Risk control system monitoring method and device, storage medium and program product

Abstract

The invention discloses a risk control system monitoring method and device, a storage medium and a program product, and is used for timely and accurately identifying whether the risk control system is abnormal or not and timely outputting alarm information when the risk control system is abnormal. The method comprises the following steps: in response to a received first credit output by a risk control system, predicting a possible third credit of a first user at a second moment based on a possible second credit of the first user at the first moment; the first credit is obtained by evaluating the first user by the risk control system based on the user characteristics of the first user at the second moment, and the first moment is before the second moment; determining the probability of the first credit based on the third credit; based on the probability, determining an abnormal degree of the risk control system at the second moment; and if the abnormal degree is greater than or equal to an abnormal degree threshold value, outputting first alarm information.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a monitoring method, device, storage medium, and program product for a risk control system. Background Art

[0002] In risk control scenarios, risk control systems are often relied upon to assess user creditworthiness, enabling comprehensive risk management for operations such as resource applications. However, the accuracy of risk control systems' assessments depends on their proper functioning. Anomalies in the risk control system can lead to erroneous assessments, impacting the normal processing of business transactions.

[0003] Therefore, it is an urgent problem to monitor the risk control system and issue an alarm when the risk control system is abnormal, so as to ensure the normal operation of the risk control system and improve the accuracy of credit assessment. Summary of the Invention

[0004] The purpose of the embodiments of the present application is to provide a monitoring method, device, storage medium and program product for a risk control system, which is used to promptly and accurately identify whether there is an abnormality in the risk control system, and to promptly output alarm information when the risk control system is abnormal, so as to avoid false alarms or missed alarms caused by "jump point" abnormalities in the credit output of the risk control system.

[0005] In order to achieve the above objectives, the embodiments of the present application adopt the following technical solutions:

[0006] In a first aspect, an embodiment of the present application provides a monitoring method for a risk control system, comprising:

[0007] In response to receiving a first credit score output by the risk control system, predicting a possible third credit score of the first user at a second moment based on a possible second credit score of the first user at a first moment; the first credit score being obtained by the risk control system through an assessment of the first user based on user characteristics of the first user at the second moment, the first moment being before the second moment;

[0008] Determining a probability of the first credit score based on the third credit score; and determining a degree of abnormality of the risk control system at the second moment based on the probability;

[0009] If the abnormality level is greater than or equal to the abnormality level threshold, a first alarm message is output.

[0010] In a second aspect, an embodiment of the present application provides a monitoring device for a risk control system, comprising:

[0011] a first prediction module configured to, in response to receiving a first credit score output by the risk control system, predict a possible third credit score of the first user at a second moment based on a possible second credit score of the first user at a first moment; the first credit score being obtained by the risk control system through an assessment of the first user based on user characteristics of the first user at the second moment, the first moment being before the second moment;

[0012] a first determining module, configured to determine a probability of the first creditworthiness based on the third creditworthiness;

[0013] The second determination module is used to determine the abnormality level of the risk control system at the second moment based on the probability; the output module is used to output the first alarm information if the abnormality level is greater than or equal to the abnormality level threshold.

[0014] In a third aspect, an embodiment of the present application provides an electronic device comprising: a processor; and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the monitoring method of the risk control system provided in the first aspect.

[0015] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium. When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device can execute the monitoring method of the risk control system provided in the second aspect.

[0016] In a fifth aspect, an embodiment of the present application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute some or all of the steps in the monitoring method of the risk control system provided in the first aspect.

[0017] At least one of the above technical solutions adopted in the embodiments of the present application can achieve the following beneficial effects:

[0018] During the credit assessment of a first user by the risk control system, the system receives in real time the first credit score output by the risk control system and, based on the first user's possible second credit score at the previous moment, predicts the first user's possible third credit score at the current moment. Because the third credit score reflects the probability distribution of the first credit score, the probability of the first credit score can be determined based on the third credit score, which reflects whether the first credit score is an abnormal "jump point." Furthermore, based on the probability of the first credit score, the system determines the degree of abnormality of the risk control system at the current moment. If the degree of abnormality is significant, the system determines that a systemic abnormality exists, and outputs a corresponding alarm. This distinguishes between "jump point" anomalies and systemic anomalies in the credit score output by the risk control system, and issues an alarm when a systemic anomaly occurs, effectively preventing false alarms or missed alarms caused by "jump point" anomalies, thereby improving the timeliness and accuracy of alarms. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0020] Figure 1 A schematic diagram of an implementation environment provided for one embodiment of the present application;

[0021] Figure 2 A flow chart of a monitoring method for a risk control system provided in one embodiment of the present application;

[0022] Figure 3 A flow chart of a monitoring method for a risk control system provided in another embodiment of the present application;

[0023] Figure 4 A flowchart of a monitoring method for a risk control system provided in yet another embodiment of the present application;

[0024] Figure 5 A schematic diagram of an experimental result of drift anomaly provided in one embodiment of the present application;

[0025] Figure 6 A schematic diagram of an experimental result of abnormal data fluctuation provided by an embodiment of the present application;

[0026] Figure 7 A schematic diagram of an experimental result of a failure anomaly provided in one embodiment of the present application;

[0027] Figure 8 A schematic diagram of an experimental result of a "jump point" anomaly provided in accordance with an embodiment of the present application;

[0028] Figure 9 A schematic diagram of the structure of a monitoring device for a wind control system provided in one embodiment of the present application;

[0029] Figure 10 A schematic structural diagram of an electronic device provided in accordance with an embodiment of the present application. DETAILED DESCRIPTION

[0030] To make the purpose, technical solutions, and advantages of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the specific embodiments of this application and the corresponding drawings. Obviously, the embodiments described are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0031] Those skilled in the art will appreciate that, with the development of technology and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0032] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequential order. It should be understood that the terms used in this way can be interchangeable under appropriate circumstances, and this is merely a way of distinguishing the objects of the same attributes when describing them in the embodiments of the present application. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, so that the process, method, system, product or equipment comprising a series of units need not be limited to those units, but may include other units that are not clearly listed or inherent to these processes, methods, products or equipment.

[0033] Key terms explained:

[0034] Long Short-term Memory (LSTM): A special type of recurrent neural network that can learn long-term dependencies in data and is a neural network with enhanced temporal information.

[0035] Particle filtering (PF) is a Monte Carlo-based filtering technique used to estimate the instability of system states. It is particularly suitable for systems with nonlinear or non-Gaussian noise. The core idea of ​​particle filtering is to use a set of random samples (particles) to represent the probability distribution of the state and use these samples to estimate statistical properties such as the expectation and variance of the state.

