CN117436966A - Credit qualification detection method, credit risk model determination method, device and equipment - Google Patents
Credit qualification detection method, credit risk model determination method, device and equipment Download PDFInfo
- Publication number
- CN117436966A CN117436966A CN202310930862.3A CN202310930862A CN117436966A CN 117436966 A CN117436966 A CN 117436966A CN 202310930862 A CN202310930862 A CN 202310930862A CN 117436966 A CN117436966 A CN 117436966A
- Authority
- CN
- China
- Prior art keywords
- updating
- credit
- risk model
- optimal position
- credit risk
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000012797 qualification Methods 0.000 title claims abstract description 118
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000001514 detection method Methods 0.000 title claims abstract description 37
- 239000002245 particle Substances 0.000 claims abstract description 179
- 230000008859 change Effects 0.000 claims description 9
- 238000005516 engineering process Methods 0.000 abstract description 5
- 238000012549 training Methods 0.000 description 28
- 230000008569 process Effects 0.000 description 17
- 238000004590 computer program Methods 0.000 description 11
- 238000005457 optimization Methods 0.000 description 9
- 238000012804 iterative process Methods 0.000 description 8
- 238000012545 processing Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 230000001149 cognitive effect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000006870 function Effects 0.000 description 5
- 238000011478 gradient descent method Methods 0.000 description 4
- 238000003062 neural network model Methods 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000012552 review Methods 0.000 description 3
- 238000013527 convolutional neural network Methods 0.000 description 2
- 230000000875 corresponding effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000009514 concussion Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000011664 signaling Effects 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0609—Buyer or seller confidence or verification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
- G06N3/0442—Recurrent networks, e.g. Hopfield networks characterised by memory or gating, e.g. long short-term memory [LSTM] or gated recurrent units [GRU]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
Abstract
The present disclosure relates to the technical field of financial science and technology, and in particular, to a credit qualification detection method, a credit risk model determination device, and a credit risk model determination device. The method comprises the following steps: detecting the credit qualification of the platform through a credit risk model according to the qualification data of the platform; the credit risk model is obtained by: determining the fitness of particles in the particle swarm; updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness; updating inertia factors and learning factors according to the current iteration times; updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor; updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks. The embodiment of the specification can accurately and efficiently detect the credit qualification of the third party platform.
Description
Technical Field
The present disclosure relates to the technical field of financial science and technology, and in particular, to a credit qualification detection method, a credit risk model determination device, and a credit risk model determination device.
Background
With the development of information technology, financial institutions may cooperate with third party platforms in order to increase the level of service. Thus, the customer can initiate a transaction on the third party platform and flow to the business system of the financial institution for processing. The third party platform connects the financial institution and the customer, and plays a vital role in transaction submitting, processing and forwarding stages.
When selecting a cooperating third party platform, the credit worthiness of the third party platform needs to be assessed to avoid the occurrence of credit risk. The third party platform has more qualification related information, is mixed and scattered in information channel distribution, and the platform is selected without unified standard. In the related art, the credit worthiness of the third party platform is often reviewed empirically by an expert. However, in the expert review mode, a plurality of links of review processes are needed, so that the labor cost is high and the efficiency is low. Moreover, the expert review mode is adopted, so that influence of subjective factors is easy to occur, and the credit qualification assessment result is inaccurate.
Disclosure of Invention
The embodiment of the specification provides a credit qualification detection method, a credit risk model determination device and credit qualification detection equipment, so as to accurately and efficiently detect the credit qualification of a third party platform. The technical solutions of the embodiments of the present specification are as follows.
The embodiment of the specification provides a credit qualification detection method, which comprises the following steps:
detecting the credit qualification of the platform through a credit risk model according to the qualification related data of the platform;
wherein the credit risk model is obtained by:
determining fitness of particles in a particle swarm, wherein the particles have positions and speeds, the positions represent parameters of a credit risk model, the speeds represent the variation degree of the parameters, and the fitness represents the quality degree of the parameters;
updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
updating inertia factors and learning factors according to the current iteration times;
updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks.
The embodiment of the specification provides a credit risk model determining method, which comprises the following steps:
determining fitness of particles in a particle swarm, wherein the particles have positions and speeds, the positions represent parameters of a credit risk model, the speeds represent the variation degree of the parameters, and the fitness represents the quality degree of the parameters;
updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
updating inertia factors and learning factors according to the current iteration times;
updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks.
