the processor implementing the risk prevention and control decision method in the above embodiments when executing the instructions, such as:

acquiring risk event data to be prevented and controlled;

Fig. 6 is a schematic structural diagram of a risk prevention and control system provided in an embodiment of the present specification, and as shown in fig. 6, a risk prevention and control decision system in an embodiment of the present specification may include

and each risk prevention and control layer sequentially calls the risk decision model according to the order of the grades of the risk prevention and control layers from low to high to perform risk prevention and control on the risk event data to be prevented and controlled, wherein the risk prevention and control result output by the lower-grade risk prevention and control layer is the risk event data of a non-safety event and is used as the input risk event data of the higher-grade risk prevention and control layer until the target risk prevention and control layer outputs the risk prevention and control result.

In addition, the system can also comprise a risk decision constraint module, wherein the risk decision constraint module is provided with risk decision constraint conditions and is used for taking the constraint conditions as output risk prevention and control results when each layer of risk prevention and control layer calls the risk decision model to perform risk prevention and control on input risk event data;

the risk decision constraints include:

In the embodiment of the specification, a decision framework of the wind control system is redefined innovatively through reinforcement learning, a decision space problem meeting multiple targets is solved innovatively from an optimization algorithm, and the decision space problem returns to the decision problem. By using reinforcement learning, the risk prevention and control decision has high antagonism and can adapt to the change of risk situations. By uniformly considering and deciding the whole link, the global optimal decision problem is realized, and the decision is prevented from being split in the layered control of the wind control system and is the local optimal decision problem. State, Action and Reward in a wind control scene are innovatively designed and defined. The multi-target constraint problem in the Optimization process is solved by innovatively combining a Reward Constrained Policy Optimization algorithm. The multi-sequence decision problem is innovatively realized for the first time, the consideration of the current decision is brought into consideration of the multi-transaction and control conditions of the user in the same day, and the investment loss and the disturbance to the user are greatly reduced.

It should be noted that the above-described processing device and system may also include other embodiments according to the description of the method embodiment. The specific implementation manner may refer to the description of the above corresponding method embodiment, and is not described in detail herein.

The risk prevention and control decision device or processing device or system provided by the specification can also be applied to various data analysis and processing systems. The system or apparatus or processing device may include any of the risk prevention decision devices described in the embodiments above. The system or apparatus or processing device may be a single server, or may include a server cluster, a system (including a distributed system), software (applications), an actual operation device, a logic gate device, a quantum computer, etc. using one or more of the methods or one or more of the embodiments of the present disclosure, and a terminal device incorporating necessary hardware for implementation. The system for checking for discrepancies may comprise at least one processor and a memory storing computer-executable instructions that, when executed by the processor, implement the steps of the method of any one or more of the embodiments described above.

The method embodiments provided by the embodiments of the present specification can be executed in a mobile terminal, a computer terminal, a server or a similar computing device. Taking an example of the present invention running on a server, fig. 7 is a block diagram of a hardware structure of a risk prevention and control decision processing server in an embodiment of the present invention, where the server may be a risk prevention and control decision device, a risk prevention and control decision processing apparatus, or a system in the foregoing embodiments. As shown in fig. 7, the server 10 may include one or more (only one shown) processors 100 (the processors 100 may include, but are not limited to, a processing device such as a microprocessor MCU or a programmable logic device FPGA, etc.), a memory 200 for storing data, and a transmission module 300 for communication functions. It will be understood by those skilled in the art that the structure shown in fig. 7 is merely an illustration and is not intended to limit the structure of the electronic device. For example, the server 10 may also include more or fewer components than shown in FIG. 7, and may also include other processing hardware, such as a database or multi-level cache, a GPU, or have a different configuration than shown in FIG. 7, for example.

The memory 200 may be used to store software programs and modules of application software, such as program instructions/modules corresponding to the risk prevention and control decision method in the embodiment of the present specification, and the processor 100 executes various functional applications and resource data updates by executing the software programs and modules stored in the memory 200. Memory 200 may include high speed random access memory and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some examples, memory 200 may further include memory located remotely from processor 100, which may be connected to a computer terminal through a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.

The transmission module 300 is used for receiving or transmitting data via a network. Specific examples of the network described above may include a wireless network provided by a communication provider of the computer terminal. In one example, the transmission module 300 includes a Network adapter (NIC) that can be connected to other Network devices through a base station so as to communicate with the internet. In one example, the transmission module 300 may be a Radio Frequency (RF) module, which is used for communicating with the internet in a wireless manner.

The foregoing description has been directed to specific embodiments of this disclosure. Other embodiments are within the scope of the following claims. In some cases, the actions or steps recited in the claims may be performed in a different order than in the embodiments and still achieve desirable results. In addition, the processes depicted in the accompanying figures do not necessarily require the particular order shown, or sequential order, to achieve desirable results. In some embodiments, multitasking and parallel processing may also be possible or may be advantageous.