[0036] Through extensive research, the inventors have discovered that the following two types of anomalies often occur during the credit assessment process of a risk control system, making monitoring of the risk control system very difficult:

[0037] First, the risk control system may experience systemic anomalies, resulting in long-term drift, failure, or fluctuation in the creditworthiness assessed by the risk control system. This could be due to significant anomalies in the user characteristics entering the risk control system (e.g., errors in structured data such as personal attributes, behavior, and external data, or anomalies in large amounts of unstructured data such as voice and text), anomalies in the credit assessment models used in the risk control system, or anomalies in the users assessed by the risk control system. In such cases, an alarm must be issued, and emergency investigations and repairs must be carried out on the risk control system. If the risk control system is not repaired promptly, it may continue to output incorrect creditworthiness, impacting subsequent business processing results. For example, in resource application transactions, if the risk control system consistently outputs incorrect creditworthiness, it will misjudge the user's creditworthiness before, during, and after the application, leading to incorrect resource allocation decisions and significant resource losses.

[0038] Second, during the ongoing credit assessment process, the risk control system may output an abnormally low or high credit score. This type of anomaly is commonly referred to as a "jump point" anomaly. This may be due to an unusual change in the user characteristics used by the risk control system for credit assessment, or it may be caused by user anomalies. While this is a low-probability event, it occurs relatively frequently over a large number of users. In this case, if an anomaly alert is issued, repeated alerts will occur, significantly increasing the number of alerts, potentially drowning out systemic anomalies and increasing the workload of business personnel to investigate the anomaly. If an anomaly alert is not issued, individual abnormal users may be missed, which in turn may affect subsequent business processing results. Therefore, it is necessary to distinguish between "jump point" anomalies and systemic anomalies.

[0039] Among related technologies, outlier detection and fixed threshold methods are currently the most commonly used monitoring methods. The core of anomaly detection is to monitor for unlikely extreme situations, such as when the risk control system outputs a negative or zero credit score, and issue an alarm if such an extreme situation occurs. However, this approach has two drawbacks: First, beyond extreme situations, a large number of abnormal conditions remain undetected, such as the systemic anomalies described above; second, these extreme conditions may be "jump point" anomalies, which, because they are not distinguished from systemic anomalies, can overwhelm systemic anomalies or trigger false alarms.

[0040] The fixed threshold method sets a fixed threshold based on prior experience monitoring risk control systems. When the credit score output by the risk control system exceeds the threshold, an alarm is issued. However, this approach also has two drawbacks: First, within a certain time dimension, a user's credit score is time-series data, and the threshold will fluctuate as it evolves. Fixed thresholds may cause false alarms or missed alarms, affecting the accuracy of the alarms. Second, it also fails to distinguish between "jump point" anomalies and systemic anomalies, which may drown out systemic anomalies or cause no alarms.

[0041] In view of this, embodiments of the present application propose a monitoring method for a risk control system. During the risk control system's credit assessment of a user, at each moment, the user's possible creditworthiness at that moment is treated as a particle. A particle filtering algorithm is used to predict the current moment's particles based on the particles at the previous moment, yielding the user's possible creditworthiness at the current moment. These possible creditworthinesses reflect the probability distribution of the creditworthiness output by the risk control system at the current moment. Furthermore, based on the possible creditworthiness at the current moment, the probability of the creditworthiness output by the risk control system can be determined. This probability reflects whether the creditworthiness output by the risk control system exhibits a "jump point" anomaly. Furthermore, based on the probability of the creditworthiness output by the risk control system, the degree of anomaly in the risk control system at the current moment is determined. If the degree of anomaly is significant, a systemic anomaly is determined in the risk control system, and a corresponding alarm is output. In this way, a distinction is made between "jump point" anomalies and systemic anomalies in the creditworthiness output by the risk control system, and an alarm is issued when a systemic anomaly occurs, effectively preventing false or missed alarms caused by "jump point" anomalies, thereby improving the timeliness and accuracy of alarms.

[0042] It should be understood that the monitoring method of the wind control system proposed in the embodiments of the present application can be executed by an electronic device. As an example, it can be executed by software in the electronic device. The electronic device mentioned here can be.

[0043] Before introducing the monitoring method of the risk control system provided by the embodiment of the present application in detail, a brief introduction to the implementation environment involved in the embodiment of the present application is given. Figure 1 , is a schematic diagram of an implementation environment provided by an embodiment of the present application, the implementation environment includes a risk control system 10 and a monitoring system 20. The risk control system 10 is connected to the monitoring system 20 via a wireless network or a wired network.

[0044] Risk control system 10 includes at least one of a single server, multiple servers, a cloud computing platform, and a virtualization center. Risk control system 10 is used to continuously evaluate user creditworthiness and transmit this information to monitoring system 20, which monitors and issues anomaly alerts. For example, risk control system 10 deploys a trained risk control model that assesses user creditworthiness in real time based on real-time user characteristics, such as personal attributes, behavioral data, external data, and other business-related structured data, as well as large amounts of unstructured data such as voice and text messages. This model then transmits this information to monitoring system 20.

[0045] The monitoring system 20 includes one or more electronic devices. The electronic devices herein may include terminal devices such as smartphones, tablets, laptops, desktop computers, intelligent voice interaction devices, smart home appliances, smart watches, vehicle-mounted terminals, aircraft, etc.; alternatively, the electronic devices may also include at least one of a server, a cloud computing platform, and a virtualization center. Upon receiving the credit score sent by the risk control system 10, the monitoring system 20 determines whether the risk control system 10 is abnormal based on the currently received credit score. If the monitoring system 10 is abnormal, it outputs an alarm message to prompt business personnel to analyze the risk control system 10 and determine the factors that caused the abnormality, thereby improving the assessment accuracy of the risk control system 10.

[0046] Based on the implementation environment introduced above, the monitoring method of the risk control system provided by the embodiment of the present application is described in detail with reference to the accompanying drawings.

[0047] Please refer to Figure 2 , is a flow chart of a method for monitoring a risk control system provided in one embodiment of the present application, the method comprising the following steps:

[0048] S202 : In response to receiving the first credit rating output by the risk control system, predicting a possible third credit rating of the first user at a second moment based on a possible second credit rating of the first user at the first moment.