The embodiment of the specification also provides a credit qualification detection device, which comprises:
the detection unit is used for detecting the credit qualification of the platform through a credit risk model according to the qualification related data of the platform;
wherein the credit risk model is obtained by:
determining fitness of particles in a particle swarm, wherein the particles have positions and speeds, the positions represent parameters of a credit risk model, the speeds represent the variation degree of the parameters, and the fitness represents the quality degree of the parameters;
updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
updating inertia factors and learning factors according to the current iteration times;
updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks.
The embodiment of the specification also provides a credit risk model determining device, which comprises:
a determining unit, configured to determine fitness of particles in a particle swarm, where the particles have a position and a velocity, the position represents a parameter of a credit risk model, the velocity represents a degree of change of the parameter, and the fitness represents a degree of merit of the parameter;
a first updating unit for updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
the second updating unit is used for updating the inertia factor and the learning factor according to the current iteration times;
the third updating unit is used for updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
and the fourth updating unit is used for updating parameters of the credit risk model constructed according to the global optimal position through the qualification related data of the sample platform and a label, wherein the label is used for indicating whether the credit qualification of the sample platform has risks.
The embodiment of the specification also provides a computer device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the credit qualification detection method and the credit risk model determination method when executing the computer program.
Embodiments of the present description may employ Particle Swarm Optimization (PSO) to improve the training process of the credit risk model. The qualification data of the platform often has the characteristics of high dimensionality, high noise, nonlinearity and the like. By combining the particle swarm optimization algorithm with the credit risk model training process, on one hand, the credit risk model can have higher convergence rate, and on the other hand, the fitting capacity of the credit risk model to high-latitude, high-noise and nonlinear qualification related data can be improved, so that the credit risk model obtained through training can have higher detection precision. In addition, the inertia factor and the learning factor are updated by the number of current iterations. So that the inertia factor and the learning factor can adaptively change following the change in the number of iterations. And the global searching capability of the particle swarm optimization algorithm is effectively improved, and the convergence speed and convergence accuracy of the particle swarm are also enhanced.
Drawings
In order to more clearly illustrate the embodiments of the present description or the solutions in the prior art, the drawings that are required for the embodiments or the description of the prior art will be briefly described, the drawings in the following description are only some embodiments described in the present description, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a flow chart of a training method of a credit risk model according to an embodiment of the present disclosure;
FIG. 2 is a flow chart of a training method of a credit risk model according to an embodiment of the present disclosure;
FIG. 3 is a flowchart of a training method of a credit risk model according to an embodiment of the present disclosure;
FIG. 4 is a flow chart of a method for detecting credit worthiness in an embodiment of the present disclosure;
FIG. 5 is a schematic structural diagram of a training device for a risk model for credit in an embodiment of the disclosure;
fig. 6 is a schematic structural diagram of a signaling qualification detecting device according to an embodiment of the present disclosure.
Detailed Description
The technical solutions of 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 apparent that the described embodiments are only some embodiments of the present specification, not all embodiments. The specific embodiments described herein are to be considered in an illustrative rather than a restrictive sense. All other embodiments derived by a person of ordinary skill in the art based on the described embodiments of the present disclosure fall within the scope of the present disclosure. In addition, relational terms such as "first" and "second", and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions.
The third party platform may comprise a SAAS (Software-as-a-Service) platform. The cloud interconnection of the financial institutions refers to embedding services provided by the financial institutions into various SAAS platforms through cloud interconnection technology to meet the requirements of SAAS cloud service clients for using services provided by the financial institutions, so as to provide comprehensive financial services for the clients.
The data related to the credit qualification detection scenario is qualification related data. The qualification related data has the characteristics of high dimensionality, high noise, nonlinearity and the like. And credit risk models such as BP neural network model, convolutional neural network model and the like can be adopted to detect and evaluate the credit qualification of the platform. In the related technology, a gradient descent method can be adopted to train the credit risk model, and the credit qualification of the platform is detected through the trained credit risk model. However, due to the influence of high dimensionality, high noise and nonlinearity, the parameters of the credit risk model are easy to fall into local minima in the training process, so that the fitting capability of the credit risk model on high latitude, high noise and nonlinearity qualification related data is poor. And the training process converges slowly, possibly causing concussion or divergence. If the Particle Swarm Optimization (PSO) can be combined with the training process of the credit risk model, the searching range of the credit risk model parameters in the training process can be enlarged, the problem that the parameters fall into local minima due to the influence of high latitude, high noise and nonlinearity in the training process is avoided, and the fitting capacity of the credit risk model to the high latitude, high noise and nonlinearity qualification related data is improved. Thereby improving the training process of the credit risk model and avoiding the problems in the training process.