The method or apparatus provided by the present specification and described in the foregoing embodiments may implement service logic through a computer program and record the service logic on a storage medium, where the storage medium may be read and executed by a computer, so as to implement the effect of the solution described in the embodiments of the present specification.

The storage medium may include a physical device for storing information, and typically, the information is digitized and then stored using an electrical, magnetic, or optical media. The storage medium may include: devices that store information using electrical energy, such as various types of memory, e.g., RAM, ROM, etc.; devices that store information using magnetic energy, such as hard disks, floppy disks, tapes, core memories, bubble memories, and usb disks; devices that store information optically, such as CDs or DVDs. Of course, there are other ways of storing media that can be read, such as quantum memory, graphene memory, and so forth.

The risk prevention and control decision method or apparatus provided in the embodiments of the present specification may be implemented in a computer by a processor executing corresponding program instructions, for example, implemented in a PC end using a c + + language of a windows operating system, implemented in a linux system, or implemented in an intelligent terminal using android and iOS system programming languages, implemented in processing logic based on a quantum computer, or the like.

It should be noted that descriptions of the apparatus, the computer storage medium, and the system described above according to the related method embodiments may also include other embodiments, and specific implementations may refer to descriptions of corresponding method embodiments, which are not described in detail herein.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for the hardware + program class embodiment, since it is substantially similar to the method embodiment, the description is simple, and the relevant points can be referred to only the partial description of the method embodiment.

The embodiments of the present description are not limited to what must be consistent with industry communications standards, standard computer resource data updating and data storage rules, or what is described in one or more embodiments of the present description. Certain industry standards, or implementations modified slightly from those described using custom modes or examples, may also achieve the same, equivalent, or similar, or other, contemplated implementations of the above-described examples. The embodiments using the modified or transformed data acquisition, storage, judgment, processing and the like can still fall within the scope of the alternative embodiments of the embodiments in this specification.

In the 90 s of the 20 th century, improvements in a technology could clearly distinguish between improvements in hardware (e.g., improvements in circuit structures such as diodes, transistors, switches, etc.) and improvements in software (improvements in process flow). However, as technology advances, many of today's process flow improvements have been seen as direct improvements in hardware circuit architecture. Designers almost always obtain the corresponding hardware circuit structure by programming an improved method flow into the hardware circuit. Thus, it cannot be said that an improvement in the process flow cannot be realized by hardware physical modules. For example, a Programmable Logic Device (PLD), such as a Field Programmable Gate Array (FPGA), is an integrated circuit whose Logic functions are determined by programming the Device by a user. A digital system is "integrated" on a PLD by the designer's own programming without requiring the chip manufacturer to design and fabricate application-specific integrated circuit chips. Furthermore, nowadays, instead of manually making an integrated Circuit chip, such Programming is often implemented by "logic compiler" software, which is similar to a software compiler used in program development and writing, but the original code before compiling is also written by a specific Programming Language, which is called Hardware Description Language (HDL), and HDL is not only one but many, such as abel (advanced Boolean Expression Language), ahdl (alternate Language Description Language), traffic, pl (core unified Programming Language), HDCal, JHDL (Java Hardware Description Language), langue, Lola, HDL, laspam, hardsradware (Hardware Description Language), vhjhd (Hardware Description Language), and vhigh-Language, which are currently used in most common. It will also be apparent to those skilled in the art that hardware circuitry that implements the logical method flows can be readily obtained by merely slightly programming the method flows into an integrated circuit using the hardware description languages described above.

The controller may be implemented in any suitable manner, for example, the controller may take the form of, for example, a microprocessor or processor and a computer-readable medium storing computer-readable program code (e.g., software or firmware) executable by the (micro) processor, logic gates, switches, an Application Specific Integrated Circuit (ASIC), a programmable logic controller, and an embedded microcontroller, examples of which include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320, the memory controller may also be implemented as part of the control logic for the memory. Those skilled in the art will also appreciate that, in addition to implementing the controller as pure computer readable program code, the same functionality can be implemented by logically programming method steps such that the controller is in the form of logic gates, switches, application specific integrated circuits, programmable logic controllers, embedded microcontrollers and the like. Such a controller may thus be considered a hardware component, and the means included therein for performing the various functions may also be considered as a structure within the hardware component. Or even means for performing the functions may be regarded as being both a software module for performing the method and a structure within a hardware component.

The systems, devices, modules or units illustrated in the above embodiments may be implemented by a computer chip or an entity, or by a product with certain functions. One typical implementation device is a computer. In particular, the computer may be, for example, a personal computer, a laptop computer, a vehicle-mounted human-computer interaction device, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

Although one or more embodiments of the present description provide method operational steps as described in the embodiments or flowcharts, more or fewer operational steps may be included based on conventional or non-inventive approaches. The order of steps recited in the embodiments is merely one manner of performing the steps in a multitude of orders and does not represent the only order of execution. When the device or the end product in practice executes, it can execute sequentially or in parallel according to the method shown in the embodiment or the figures (for example, in the environment of parallel processors or multi-thread processing, even in the environment of distributed resource data update). The terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, the presence of additional identical or equivalent elements in a process, method, article, or apparatus that comprises the recited elements is not excluded. The terms first, second, etc. are used to denote names, but not any particular order.