[0049] The first user can be any user to be evaluated. There can be multiple first users, and this embodiment of the present application does not limit this. The first credit rating is obtained by the risk control system based on the first user's user characteristics at the second moment. User characteristics may include, but are not limited to, at least one of the following information: user's personal attributes, behavioral data, external data, and other business-related structured data, as well as a large amount of user voice and text data, such as unstructured business-related data.

[0050] Exemplarily, the risk control system performs big data analysis on the user characteristics and credit scores of a large number of users to obtain a correspondence between user characteristics and credit scores; further, based on the correspondence and the user characteristics of the first user at the second moment, the first credit score of the first user at the second moment is determined.

[0051] For example, a risk control system deploys a trained risk control model. The user characteristics of a first user at a second moment are input into the risk control model for credit assessment, resulting in a first credit score for the first user at the second moment. Specifically, the risk control model can be a logistic regression model, which performs a regression operation based on the user characteristics of the first user at the second moment to determine the probability that the user presents credit risk, and then determines the first credit score for the first user at the second moment based on this probability.

[0052] Alternatively, the risk control model can also be a decision tree model, which divides the first user's user characteristics at the second moment into groups. This division process is based on multi-dimensional data, including but not limited to the user's credit history, income level, consumption habits, social network information, etc., thereby constructing a tree structure. In this tree structure, each node represents a judgment point for a specific feature value or feature condition. For example, one node represents "whether the user's monthly income exceeds a certain amount" and another node represents "whether the user has a record of overdue payments in the past year." These nodes are layered according to logical relationships and importance, forming branches. As the decision tree continues to deepen, the model will eventually reach a leaf node based on the path of the first user's user characteristics at the second moment in the tree structure. This leaf node represents the final evaluation result based on the user characteristics division, that is, the first credit score of the first user at the second moment.

[0053] The second moment is after the first moment. Exemplarily, the second moment is the moment before the first moment, wherein the first moment is recorded as t-1, the second moment is recorded as t, and t is a positive integer.

[0054] In this embodiment of the present application, a first user may have multiple possible second credit ratings at a first moment, and multiple possible third credit ratings at a second moment. Each second credit rating has a corresponding third credit rating. In other words, for each second credit rating, a corresponding third credit rating can be predicted based on the second credit rating.

[0055] Regarding S202, in one implementation, the PF algorithm can be used to treat each second credit as a particle in the state space at the first moment. Based on the state transition equation and the state of each particle, the particle at the second moment is predicted. Each particle at the second moment is a possible third credit of the first user at the second moment. The state transition equation is used to describe the law of how the state of a particle changes from one moment to another, and is expressed as in, represents the possible second credit of the first user at the first moment, represents the third possible credit of the first user at the first moment, n (t) represents the process noise at the second moment, which reflects the noise generated by physical factors in the time-varying process, and f(·) represents the state transfer equation.

[0056] In actual applications, the risk control system predicts the first user's current credit score based on the first user's user characteristics at the previous moment at a preset interval. Simultaneously, after receiving the current first credit score, the monitoring system predicts the first user's current possible third credit score based on the first user's possible second credit score at the previous moment. In this scenario, the previous moment is the first moment, and the current moment is the second moment.

[0057] The possible third credit of the first user at the first moment is predicted based on the initial second credit of the first user. The initial second credit can be set in various appropriate ways. For example, the initial second credit is generated by the prior probability density function p(x), which is denoted as Where N is the number of initial second credits, and the prior probability density function p(x) is usually regarded as a normal distribution with a mean of the third credit at the first moment predicted by the state transition equation and a variance of σ0.

[0058] In another embodiment, considering that the neural network meets the computational requirements of the state transfer equation in the PF algorithm and can process long time series data very well, the trained neural network is used as the state transfer equation to improve the accuracy of the third credit.

[0059] Specifically, the first model can be used as a state transition equation, that is, the third credit score is predicted by the first model based on the second credit score. The first model is trained as follows:

[0060] Step A1: predicting the eighth credit score of the second user at the fifth moment based on the seventh credit score of the second user at the fourth moment using the first model.

[0061] The second user can be any normal user, meaning that the second user is applicable to the risk control model deployed in the risk control system, the credit rating assessed by the risk control model for the second user does not exhibit any "jump point" anomalies, and the second user's user characteristics do not exhibit any anomalies. The fifth and fourth moments are prior to the first moment, with the fourth moment preceding the fifth moment. The seventh credit rating is determined by the risk control system based on the second user's user characteristics at the fourth moment. It is worth noting that the specific implementation of the credit rating of the second user by the risk control system is similar to that of the first user by the risk control system and will not be further elaborated here.

[0062] In the embodiment of the present application, the first model can be any neural network with time series prediction capabilities. For example, the first model can be LSTM. Since LSTM is a recursive neural network, it is more in line with the computational requirements of the state transfer equation in the PF algorithm and can better process long-term time series data. Therefore, using the trained LSTM as the state transfer equation can more accurately predict the user's possible credit at the current moment based on the user's possible credit at the previous moment. In this case, the state transfer equation can be written as

[0063] Step A2: adjusting the parameters of the first model based on the difference between the eighth credit score and the ninth credit score of the second user at the fifth moment.

[0064] The ninth credit rating is obtained by the risk control system through an evaluation of the second user based on the user characteristics of the second user at the fifth moment.

[0065] Specifically, based on the difference between the eighth credit and the ninth credit, the loss of the first model is determined, that is, Among them, Loss represents the loss of the first model, y p represents the eighth credit, N represents the number of eighth credits, y m represents the ninth credit level; then, minimizing the loss of the first model is used as the optimization goal, and the parameters of the first model are adjusted using the backpropagation algorithm. Steps A1 to A2 are repeated multiple times until the training stop condition is met, thus completing the training of the first model. The training stop condition can be set as needed, such as when the number of training times reaches a preset threshold or when the loss of the first model converges, and is not limited in this embodiment of the present application.

[0066] S204: Determine the probability of the first credit level based on the third credit level.

[0067] There are multiple third credit ratings, which reflect the probability distribution of the first user. Based on these third credit ratings, the probability of the first credit rating can be determined. The greater the probability of the first credit rating, the more reliable it is; conversely, the less reliable it is. Therefore, the probability of the first credit rating reflects whether the first credit rating has "jump point" anomalies, that is, whether the first credit rating is an abnormal "jump point."

[0068] In one implementation, the above S204 includes the following steps:

[0069] S241 : Determine a fourth credit score of the first user at the second moment based on the third credit score and the measurement noise at the second moment.