Please refer to fig. 1. The embodiment of the specification provides a method for determining a credit risk model, which can be applied to computer equipment such as a server, a server cluster or a personal computer. The method may comprise the following steps.
Step 11: the fitness of particles in a particle population is determined, the particles have positions and speeds, the positions represent parameters of a credit risk model, the speeds represent the change degree of the parameters, and the fitness represents the quality degree of the parameters.
In some embodiments, the credit risk model is used to perform a detection assessment of credit worthiness of a third party platform. The third party platform may include a SAAS platform or the like. The credit risk model may include a neural network model, a support vector machine model, and the like. The neural network model may include a BP neural network model, a convolutional neural network model, or the like.
In some embodiments, the credit risk model is used for detection from qualification-related data of the platform. The qualification-related data may include at least one sub-qualification data for representing credit qualification of the platform from at least one dimension. For example, the qualification related data may include a plurality of sub-qualification data as shown in table 1 below.
TABLE 1
In some embodiments, the credit risk model may have a plurality of parameters. The parameters include parameters for processing qualification related data. For example, the parameters include weights and thresholds for processing the qualification related data, etc.
In some embodiments, the population of particles may include a plurality of particles therein. The number of particles can be flexibly set according to the needs. For example, if the parameters of the credit risk model are required to have a wider range of value searches, the number of particles can be set to be larger, but the training efficiency is slower. If faster training efficiency is required, the particle number can be set smaller, but the value search range of the credit risk model parameter is smaller. In practice, the particle number can be comprehensively determined according to the value searching range and the training efficiency.
Each particle of the population of particles may have a location. The location is used to represent parameters of a credit risk model. For example, the location may include a location vector, the number of data elements in the location vector may be equal to the number of parameters of the credit risk model, and each data element in the location vector may correspond to a parameter, and may specifically be a value of the parameter. Thus, the location of each particle may be understood as a combination of values of a plurality of parameters of the credit risk model, and the location of a plurality of particles may be understood as a combination of values of a plurality of parameters of the credit risk model. Each particle of the population of particles may also have a velocity. The speed is used for representing the variation degree of the parameter, and the variation degree can comprise the variation speed degree of the parameter value. For example, the speed may include a speed vector, the number of data elements in the speed vector may be equal to the number of parameters of the credit risk model, and each data element in the speed vector may correspond to a parameter and may be used to represent how fast a certain value of the parameter changes. The value may comprise a data element in the position vector of the particle corresponding to the parameter.
For example, the credit risk model has 20 parameters. The particle group comprises 100 particles. The ith particle in the particle group may have a position vector x i And velocity vector v i . Position vector x i And velocity vector v i There may be 20 data elements, respectively. Position vector x i The j-th data element x in (2) ij A value representing the j-th parameter of the credit risk model, velocity vector v i The j-th data element v in (1) ij Value x representing jth parameter of credit risk model ij Is a degree of variation of (a).
In some embodiments, the fitness is used to represent the goodness of the parameter. The fitness is used for representing the goodness of a value combination represented by the position vector, and the goodness can comprise the accuracy of credit risk model on credit qualification detection. The magnitude of the fitness may be positively correlated with the degree of merit.
The fitness may be determined from the location of the particles. The fitness can be determined by an objective function based on the position of the particles. For example, the position of the particle may be substituted into the objective function, and the value of the objective function may be obtained as the fitness. Alternatively, a credit risk model may also be constructed based on the location of the particles; the performance index of the constructed credit risk model may be obtained as fitness. The performance indicators may include Precision (Accuracy), recall (Recall), precision (Precision), F1 Score (F1-Score), and the like. For example, the qualification related data of the test platform can be acquired, and the qualification related data of the test platform can be input into the constructed credit risk model to obtain a credit qualification detection result; and determining the performance index of the constructed credit risk model according to the label of the test platform and the credit qualification detection result. The tag is used to indicate whether the credit worthiness of the test platform is at risk.