For convenience of description, the above devices are described as being divided into various modules by functions, and are described separately. Of course, when implementing one or more of the present description, the functions of each module may be implemented in one or more software and/or hardware, or a module implementing the same function may be implemented by a combination of multiple sub-modules or sub-units, etc. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units is only one logical division, and other divisions may be realized in practice, for example, a plurality of units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable resource data updating apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable resource data updating apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable resource data update apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable resource data update apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

In a typical configuration, a computing device includes one or more processors (CPUs), input/output interfaces, network interfaces, and memory.

The memory may include forms of volatile memory in a computer readable medium, 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 a computer-readable medium.

Computer-readable media, including both non-transitory and non-transitory, removable and non-removable media, may implement information storage by any method or technology. The information may be computer readable instructions, data structures, modules of a program, or other data. Examples of 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, compact disc read only memory (CD-ROM), Digital Versatile Discs (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage, graphene storage or other magnetic storage devices, or any other non-transmission medium that can be used to store information that can be accessed by a computing device. As defined herein, a computer readable medium does not include a transitory computer readable medium such as a modulated data signal and a carrier wave.

As will be appreciated by one skilled in the art, one or more embodiments of the present description may be provided as a method, system, or computer program product. Accordingly, one or more embodiments of the present description may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, one or more embodiments of the present description may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

One or more embodiments of the present description may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. One or more embodiments of the present specification can also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.

The embodiments in the present specification are described in a progressive manner, and the same and similar parts among the embodiments may be referred to each other, and each embodiment focuses on the differences from the other embodiments. In particular, for system embodiments, since they are substantially similar to method embodiments, the description is relatively simple, and the relevant points can be referred to only part of the description of the method embodiments. In the description of the specification, reference to the description of the term "one embodiment," "some embodiments," "an example," "a specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the specification. In this specification, the schematic representations of the terms used above are not necessarily intended to refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples. Furthermore, various embodiments or examples and features of different embodiments or examples described in this specification can be combined and combined by one skilled in the art without contradiction.

The above description is merely exemplary of one or more embodiments of the present disclosure and is not intended to limit the scope of one or more embodiments of the present disclosure. Various modifications and alterations to one or more embodiments described herein will be apparent to those skilled in the art. Any modification, equivalent replacement, improvement or the like made within the spirit and principle of the present specification should be included in the scope of the claims.

Claims

1. A risk prevention and control decision method, comprising:

acquiring risk event data to be prevented and controlled;

2. The method of claim 1, the risk prevention and control system comprising a plurality of risk prevention and control layers, the risk prevention and control of the risk event data using a risk decision model in the risk prevention and control system comprising:

3. The method of claim 1, wherein the risk decision model includes risk decision constraints, and the setting method of the risk decision constraints comprises:

4. The method of claim 3, wherein each risk prevention and control layer calls the risk decision model to perform risk prevention and control on risk event data, and the method comprises the following steps:

5. The method of claim 1, the process of building the risk decision model using a reinforcement learning algorithm comprising:

the reward parameters include reward values of different sizes;

6. The method of claim 5, wherein returning a corresponding reward value according to the risk prevention and control result output by the risk decision model and the risk result corresponding to the historical risk event data comprises:

7. The method of claim 5, further comprising:

8. A risk prevention decision-making apparatus comprising:

9. The apparatus of claim 8, the risk prevention and control system comprising a plurality of risk prevention and control layers, the risk decision module being specifically configured to:

10. The apparatus of claim 8, wherein the risk decision model comprises risk decision constraints, and the setting method of the risk decision constraints comprises:

11. The apparatus of claim 8, the risk decision module to:

12. The apparatus of claim 8, further comprising a model construction module to construct the risk decision model using the following method:

the reward parameters include reward values of different sizes;

13. The apparatus of claim 12, the model building module to be specifically configured to:

14. The apparatus of claim 12, the risk decision module further to:

15. A risk prevention and control system comprising: the risk decision model is obtained by using a reinforcement learning algorithm and taking historical risk event data and risk results corresponding to the historical risk event data as sample data and conducting model training construction;

16. The system according to claim 15, further comprising a risk decision constraint module, wherein a risk decision constraint condition is set in the risk decision constraint module, and the risk decision constraint module is used as a constraint condition of an output risk prevention and control result when each layer of risk prevention and control layer calls the risk decision model to perform risk prevention and control on input risk event data;

the risk decision constraints include:

different disturbance rates, case rates and transaction failure rates are set based on a service scene, user grouping and a transaction channel;

different risk verification methods are set based on the transaction channel and the risk type.

17. A risk prevention decision processing device, comprising: at least one processor and a memory for storing processor-executable instructions, the processor implementing the method of any one of claims 1-7 when executing the instructions.