[0070] The measurement noise is obtained by sensor measurement. By introducing the measurement noise at the second moment into the third credit through the preset observation equation, the fourth credit of the first user at the second moment can be obtained, which is recorded as in, represents the fourth credit of the first user at the second moment, h(·) represents the observation equation, v (t) represents the measurement noise at the second moment, Indicates the third credit level of the first user at the second moment.

[0071] S242: Determine a fifth credit level from the fourth credit level.

[0072] The fifth credit is less than or equal to the first credit. For example, for each fourth credit, it is recorded as If the fourth credit is less than or equal to the first credit (denoted as ),Right now The fourth credit level is determined as the fifth credit level.

[0073] S243 : Determine the probability of the first credit level based on the ratio between the number of the fifth credit level and the number of the fourth credit level.

[0074] Specifically, if the first creditworthiness is less than or equal to the expectation of the fourth creditworthiness, the ratio is determined as the probability of the first creditworthiness; if the first creditworthiness is greater than the expectation of the fourth creditworthiness, the difference between the ratio threshold and the ratio is determined as the probability of the first creditworthiness.

[0075] Among them, the ratio threshold can be set according to actual needs, such as the ratio threshold is 1, etc., which is not limited in the embodiment of the present application.

[0076] For example, under normal circumstances, the probability density distribution of the first creditworthiness can be used to obtain the probability of the first creditworthiness, as shown in the following formula (1).

[0077]

[0078] in, represents the probability of the first credit, g(y) t represents the probability distribution density of the first credit, Indicates the expectation of the fourth degree of credit, Indicates the first level of credit.

[0079] In the PF algorithm, the probability density distribution consists of discrete particles, and the weight of each particle is equal, so It can be approximated as N m / N, where, since each particle represents a possible credit, N m represents the number of the fifth credit, and N represents the number of the fourth credit. Therefore, the above formula (1) can be further rewritten as the following formula (2).

[0080]

[0081] In this way, the probability of the first credit rating can be determined more quickly and accurately, providing reliable data support for subsequent abnormal warnings of the risk control system.

[0082] In another implementation, the above S204 includes the following steps: determining a fifth credit level from the third credit level, the fifth credit level being less than or equal to the first credit level; and determining the probability of the first credit level based on a ratio between the fifth credit level and the fourth credit level.

[0083] The above describes some implementations of the above S204. Of course, it should be understood that the above S204 can also be implemented in other ways, which are not limited in the present embodiment.

[0084] In another embodiment of the present application, the credit score output by the risk control system serves as the particle's observed value in the PF algorithm, used for particle resampling and prediction of the particle's state value at the next moment. In other words, the credit score output by the risk control system is used to predict the user's likely credit score at the next moment. To avoid introducing erroneous observations into subsequent calculations and affecting the final processing results, the risk control system monitoring method proposed in this embodiment of the application also introduces corrections for abnormal credit scores output by the risk control system.

[0085] Specifically, after the above S243, it also includes: if the probability of the first credit level is less than or equal to the first probability threshold, the first credit level is corrected based on the probability to obtain a sixth credit level; based on the difference between the sixth credit level and the fourth credit level, the third credit level is sampled for predicting the possible credit level of the first user at a third moment, wherein the third moment is after the second moment.

[0086] As described above, when the probability of a first credit score is less than or equal to the first probability threshold, the first credit score can be determined to be an abnormal "jump point." During the credit assessment process, although users' credit scores often experience extremely high or low values, such occurrences are actually rare events under normal operating conditions of the risk control system. Statistically, for such rare events, the first probability threshold is generally set to less than 1%, i.e., ε = 1%. Secondly, completely discarding an abnormal first credit score is unreasonable because, while the first credit score is erroneous, the trend is correct to some extent. Therefore, when the first credit score is less than or equal to the first probability threshold, the first credit score can be corrected and then used to predict the possible credit score at the next moment. This effectively utilizes the first credit score with a correct trend and avoids completely discarding the abnormal first credit score.

[0087] Specifically, the above-mentioned correction of the first credit level based on the probability to obtain the sixth credit level includes the following steps: based on the relationship between the expectation of the fourth credit level and the first credit level, determining the weight of the first credit level and the weight of the expectation; based on the weight of the first credit level and the weight of the expectation, determining the weighted sum of the first credit level and the expectation to obtain the sixth credit level.

[0088] If the first credit rating is less than or equal to the expectation, the probability is determined as the weight of the first credit rating, and the difference between the second probability threshold and the probability is determined as the weight of the expectation; if the first credit rating is greater than the expectation, the difference between the second probability threshold and the probability is determined as the weight of the expectation, and the probability is determined as the weight of the expectation.

[0089] Exemplarily, the sixth credit can be determined by the following formula (3).

[0090]

[0091] in, Indicates the sixth degree of credit, Indicates the first credit level, Indicates the fourth degree of credit, Indicates the expectation of the fourth degree of credit, Indicates the probability of the first credit level.

[0092] Because the expected fourth creditworthiness reflects the average level of the first user's possible creditworthiness at the second moment, if the first creditworthiness does not exceed the expected fourth creditworthiness, the expected fourth creditworthiness is closer to the first user's actual creditworthiness. Therefore, a relatively small weight is assigned to the first creditworthiness, while a relatively high weight is assigned to the expected fourth creditworthiness. If the first creditworthiness exceeds the expected fourth creditworthiness, the first creditworthiness is closer to reflecting the first user's actual creditworthiness. Therefore, a relatively large weight is assigned to the first creditworthiness, while a relatively small weight is assigned to the expected fourth creditworthiness. The sixth creditworthiness thus obtained can more accurately reflect the first user's actual creditworthiness, providing reliable data support for subsequent accurate prediction of the first user's possible creditworthiness at the next moment.

[0093] The above-mentioned method of determining the weight of the first creditworthiness and the expected weight based on the magnitude relationship between the expected fourth creditworthiness and the first creditworthiness includes the following steps: determining the weight of the third creditworthiness based on the difference between the sixth creditworthiness and the fourth creditworthiness; deleting third creditworthinesses with weights less than or equal to a weight threshold, and copying third creditworthinesses with weights greater than or equal to the weight threshold, to obtain a sampling result.

[0094] For example, the weight of the third credit level can be determined by the following formula (4).

[0095]

[0096] in, represents the weight of the third credit, Indicates the fourth degree of credit, Indicates the sixth degree of credit.