Step 12: and updating the individual optimal positions of the particles and the global optimal positions of the particle swarm according to the fitness.
In some embodiments, the method of training the credit risk model may include one or more iterative processes. Each iteration process may record an individual optimal position of the particles and a global optimal position of the particle swarm. Each particle has at least one historical position prior to the current iterative process. The individual optimal positions of the particles include historical positions of the particles with optimal fitness prior to the current iterative process. The global optimal position of the particle swarm comprises an individual optimal position with optimal fitness in the particle swarm.
It should be noted that, in the first iteration, the position and the velocity of the particles may be preset. For example, the position and velocity of the particles may be set randomly. In addition, during the training of the credit risk model, the position of each particle may be within a set position range. For example, the position of the particles may be less than or equal to x max Greater than or equal to-x max . The velocity of each particle may be within a set velocity range. For example, the velocity of the particles may be less than or equal to v max Greater than or equal to-v max 。
In some embodiments, each particle in a population of particles may have a location and an individual optimal location during the current iteration. The individual optimal positions include historical positions of the particles with optimal fitness prior to the current iterative process. The particle swarm may have a globally optimal position. The global optimal position comprises an individual optimal position with optimal fitness in the particle swarm.
The individual optimal position of each particle may be updated according to the fitness. Specifically, for each particle in the particle swarm, the fitness of the particle can be compared with the fitness of the optimal position of the individual; if the fitness of the particle is better than that of the individual optimal position, the position of the particle can be used as a new individual optimal position; if the fitness of the particles is inferior to that of the individual optimal positions, the individual optimal positions can be kept unchanged. Updating the individual optimal positions of the particles in the particle swarm; the global optimal position of the particle swarm can be updated according to the individual optimal fitness of each particle after being updated. Specifically, an individual optimum position having an optimum fitness may be selected from individual optimum fitness of each particle; the fitness of the selected individual optimal position can be compared with the fitness of the global optimal position; if the fitness of the selected individual optimal position is better than that of the global optimal position, the selected individual optimal position can be used as a new global optimal position; if the fitness of the selected individual optimal position is inferior to that of the global optimal position, the global optimal position can be kept unchanged.
Step 13: and updating the inertia factor and the learning factor according to the current iteration times.
In some embodiments, the inertia factor and the learning factor are updated by depending on the current number of iterations. So that the inertia factor and the learning factor can adaptively change following a change in the number of iterations. And the global searching capability of the particle swarm optimization algorithm is effectively improved, and the convergence speed and convergence accuracy of the particle swarm are also enhanced.
Formulas for updating particle velocity include memory terms and cognitive terms, including self-cognitive terms and population-cognitive terms. For example, the formula for updating the particle velocity includes Representing memory items->Representing self-cognition items->Representing a population cognitive term. The inertia factor is used for adjusting the proportion of the memory term, c 1 For adjusting the proportion of the self-cognition item, c 2 Is used for adjusting the proportion of the cognitive items of the population.
In practical application, the ratio between the current iteration number and the maximum iteration number can be obtained; the inertia factor and the learning factor may be updated based on the ratio. The maximum number of iterations may be preset. For example, the formula can be based on Updating an inertia factor; can be->Andthe learning factor is updated. Wherein w represents an updated inertia factor, c 1 And c 2 Represents the updated learning factor, g represents the current iteration number, g max Represents the maximum number of iterations, a, b, d, e 1 、e 2 、f 1 、f 2 Representing the parameters.