[0097] Furthermore, based on the principle that the total number of third credits remains unchanged after sampling, by replicating third credits with high weights and discarding those with low weights, the third credits are concentrated in high-probability areas, preventing third credit degradation. This process is also called resampling or importance sampling.

[0098] S206: Determine the abnormality level of the risk control system at the second moment based on the probability of the first credit rating.

[0099] The degree of abnormality in the risk control system reflects the possibility of systemic abnormalities in the risk control system.

[0100] As described above, the probability of the first creditworthiness reflects whether there is a "jump point" anomaly in the first creditworthiness, that is, whether the first creditworthiness is an abnormal "jump point". Usually, the "jump point" anomaly is a small-probability event and will not occur continuously. Once the "jump point" anomaly occurs continuously for multiple times, it can be determined that there is a systematic anomaly in the risk control system. Based on this, by setting an initial degree of anomaly, when a "jump point" anomaly appears in the creditworthiness output by the risk control system, the degree of anomaly is accumulated, so that the degree of anomaly increases continuously with the increase in the number of accumulated "jump point" anomalies and decreases with the increase in the number of accumulated normal times, which is equivalent to achieving the dynamic accumulation of "jump point" anomalies, thereby realizing the distinction between "jump point" anomalies and systematic anomalies and effectively preventing the mis-triggering of alarms for "jump point" anomalies. In addition, the degree of anomaly is also reduced, which is equivalent to endowing the degree of anomaly with a "slow decline" mechanism and improving the response rate of the degree of anomaly.

[0101] Specifically, if the probability of the first creditworthiness is less than or equal to the first probability threshold, the sum of the degree of anomaly of the risk control system at the first moment and the first change amount is determined as the degree of anomaly of the risk control system at the second moment; if the probability of the first creditworthiness is greater than the first probability threshold, the difference between the degree of anomaly of the risk control system at the first moment and the second change amount is determined as the degree of anomaly of the risk control system at the second moment, and the second change amount is less than the first change amount.

[0102] Exemplarily, the degree of anomaly of the risk control system is initialized, that is, the initial degree of anomaly of the risk control system is set to 0, denoted as Furthermore, the degree of anomaly of the risk control system at the second moment can be determined by the following formula (5).

[0103]

[0104] Among them, represents the degree of anomaly of the risk control system at the second moment; represents the degree of anomaly of the risk control system at the first moment; a represents the growth coefficient, and a > 1, generally taken as 4; r represents the reduction coefficient, and 0 < r < 1, generally taken as 0.05. Such values can bring a "slow decline" mechanism to the degree of anomaly and improve the response speed of the degree of anomaly. In this case, the product of the growth coefficient and the reduction coefficient is the first change amount, and the reduction coefficient is the second change amount.

[0105] S208, if the degree of anomaly of the risk control system at the second moment is greater than or equal to the anomaly degree threshold, output the first alarm message.

[0106] The greater the degree of anomaly of the risk control system at the second moment, the greater the possibility that there is a systematic anomaly in the risk control system; when the degree of anomaly at the second moment reaches a threshold, it can be determined that there is a systematic anomaly in the risk control system, that is, Wherein, represents the abnormality threshold, which can be set according to actual needs and is not limited in the present embodiment. For example, if a = 4, r = 0.05, and β = 1, then after 5 consecutive "jump point" anomalies, it is determined that there is a systemic abnormality in the risk control system, and the first alarm information is output.

[0107] One or more embodiments of the present application provide a monitoring method for a risk control system. During the process of the risk control system performing a credit assessment on a first user, the method receives a first credit score output by the risk control system in real time, and predicts the first user's possible third credit score at the current moment based on the first user's possible second credit score at the previous moment. Because the third credit score reflects the probability distribution of the first credit score, the probability of the first credit score can be determined based on the third credit score, which reflects whether the first credit score is an abnormal "jump point". Furthermore, based on the probability of the first credit score, the abnormality level of the risk control system at the current moment is determined. If the abnormality level is large, it is determined that the risk control system has a systemic abnormality, and corresponding alarm information is output. In this way, the "jump point" abnormality of the credit score output by the risk control system is distinguished from a systemic abnormality, and an alarm is issued when a systemic abnormality occurs, effectively preventing false alarms or missed alarms caused by "jump point" abnormalities, thereby improving the timeliness and accuracy of alarms.

[0108] The monitoring method for the risk control system provided in the embodiments of the present application can be applied to various business scenarios. The following is an example of a resource application scenario. It should be understood that the application of this method to the resource application scenario is only an exemplary description and should not be understood as a limitation on the application scenario of the method.

[0109] In the resource application scenario, the first user is the user who submits the resource application. Figure 3 As shown, the risk control system periodically performs a credit assessment on the first user based on the user characteristics of the first user at the current moment at a preset interval, obtains the first credit score of the first user at the current moment, and sends it to the monitoring system.

[0110] In response to receiving the first credit score, the monitoring system predicts the first user's possible third credit score at the current moment based on the first user's possible second credit score at the previous moment and the state transition equation. It also predicts the first user's fourth credit score at the current moment based on the third credit score and the observation equation. The monitoring system then determines the probability of the first credit score based on the magnitude relationship between the expected fourth credit score and the first credit score. If this probability is less than or equal to a first probability threshold, the first credit score is determined to be an abnormal "jump point," indicating that the risk control system currently has a "jump point" anomaly. This triggers a correction procedure for the first credit score and an abnormality warning procedure.

[0111] Specifically, the correction process for the first credit rating includes: determining the weight of the first credit rating and the expected weight based on the size relationship between the expected fourth credit rating and the first credit rating, and then performing a weighted sum of the first credit rating and the expected weight based on these weights to obtain a sixth credit rating as the corrected first credit rating.

[0112] Furthermore, the risk control system determines a weight of the third credit rating based on the difference between the sixth credit rating and the fourth credit rating, normalizes the weight, resamples the third credit rating based on the normalized weight to obtain a new third credit rating, and uses the new third credit rating to predict the possible credit rating of the first user at the next moment.

[0113] The abnormality warning process for the risk control system includes: if the probability of the first credit score is less than or equal to the first probability threshold, it is determined that the risk control system has a "jump point" abnormality at the current moment, and the sum of the abnormality level of the risk control system at the previous moment and the first change is determined as the abnormality level of the risk control system at the current moment; if the probability of the first credit score is greater than the first probability threshold, it is determined that the risk control system does not have a "jump point" abnormality at the current moment, and the difference between the abnormality level of the risk control system at the previous moment and the second change is determined as the abnormality level of the risk control system at the current moment. If the abnormality level of the risk control system at the current moment is greater than or equal to the abnormality level threshold, it is determined that the risk control system has a systemic abnormality at the current moment, and the first alarm information is output.

[0114] like Figure 4 As shown, if the risk control system has a "jump point" anomaly at the current moment, the risk control system uses other risk control models that match the first user to perform a credit assessment on the first user or find other reasons to obtain the first user's new first credit score at the current moment; if the risk control system has a systemic anomaly at the current moment, the risk control model used by the risk control system at the current moment is repaired, and the credit assessment of the first user is performed again to obtain the first user's new first credit score at the current moment.

[0115] Afterwards, the risk control system can make decisions related to resource application services based on the first user's new first credit score at the current moment, such as whether to issue resources to the first user or whether to withdraw resources issued to the first user.

[0116] The inventors conducted experiments on the monitoring method of the risk control system proposed in the embodiments of this application and obtained experimental results on "jump point" anomalies and systemic anomalies. Among them, systemic anomalies include drift anomalies, data fluctuation anomalies, and failure anomalies. The details are as follows:

[0117] The experimental results of drift anomaly are as follows Figure 5As shown, at the 21st time step, that is, the 21st moment, the credit drift output by the risk control system exceeds the preset confidence range, and the early warning program for the risk control system is activated; at the 25th time step, the probability of the credit output by the risk control system is less than or equal to the first probability threshold, and the monitoring system outputs the first alarm information to prompt that there is a systemic abnormality in the risk control system.

[0118] It can be seen that the credit range remains stable when the credit output by the risk control system is abnormal. This is because the influence of abnormal credit is eliminated and the accuracy of the PF algorithm is maintained.

[0119] The experimental results of abnormal data fluctuations are as follows Figure 6 As shown, at the 19th time step, the credit fluctuation of the risk control system output exceeds the preset confidence range, and the early warning program for the risk control system is activated. At the 22nd time step, the credit fluctuation of the risk control system output returns to the confidence range, and the abnormality level of the risk control system (i.e., the early warning indicator) begins to decline. However, the difference is that the abnormality level decreases slowly due to the slow-down mechanism, while the abnormality level without the slow-down mechanism decreases rapidly. As a result, the abnormality level of the former reaches the threshold and begins to output the first alarm information 10 time steps after the early warning program begins, while the latter does not output the first alarm information until 17 time steps after the early warning program begins. Therefore, the abnormality level with the slow-down mechanism shortens the alarm time of systemic abnormalities.

[0120] It can be seen that the confidence range remains stable when the credit output of the risk control system fluctuates. This is also because the influence of abnormal credit is eliminated and the accuracy of the PF algorithm is maintained.

[0121] The experimental results of failure anomalies are as follows Figure 7 As shown in the figure, when the credit score output by the risk control system fails at the 21st time step and exceeds the credit confidence range, the early warning program for the risk control system is activated. At the 22nd time step, the credit score output by the risk control system returns to the credit confidence range, and the abnormality level begins to decrease. At the 7th time step after the early warning program is activated, the monitoring system outputs the first alarm information, indicating that there is a systemic abnormality in the risk control system.

[0122] It can be seen that the confidence range remains stable when the credit output by the risk control system fails. This is also because the influence of abnormal credit is discarded and the accuracy of the PF algorithm is maintained.

[0123] The experimental results of the risk control system showing abnormal “jump points” are as follows: Figure 8As shown in the figure, when the credit score output by the risk control system "jumps" at the 22nd time step, the "jump" exceeds the confidence range, and the early warning program for the risk control system is activated. Subsequently, because no "jumps" occur again, the abnormality level of the risk control system decays to 0. At the 36th time step, the credit score output by the risk control system "jumps" again, and the early warning program for the risk control system is activated. Subsequently, due to the emergence of a new "jump", the abnormality level of the risk control system decays to 0. This shows that the "jump" anomaly of the risk control system is accurately distinguished from systemic anomalies, and the confidence range remains stable even when the risk control system "jumps". This is also because the influence of abnormal credit scores is eliminated, maintaining the accuracy of the PF algorithm.

[0124] The foregoing description of this specification describes specific embodiments. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims can be performed in an order different from that described in the embodiments and still achieve the desired results. Furthermore, the processes depicted in the accompanying drawings do not necessarily require the specific order shown or the sequential order to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0125] Based on the same inventive concept, the embodiment of the present application also provides a monitoring device for a wind control system. Figure 9 , is a schematic structural diagram of a monitoring device 900 for a risk control system provided in one embodiment of the present application, the device 900 comprising:

[0126] The first prediction module 910 is configured to, in response to receiving a first credit score output by the risk control system, predict a third credit score of the first user at a second moment based on the second credit score of the first user at the first moment; the first credit score is obtained by evaluating the first user based on the user characteristics of the first user at the second moment, and the second moment is located after the first moment.

[0127] The first determining module 920 is configured to determine the probability of the first creditworthiness based on the third creditworthiness.

[0128] The second determination module 930 is configured to determine the abnormality level of the risk control system at the second moment based on the probability.

[0129] The output module 940 is configured to output a first warning message if the abnormality level is greater than or equal to an abnormality level threshold.

[0130] In another embodiment, the second determination module is used to: if the probability is less than or equal to a first probability threshold, determine the sum of the abnormality level of the risk control system at the first moment and the first change as the abnormality level of the risk control system at the second moment; if the probability is greater than the first probability threshold, determine the difference between the abnormality level of the risk control system at the first moment and the second change as the abnormality level of the risk control system at the second moment, and the second change is less than the first change.

[0131] In another embodiment, the first determination module is used to: determine a fourth credit level of the first user at the second moment based on the third credit level and the measurement noise at the second moment; determine a fifth credit level from the fourth credit level, wherein the fifth credit level is less than or equal to the first credit level; and determine a probability of the first credit level based on a ratio between the number of the fifth credit level and the number of the fourth credit level.

[0132] In another embodiment, when determining the probability of the first credit level based on the ratio between the fifth credit level and the fourth credit level, the first determining module performs the following steps:

[0133] If the first creditworthiness is less than or equal to the expectation of the fourth creditworthiness, determining the ratio as the probability of the first creditworthiness;

[0134] If the first creditworthiness is greater than the expectation of the fourth creditworthiness, the difference between the ratio threshold and the ratio is determined as the probability of the first creditworthiness.