The inertia factor w may thus increase with increasing number of iterations. Specifically, when the iteration times are smaller, the inertia factor w is smaller, the proportion of the memory term is smaller when the particle speed is updated, and the updating amplitude of the particle speed is larger. When the iteration times are larger, the inertia factor w is larger, the proportion of the memory term is larger when the particle speed is updated, and the updating amplitude of the particle speed is smaller. The particle speed is updated greatly firstly and then updated slightly, so that the convergence speed of the particle swarm is improved. And is also provided withThe convergence speed is further increased in the form of a power function. e, e 1 -e 2 May be negative. Thus learn factor c 1 Can be reduced with the increase of the iteration times. When the iteration times are smaller, the self-cognition item accounts for larger proportion when the particle speed is updated. When the iteration times are large, the proportion of self-cognition items is small when the particle speed is updated. f (f) 1 -f 2 May be a positive number. Thus learn factor c 2 May increase with increasing number of iterations. And when the iteration times are smaller, the proportion of the group cognitive items is smaller when the particle speed is updated. And when the iteration times are larger, the proportion of the group cognitive items is larger when the particle speed is updated. Considering that the difference between the global optimal position and the true optimal position of the particle population is larger in the early stage of the iterative process, and the difference between the global optimal position and the true optimal position of the particle population is smaller in the later stage of the iterative process. Thus learning factor c 1 Decreasing with increasing iteration number, learning factor c 2 The convergence speed of the particle swarm can be increased along with the increase of the iteration number, and the convergence accuracy can also be improved.
Step 14: and updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor.
In some embodiments, the position and velocity of each particle in the particle swarm may be further updated according to the updated individual optimal position, the global optimal position, the inertia factor, and the learning factor. For example, the velocity of each particle may be updated according to the updated individual optimum position, global optimum position, inertia factor, and learning factor by the following formula:k represents the current iteration number, k+1 represents the next iteration number, +.>Indicating the speed of the i-th particle after updating, < >>Indicating the speed of the ith particle, w indicating the inertia factor, which can be obtained in particular by step 13,/i>Represents the optimal position of the individual, gbest k Representing a global optimum position, c 1 And c 2 Representing learning factors, which can be obtained in step 13, r 1 And r 2 Representing a random number. The position of each particle can be updated according to the updated velocity by the following formula: />k represents the current iteration number, k+1 represents the next iteration number, +.>Indicating the speed of the i-th particle after updating, < >>Represents the position of the ith particle, +.>Indicating the updated position of the ith particle.
Step 15: updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks.
In some embodiments, a credit risk model may be constructed from the globally optimal location; the constructed credit risk model may be trained based on qualification data of the sample platform and the label of the sample platform. The tag is used to indicate whether credit qualification of the sample platform is at risk. Specifically, the qualification data of the sample platform can be input into the constructed credit risk model to obtain a credit qualification detection result; loss information can be calculated according to the credit qualification detection result and the label; parameters of the constructed credit risk model can be updated by a gradient descent method according to the loss information. The credit qualification detection result can be credit qualification or credit qualification. Of course, the credit worthiness detection result may be a certain credit worthiness level of a plurality of credit worthiness levels. Alternatively, the credit qualification test result may be a credit score. The larger the credit score, the better the credit qualification, and the smaller the probability of a credit risk. A smaller credit score indicates poorer credit qualification and a greater probability of occurrence of credit risk.
In some embodiments, please refer to fig. 2. Steps 10-14 may be iteratively performed until the first set condition is met. The first setting condition may include that the number of iterations reaches a preset maximum number of iterations. After the first setting condition is satisfied, step 15 may be performed. Step 15 may further include a plurality of iterative processes, thereby facilitating training using qualification data of a plurality of sample platforms. The embodiment of fig. 1 as a whole may thus comprise 2 separate multiple iterative processes.
In some embodiments, please refer to fig. 3. The training method may further include step 16: and determining a new global optimal position of the particle swarm, wherein the new global optimal position can comprise parameters updated by the credit risk model. The new global optimal location may include parameters of the credit risk model after updating. That is, the updated parameters of the credit risk model may be used as new global optimal locations. Steps 11-16 may be iteratively performed until the second set condition is met. The second setting condition comprises that the iteration times reach the preset maximum iteration times. The corresponding embodiment of fig. 1 thus comprises 1 independent multiple iteration process.
The method for determining the credit risk model according to the embodiment of the specification can determine the fitness of particles in a particle swarm, wherein the particles have positions and speeds, the positions represent parameters of the credit risk model, the speeds represent the variation degree of the parameters, and the fitness represents the goodness of the parameters; the individual optimal position of the particles and the global optimal position of the particle swarm can be updated according to the fitness; the inertia factor and the learning factor can be updated according to the current iteration times; the position and the speed of the particles can be updated according to the individual optimal position, the global optimal position, the inertia factor and the learning factor; parameters of a credit risk model constructed according to the global optimal position can be updated through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks. Thus, particle Swarm Optimization (PSO) may be employed to improve the training process of the credit risk model. The qualification data of the platform often has the characteristics of high dimensionality, high noise, nonlinearity and the like. By combining the particle swarm optimization algorithm with the training process of the credit risk model, the searching range of the parameters of the credit risk model in the training process is enlarged, the problem that the parameters are trapped into local minima due to the influence of high latitude, high noise and nonlinearity in the training process can be avoided, the optimal parameters of the credit risk model can be obtained, and the fitting capacity of the credit risk model to high latitude, high noise and nonlinearity qualification data is improved. In addition, the inertia factor and the learning factor are updated by the number of current iterations. So that the inertia factor and the learning factor can adaptively change following a change in the number of iterations. And the global searching capability of the particle swarm optimization algorithm is effectively improved, and the convergence speed and convergence accuracy of the particle swarm are also enhanced.
Please refer to fig. 4. The embodiment of the specification also provides a credit qualification detection method. The method can be applied to computer equipment such as servers, server clusters or personal computers. The method may comprise the following steps.
Step 41: and detecting the credit qualification of the platform through a credit risk model according to the qualification related data of the platform.
In some embodiments, the qualification-related data may include at least one sub-qualification data for representing credit qualification of the platform from at least one dimension. The credit risk model may be obtained by:
determining fitness of particles in a particle swarm, wherein the particles have positions and speeds, the positions represent parameters of a credit risk model, the speeds represent the variation degree of the parameters, and the fitness represents the quality degree of the parameters;
updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
updating inertia factors and learning factors according to the current iteration times;
updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks.
In some embodiments, the qualification-related data may be input into a credit risk model to obtain a credit qualification detection result for the platform. The credit qualification detection result can be credit qualification or credit qualification. Of course, the credit worthiness detection result may be a certain credit worthiness level of a plurality of credit worthiness levels. Alternatively, the credit qualification test result may be a credit score. The larger the credit score, the better the credit qualification, and the smaller the probability of a credit risk. A smaller credit score indicates poorer credit qualification and a greater probability of occurrence of credit risk. For example, the credit risk model may includeWherein w is i Parameters representing a credit risk model, x i Representing sub-qualification data, n representing the number of sub-qualification data in the qualification-related data, i.e. the number of dimensions, θ representing a threshold value, f representing an activation function of the credit risk model, y representing a credit score.
According to the credit qualification detection method, the credit qualification detection result of the platform can be determined through the credit risk model according to the qualification data of the platform. The trained credit risk model has stronger adaptability to high-latitude, high-noise and nonlinear qualification data. The accuracy of credit qualification detection can be improved through the trained credit risk model.
In some scenario examples, particle swarm optimization algorithms (PSOs) may be combined with a training process of a credit risk model, such as a gradient descent method. The credit risk model obtained through training can detect and evaluate the credit qualification of the third party platform. Compared with a gradient descent method, the credit risk model has the advantages that on one hand, the credit risk model has higher convergence rate, and on the other hand, the fitting capability of the credit risk model to high-latitude, high-noise and nonlinear qualification related data can be improved, so that the credit risk model obtained through training can have higher detection precision.
Please refer to fig. 5. The embodiment of the specification also provides a credit risk model determining device, which comprises the following modules.
A determining unit 51 for determining fitness of particles in a population of particles, the particles having a position and a velocity, the position representing a parameter of the credit risk model, the velocity representing a degree of variation of the parameter, the fitness representing a degree of merit of the parameter;
a first updating unit 52 for updating the individual optimal positions of the particles and the global optimal positions of the particle groups according to the fitness;
a second updating unit 53, configured to update the inertia factor and the learning factor according to the current iteration number;
a third updating unit 54 for updating the position and speed of the particles according to the individual optimum position, the global optimum position, the inertia factor and the learning factor;
a fourth updating unit 55, configured to update parameters of the credit risk model constructed according to the global optimal position by using the qualification related data of the sample platform and a label, where the label is used to indicate whether the credit qualification of the sample platform has a risk.
Please refer to fig. 6. The embodiment of the specification also provides a credit qualification detection device, which comprises the following modules.