[0135] In another embodiment, the apparatus further comprises:

[0136] a correction module, configured to correct the first credit rating based on the probability to obtain a sixth credit rating if the probability is less than or equal to a first probability threshold;

[0137] The sampling module is configured to sample the third credit score based on the difference between the sixth credit score and the fourth credit score, so as to predict the possible credit score of the first user at a third moment, where the third moment is after the second moment.

[0138] In another embodiment, the correction module is configured to: determine the weight of the first creditworthiness and the weight of the expectation based on the magnitude relationship between the expectation of the fourth creditworthiness and the first creditworthiness; and determine a weighted sum of the first creditworthiness and the expectation based on the weight of the first creditworthiness and the weight of the expectation to obtain a sixth creditworthiness.

[0139] In another embodiment, when the correction module determines the weight of the first creditworthiness and the expected weight based on the size relationship between the expectation of the fourth creditworthiness and the first creditworthiness, it performs the following steps: if the first creditworthiness is less than or equal to the expectation, the probability is determined as the weight of the first creditworthiness, and the difference between the second probability threshold and the probability is determined as the expected weight; if the first creditworthiness is greater than the expectation, the difference between the second probability threshold and the probability is determined as the expected weight, and the probability is determined as the expected weight.

[0140] In another embodiment, the sampling module is configured to: determine a weight of the third credit rating based on a difference between the sixth credit rating and the fourth credit rating; delete third credit ratings whose weights are less than or equal to a weight threshold, and copy third credit ratings whose weights are greater than or equal to the weight threshold, to obtain a sampling result.

[0141] In another embodiment, the third credit rating is predicted by the first model based on the second credit rating;

[0142] The first model is trained in the following manner: using the first model, based on the second user's seventh credit score at the fourth moment, predicting the second user's eighth credit score at the fifth moment; the fifth moment and the fourth moment are moments before the first moment, and the fourth moment is before the fifth moment, and the seventh credit score is obtained by the risk control system evaluating the second user based on the user characteristics of the second user at the fourth moment; based on the difference between the eighth credit score and the second user's ninth credit score at the fifth moment, adjusting the parameters of the first model, and the ninth credit score is obtained by the risk control system evaluating the second user based on the user characteristics of the second user at the fifth moment.

[0143] Obviously, the monitoring device of the risk control system provided in the embodiment of the present application can be used as Figure 2 The execution subject of the monitoring method of the risk control system shown is, for example Figure 2 In the monitoring method of the risk control system shown in FIG, step S202 can be performed by Figure 9 The first prediction module 910 in the monitoring device of the risk control system shown in FIG. 1 is executed, and step S204 can be performed by Figure 9 The first determination module 920 in the monitoring device of the risk control system shown in FIG. 1 is executed, and step S206 can be performed by Figure 9 The second determining module 930 in the monitoring device of the risk control system shown in FIG. 1 is executed, and step S208 can be performed by Figure 9 The output module 940 in the monitoring device of the wind control system shown is executed.

[0144] According to another embodiment of the present application, Figure 9 The various modules in the monitoring device of the risk control system shown can be individually or completely combined into one or several other modules to form a whole, or one (or some) of the modules can be further divided into multiple functionally smaller modules to form a whole, which can achieve the same operation without affecting the realization of the technical effects of the embodiments of the present application. The above-mentioned units are divided based on logical functions. In actual applications, the functions of one module can also be implemented by multiple modules, or the functions of multiple modules can be implemented by one module. In the embodiments of the present application, the monitoring device of the risk control system can also include other modules. In actual applications, these modules can also be implemented with the assistance of other modules, and can be implemented by the collaboration of multiple modules.

[0145] According to another embodiment of the present application, the system can be executed on a general computing device such as a computer including processing elements such as a CPU, RAM, ROM and storage elements. Figure 2 A computer program (including program code) for each step of the corresponding method shown in FIG. Figure 9 The computer program can be recorded on a computer-readable storage medium, for example, and transferred to an electronic device through the computer-readable storage medium and run therein.

[0146] Figure 10 This is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 10 At the hardware level, the electronic device includes a processor and, optionally, an internal bus, a network interface, and memory. The memory may include internal memory, such as high-speed random-access memory (RAM), or non-volatile memory, such as at least one disk drive. Of course, the electronic device may also include other hardware required for its services.

[0147] The processor, network interface, and memory can be interconnected via an internal bus, which can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 10Only one bidirectional arrow is used in the diagram, but this does not mean that there is only one bus or one type of bus.

[0148] The memory is used to store programs. Specifically, the program may include program code, which includes computer operating instructions. The memory may include internal memory and non-volatile memory, and provides instructions and data to the processor.

[0149] The processor reads the corresponding computer program from the non-volatile memory into the internal memory and then runs it, forming a monitoring device for the risk control system at the logical level. The processor executes the program stored in the memory and is specifically used to perform the following operations:

[0150] In response to receiving a first credit score output by the risk control system, predicting a third credit score of the first user at a second moment based on the second credit score of the first user at a first moment; the first credit score being obtained by evaluating the first user based on user characteristics of the first user at the second moment, the first moment being before the second moment;

[0151] determining a probability of the first creditworthiness based on the third creditworthiness;

[0152] determining, based on the probability, a degree of abnormality of the risk control system at the second moment;

[0153] If the abnormality level is greater than or equal to the abnormality level threshold, a first alarm message is output.

[0154] The above application Figure 2The methods performed by the monitoring device of the risk control system disclosed in the illustrated embodiments can be applied to or implemented by a processor. The processor may be an integrated circuit chip with signal processing capabilities. During implementation, each step of the above method can be completed by hardware integrated logic circuits in the processor or by software instructions. The above processor can be a general-purpose processor, including a central processing unit (CPU), a network processor (NP), etc.; it can also be a digital signal processor (DSP), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. The methods, steps, and logic block diagrams disclosed in the embodiments of this application can be implemented or executed. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly implemented and executed by a hardware decoding processor, or by a combination of hardware and software modules in the decoding processor. The software module can be located in a storage medium well-known in the art, such as random access memory, flash memory, read-only memory, programmable read-only memory, electrically erasable programmable memory, registers, etc. The storage medium is located in the memory, and the processor reads the information in the memory and, in conjunction with its hardware, completes the steps of the above method.

[0155] The electronic device may also perform Figure 2 Method, and realize the monitoring device of the risk control system Figure 2 、 Figure 3 and Figure 4 The functions of the illustrated embodiment will not be described in detail in the embodiments of the present application.

[0156] Of course, in addition to software implementation, the electronic device of this application does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc. That is to say, the execution subject of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0157] The embodiment of the present application also provides a computer-readable storage medium, which stores one or more programs, wherein the one or more programs include instructions, which, when executed by an electronic device including multiple application programs, can enable the electronic device to execute Figure 2 The method of the embodiment shown is specifically used to perform the following operations:

[0158] In response to receiving a first credit score output by the risk control system, predicting a third credit score of the first user at a second moment based on the second credit score of the first user at a first moment; the first credit score being obtained by evaluating the first user based on user characteristics of the first user at the second moment, the first moment being before the second moment;

[0159] determining a probability of the first creditworthiness based on the third creditworthiness;

[0160] determining, based on the probability, a degree of abnormality of the risk control system at the second moment;

[0161] If the abnormality level is greater than or equal to the abnormality level threshold, a first alarm message is output.

[0162] An embodiment of the present application also proposes a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute some or all of the steps in the monitoring method of the risk control system proposed in the embodiment of the present application.

[0163] In short, the above description is only a preferred embodiment of the present application and is not intended to limit the scope of protection of the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0164] The systems, devices, modules, or units described in the above embodiments may be implemented by computer chips or entities, or by products having certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smartphone, 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.

[0165] Computer-readable media includes permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. 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, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media (transitory media), such as modulated data signals and carrier waves.

[0166] It should also be noted that the terms "comprises," "includes," or any other variations thereof are intended to encompass non-exclusive inclusion, such that a process, method, commodity, or apparatus that includes a series of elements includes not only those elements but also other elements not explicitly listed, or includes elements inherent to such process, method, commodity, or apparatus. In the absence of further limitations, an element defined by the phrase "comprises a ..." does not exclude the presence of other identical elements in the process, method, commodity, or apparatus that includes the element.

[0167] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

Claims

1. A monitoring method for a risk control system, characterized in that: include: In response to receiving a first credit score output by the risk control system, predicting a third credit score of the first user at a second moment based on the second credit score of the first user at a first moment; the first credit score is obtained by the risk control system through an assessment of the first user based on the user characteristics of the first user at the second moment, the first moment being before the second moment; determining a probability of the first creditworthiness based on the third creditworthiness; determining, based on the probability, a degree of abnormality of the risk control system at the second moment; If the abnormality level is greater than or equal to the abnormality level threshold, a first alarm message is output.

2. The method according to claim 1, characterized in that Determining the abnormality level of the risk control system at the second moment based on the probability includes: If the probability is less than or equal to a first probability threshold, determining the sum of the abnormality level of the risk control system at the first moment and the first change as the abnormality level of the risk control system at the second moment; If the probability is greater than the first probability threshold, the difference between the abnormality level of the risk control system at the first moment and the second change amount is determined as the abnormality level of the risk control system at the second moment, and the second change amount is less than the first change amount.

3. The method according to claim 1, characterized in that The determining, based on the third creditworthiness, the probability of the first creditworthiness includes: determining a fourth credit score of the first user at the second moment based on the third credit score and the measurement noise at the second moment; determining a fifth credit rating from the fourth credit rating, the fifth credit rating being less than or equal to the first credit rating; The probability of the first degree of credit is determined based on a ratio between the number of the fifth degree of credit and the number of the fourth degree of credit.

4. The method according to claim 3, characterized in that The determining the probability of the first credit level based on the ratio between the number of the fifth credit level and the number of the fourth credit level includes: If the first creditworthiness is less than or equal to the expectation of the fourth creditworthiness, determining the ratio as the probability of the first creditworthiness; If the first creditworthiness is greater than the expectation of the fourth creditworthiness, the difference between the ratio threshold and the ratio is determined as the probability of the first creditworthiness.

5. The method according to claim 3, characterized in that After determining the probability of the first credit level based on the ratio between the number of the fifth credit level and the number of the fourth credit level, the method further includes: If the probability is less than or equal to a first probability threshold, modifying the first credit rating based on the probability to obtain a sixth credit rating; Based on the difference between the sixth credit score and the fourth credit score, the third credit score is sampled for predicting the possible credit score of the first user at a third moment, where the third moment is after the second moment.

6. The method according to claim 5, characterized in that The step of modifying the first credit level based on the probability to obtain a sixth credit level includes: determining a weight of the first creditworthiness and a weight of the expectation based on a magnitude relationship between the expectation of the fourth creditworthiness and the first creditworthiness; Based on the weight of the first credit level and the expected weight, a weighted sum of the first credit level and the expected value is determined to obtain a sixth credit level.

7. The method according to claim 6, characterized in that The determining of the weight of the first creditworthiness and the weight of the expectation based on the magnitude relationship between the expectation of the fourth creditworthiness and the first creditworthiness includes: If the first creditworthiness is less than or equal to the expectation, determining the probability as a weight of the first creditworthiness, and determining a difference between a second probability threshold and the probability as a weight of the expectation; If the first credit is greater than the expectation, the difference between the second probability threshold and the probability is determined as the weight of the expectation, and the probability is determined as the weight of the expectation.

8. The method according to claim 5, characterized in that The sampling of the third credit level based on the difference between the sixth credit level and the fourth credit level includes: determining a weight of the third credit rating based on a difference between the sixth credit rating and the fourth credit rating; The third credits with weights less than or equal to the weight threshold are deleted, and the third credits with weights greater than or equal to the weight threshold are copied to obtain a sampling result.

9. The method according to claim 1, characterized in that The third credit rating is predicted by the first model based on the second credit rating; The first model is trained in the following way: using the first model, predicting the eighth credit score of the second user at a fifth moment based on the seventh credit score of the second user at a fourth moment; the fifth moment and the fourth moment are moments before the first moment, and the fourth moment is before the fifth moment; and the seventh credit score is obtained by evaluating the second user based on the user characteristics of the second user at the fourth moment; The parameters of the first model are adjusted based on a difference between the eighth credit score and a ninth credit score of the second user at the fifth moment, wherein the ninth credit score is obtained by evaluating the second user based on the user characteristics of the second user at the fifth moment.

10. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the monitoring method of the risk control system according to any one of claims 1 to 9.

11. A computer-readable storage medium, characterized in that When the instructions in the storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the monitoring method of a risk control system according to any one of claims 1 to 9.

12. A computer program product, characterized in that The computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute part or all of the steps in the monitoring method of the risk control system according to any one of claims 1 to 9.

Images