A detection unit 61, configured to detect credit qualification of the platform through a credit risk model according to the qualification related data of the platform;
wherein the credit risk model is obtained by:
determining fitness of particles in a particle swarm, wherein the particles have positions and speeds, the positions represent parameters of a credit risk model, the speeds represent the variation degree of the parameters, and the fitness represents the quality degree of the parameters;
updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
updating inertia factors and learning factors according to the current iteration times;
updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks.
The embodiment of the specification also provides a computer device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor realizes the credit qualification detection method and the credit risk model determination method when executing the computer program.
The embodiments of the present specification also provide a computer readable storage medium storing a computer program which, when executed by a processor, implements the above-described credit qualification detection method and credit risk model determination method.
The embodiments of the present specification also provide a computer program product comprising a computer program which, when executed by a processor, implements the credit qualification detection method and the credit risk model determination method described above.
Those skilled in the art will appreciate that the present description may be provided as a method, system, or computer program product. The description may thus take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present description can take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
The present description is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the specification. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. The computer may be a personal computer, a laptop computer, a cellular telephone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.
Each functional unit in the embodiments of the present disclosure may be integrated in one processing unit, or each functional unit may exist alone physically, or two or more functional units may be integrated in one processing unit.
Those skilled in the art will appreciate that the descriptions of various embodiments are provided herein with respect to each of the embodiments, and that reference may be made to the relevant descriptions of other embodiments for parts of one embodiment that are not described in detail. In addition, it will be appreciated that those skilled in the art, upon reading the present specification, may conceive of any combination of some or all of the embodiments set forth herein without any inventive effort, and that such combination is within the scope of the disclosure and protection of the present specification.
Although the present specification is depicted by way of example, it will be appreciated by those skilled in the art that the above examples are merely intended to aid in understanding the core ideas of the present specification. Those skilled in the art will appreciate numerous modifications and variations therefrom. It is intended that the appended claims cover such modifications and variations as fall within the true spirit of this present description.
Claims (10)
1. A method for credit worthiness detection, comprising:
detecting the credit qualification of the platform through a credit risk model according to the qualification related data of the platform;
wherein the credit risk model is obtained by:
determining fitness of particles in a particle swarm, wherein the particles have positions and speeds, the positions represent parameters of a credit risk model, the speeds represent the variation degree of the parameters, and the fitness represents the quality degree of the parameters;
updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
updating inertia factors and learning factors according to the current iteration times;
updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks.
2. A method for determining a credit risk model, comprising:
determining fitness of particles in a particle swarm, wherein the particles have positions and speeds, the positions represent parameters of a credit risk model, the speeds represent the variation degree of the parameters, and the fitness represents the quality degree of the parameters;
updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
updating inertia factors and learning factors according to the current iteration times;
updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks.
3. The method of claim 2, wherein the step of determining the fitness comprises:
constructing a credit risk model according to the positions of particles in the particle swarm;
and acquiring a performance index of a credit risk model constructed according to the particle position as the fitness of the particles.
4. The method of claim 2, wherein the step of updating the inertia factor and the learning factor comprises:
acquiring the ratio between the current iteration number and the maximum iteration number;
and updating the inertia factor and the learning factor according to the ratio.
5. The method of claim 4, wherein the step of updating the inertia factor and the learning factor comprises:
according to the formulaUpdating an inertia factor;
according to the formulaAnd->Updating the learning factor;
wherein w represents an updated inertia factor, c 1 And c 2 Represents the updated learning factor, g represents the current iteration number, g max Represents the maximum number of iterations, a, b, d, e 1 、e 2 、f 1 、f 2 Representing the parameters.
6. The method according to claim 2, wherein the method further comprises:
iteratively executing the steps of determining the fitness, updating the individual optimal position and the global optimal position, updating the inertia factors and the learning factors and updating the particle positions and the particle speeds until the first setting condition is met;
the step of updating the parameters of the credit risk model includes:
after the iteration is finished, the parameters of the credit risk model constructed according to the global optimal position are updated.
7. The method according to claim 2, wherein the method further comprises:
determining a new global optimal position of the particle swarm, wherein the new global optimal position comprises parameters updated by a credit risk model;
the steps of determining fitness, updating the individual optimal position and the global optimal position, updating the inertia factor and the learning factor, updating the particle position and the speed, and updating the parameters of the credit risk model are iteratively executed until the second setting condition is satisfied.
8. A credit worthiness detection apparatus, comprising:
the detection unit is used for detecting the credit qualification of the platform through a credit risk model according to the qualification related data of the platform;
wherein the credit risk model is obtained by:
determining fitness of particles in a particle swarm, wherein the particles have positions and speeds, the positions represent parameters of a credit risk model, the speeds represent the variation degree of the parameters, and the fitness represents the quality degree of the parameters;
updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
updating inertia factors and learning factors according to the current iteration times;
updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
updating parameters of a credit risk model constructed according to the global optimal position through qualification related data of the sample platform and a label, wherein the label is used for indicating whether credit qualification of the sample platform has risks.
9. A credit risk model determining apparatus, comprising:
a determining unit, configured to determine fitness of particles in a particle swarm, where the particles have a position and a velocity, the position represents a parameter of a credit risk model, the velocity represents a degree of change of the parameter, and the fitness represents a degree of merit of the parameter;
a first updating unit for updating the individual optimal position of the particles and the global optimal position of the particle swarm according to the fitness;
the second updating unit is used for updating the inertia factor and the learning factor according to the current iteration times;
the third updating unit is used for updating the position and the speed of the particles according to the individual optimal position, the global optimal position, the inertia factor and the learning factor;
and the fourth updating unit is used for updating parameters of the credit risk model constructed according to the global optimal position through the qualification related data of the sample platform and a label, wherein the label is used for indicating whether the credit qualification of the sample platform has risks.
10. A computer device, comprising:
a processor;
a memory for storing processor-executable instructions;
the processor implements the method of any of claims 1-7 by executing the instructions.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310930862.3A CN117436966A (en) | 2023-07-27 | 2023-07-27 | Credit qualification detection method, credit risk model determination method, device and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310930862.3A CN117436966A (en) | 2023-07-27 | 2023-07-27 | Credit qualification detection method, credit risk model determination method, device and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117436966A true CN117436966A (en) | 2024-01-23 |
Family
ID=89552266
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310930862.3A Pending CN117436966A (en) | 2023-07-27 | 2023-07-27 | Credit qualification detection method, credit risk model determination method, device and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117436966A (en) |
-
2023
- 2023-07-27 CN CN202310930862.3A patent/CN117436966A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10943186B2 (en) | Machine learning model training method and device, and electronic device | |
| CN111275546B (en) | Financial customer fraud risk identification method and device | |
| JP6546180B2 (en) | Get Network Subject's Social Relationship Type | |
| WO2018170454A2 (en) | Using different data sources for a predictive model | |
| CN110111113B (en) | Abnormal transaction node detection method and device | |
| JP2019519021A (en) | Performance model bad influence correction | |
| US11257019B2 (en) | Method and system for search provider selection based on performance scores with respect to each search query | |
| CN110995459B (en) | Abnormal object identification method, device, medium and electronic equipment | |
| CN110929799A (en) | Method, electronic device, and computer-readable medium for detecting abnormal user | |
| CN110348471B (en) | Abnormal object identification method, device, medium and electronic equipment | |
| CN110930218B (en) | Method and device for identifying fraudulent clients and electronic equipment | |
| WO2023103527A1 (en) | Access frequency prediction method and device | |
| US20210295231A1 (en) | Machine learning model training outliers | |
| WO2017071474A1 (en) | Method and device for processing language data items and method and device for analyzing language data items | |
| US10678821B2 (en) | Evaluating theses using tree structures | |
| WO2022017082A1 (en) | Method and apparatus for detecting false transaction orders | |
| CN110866698A (en) | Device for assessing service score of service provider | |
| CN113298121A (en) | Message sending method and device based on multi-data source modeling and electronic equipment | |
| US20210357699A1 (en) | Data quality assessment for data analytics | |
| US11798540B2 (en) | Systems and methods for dynamically updating machine learning models that provide conversational responses | |
| JP2022534160A (en) | Methods and devices for outputting information, electronic devices, storage media, and computer programs | |
| CN117436966A (en) | Credit qualification detection method, credit risk model determination method, device and equipment | |
| US11580094B2 (en) | Real-time anomaly determination using integrated probabilistic system | |
| CN114298825A (en) | Method and device for extremely evaluating repayment volume | |
| CN114049202A (en) | Operation risk identification method and device, storage medium and electronic equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |