Bank risk prediction method and device, equipment and medium

Abstract

The invention discloses a bank risk prediction method, apparatus and device, and a medium. The method comprises the steps of obtaining a basic data set of a multi-objective function for bank risk prediction; the multi-objective function comprises an objective function of a bank risk coefficient and an objective function of a bank operation index; the basic data set comprises a plurality of decision variables and a solution space corresponding to each decision variable; in an iteration process of bank risk prediction, performing simulation dynamic iteration on the multi-objective function based on the basic data set of the multi-objective function, determining an optimal decision variable of the multi-objective function and an optimal simulation solution set corresponding to the optimal decision variable, and outputting an optimal risk coefficient and an optimal operation index corresponding to the optimal simulation solution set; the simulation dynamic iteration is characterized in that simulation resources utilized in the k-th iteration process are dynamically adjusted based on a simulation result of the (k-1)-th iteration. According to the method, simulation resources can be saved in the iteration process of bank risk prediction, and waste of the simulation resources is avoided.

Description

Technical Field

[0001] The present application relates to the technical field of risk prediction, and in particular to a bank risk prediction method, device, equipment and medium. Background Art

[0002] As a core component of the financial system, banks play a key role, but they also face many risks, including credit risk, market risk, liquidity risk, etc. Risk prediction plays a vital role in bank management, and improper risk prediction may cause great losses to customers.

[0003] In recent years, although there is no shortage of research on measuring bank risks through objective functions, the simulation process and the optimization process are not closely integrated when solving the objective function. The two processes run almost independently. The optimization process only includes continuing to iterate the simulation results fed back by the simulation process, and the simulation process only includes the allocation of simulation resources for the current population plan iterated by the optimization, resulting in a waste of simulation resources. Summary of the invention

[0004] Based on the above problems, the present application provides a bank risk prediction method, device, equipment and medium, the purpose of which is to save simulation resources and avoid waste of simulation resources during the iterative process of bank risk prediction.

[0005] The embodiments of the present application disclose the following technical solutions:

[0006] The first aspect of the present application provides a bank risk prediction method, the method comprising:

[0007] Obtain a basic data set of a multi-objective function for bank risk prediction; the multi-objective function includes an objective function of a bank risk coefficient and an objective function of a bank operating indicator; the basic data set includes a plurality of decision variables and a solution space corresponding to each decision variable; the solution space includes a plurality of sets of simulation solutions for the decision variables; the simulation solution set includes a solution to an objective function of a bank risk coefficient and a solution to an objective function of a bank operating indicator;

[0008] In the iterative process of bank risk prediction, the multi-objective function is simulated and dynamically iterated based on the basic data set of the multi-objective function to determine the optimal decision variables of the multi-objective function and the optimal simulation solution set corresponding to the optimal decision variables, and the optimal risk coefficient and the optimal operating indicator corresponding to the optimal simulation solution set are output; the simulation dynamic iteration represents that the simulation resources used in the kth iteration process are dynamically adjusted based on the simulation results of the k-1th iteration, k≥2.

[0009] Optionally, in the iterative process of bank risk prediction, the multi-objective function is simulated and dynamically iterated based on the basic data set of the multi-objective function, the optimal decision variables of the multi-objective function and the optimal simulation solution set corresponding to the optimal decision variables are determined, and the optimal risk coefficient and the optimal operating indicator corresponding to the optimal simulation solution set are output, which specifically includes:

[0010] Let the value of k be 2;

[0011] Generate a supplemented population individual gene sequence of the kth iteration based on the simulation result of the k-1th iteration; the supplemented population individual gene sequence includes a supplemented gene; the supplemented gene includes a decision variable and the number of simulations after the supplementation corresponding to the decision variable; the simulation result is a risk coefficient and an operation index corresponding to a set of simulation solutions of the decision variable;

[0012] Performing simulation based on the supplemented population individual gene sequences of the k-th iteration and the multi-objective function to determine the parent population of the k-th iteration; the parent population includes parent individuals; the parent individuals include decision variables and simulation results;

[0013] Determine whether the parent population of the k-th iteration meets the iteration end condition. If so, use the decision variables of the parent individuals in the parent population of the k-th iteration as the optimal decision variables, use the simulation solution set of the decision variables of the parent individuals in the parent population of the k-th iteration as the optimal simulation solution set, and output the risk coefficient and operation index corresponding to the optimal simulation solution set as the optimal risk coefficient and the optimal operation index; the iteration end condition is that the number of iterations reaches the preset number of iterations or the parent population contains a decision variable and a simulation result;

[0014] If not, the initial population individual gene sequence of the k+1th iteration is determined based on the parent population of the kth iteration, the value of k is increased by 1, and the step of generating the supplementary population individual gene sequence of the kth iteration based on the simulation result of the k-1th iteration is returned.

[0015] Optionally, the generating of the supplementary population individual gene sequences of the kth iteration based on the simulation result of the k-1th iteration specifically includes:

[0016] Obtain the supplementary simulation times index corresponding to the decision variables in the basic data set of the kth iteration;

[0017] Encoding the supplementary simulation times index corresponding to the decision variables in the basic data set of the kth iteration and the decision variables in the basic data set of the kth iteration to obtain the initial population individual gene sequence of the kth iteration; the initial population individual gene sequence includes the initial gene; the initial gene includes the decision variables and the supplementary simulation times index corresponding to the decision variables;

[0018] The supplementary population individual gene sequence of the kth iteration is determined based on the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration.

[0019] Optionally, the determining of the supplementary population individual gene sequence of the kth iteration based on the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration specifically includes:

[0020] Determine the number of supplementary simulations for the decision variables in the initial population individual gene sequence of the kth iteration based on the supplementary simulation number index corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration;

[0021] The MOCBA algorithm is used to allocate simulation resources of the multi-objective function based on the simulation results of the k-1th iteration, and the simulation times corresponding to the decision variables in the initial gene sequence of the population individuals of the kth iteration are obtained;

[0022] Based on the number of simulations corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the number of supplementary simulations of the decision variables in the initial population individual gene sequence of the kth iteration, the supplemented population individual gene sequence of the kth iteration is determined; the number of simulations after supplementation corresponding to the decision variables is determined by adding the number of supplementary simulations corresponding to the decision variables and the number of simulations.

[0023] Optionally, the method of determining the number of supplementary simulations for the decision variables in the initial population individual gene sequence of the kth iteration based on the supplementary simulation number index corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration specifically includes:

[0024] Determine whether the simulation result of the k-1th iteration is a non-dominated solution set, and obtain a first determination result;

[0025] If the first judgment result is yes, then determine whether the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration is greater than a first preset threshold value to obtain a second judgment result; the first preset threshold value is the upper limit of the first supplementary simulation boundary value;

[0026] If the second judgment result is yes, then the number of supplementary simulations of the decision variables in the initial gene sequence of the population individuals of the kth iteration is set as the upper limit of the first supplementary simulation boundary value;

[0027] If the second judgment result is no, then the number of supplementary simulations of the decision variable in the initial population individual gene sequence of the kth iteration is set to the number of supplementary simulations corresponding to the decision variable in the basic data set of the kth iteration;

[0028] If the first judgment result is no, then determine whether the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration is greater than the second preset threshold value, and obtain a third judgment result; the second preset threshold value is the upper limit of the second supplementary simulation boundary value;

[0029] If the third judgment result is yes, then the number of supplementary simulations of the decision variables in the initial gene sequence of the population individuals of the kth iteration is set as the upper limit of the second supplementary simulation boundary value;

[0030] If the third judgment result is no, then the number of supplementary simulations of the decision variable in the initial population individual gene sequence of the kth iteration is set to the number of supplementary simulations corresponding to the decision variable in the basic data set of the kth iteration; the lower limit of the first supplementary simulation boundary value is equal to the upper limit of the second supplementary simulation boundary value.

[0031] Optionally, the determining of the initial population individual gene sequences of the k+1th iteration based on the parent population of the kth iteration specifically includes:

[0032] Performing crossover mutation on the parent population of the k-th iteration to determine the child population of the k-th iteration;

[0033] Merging the parent population of the k-th iteration and the child population of the k-th iteration to obtain the parent-child population of the k-th iteration;

[0034] The parent-child population of the k-th iteration is selected by using a survival selection method to determine the initial population individual gene sequence of the k+1-th iteration.

[0035] Optionally, the method of selecting the parent-child population of the kth iteration by a survival selection method to determine the initial population individual gene sequence of the k+1th iteration specifically includes:

[0036] Performing a fast non-dominated sort on the parent-child population of the k-th iteration to obtain a sorted parent-child population of the k-th iteration;

[0037] Calculating the crowding degree of each individual in the sorted parent-child generation population of the k-th iteration, screening the individuals in the sorted parent-child generation population of the k-th iteration based on the crowding degree of each individual, and determining the screened parent-child generation population of the k-th iteration;

[0038] Determine whether there is an individual meeting a preset condition in the parent-child generation population after screening in the k-th iteration, and obtain a fourth determination result;

[0039] If the fourth judgment result is yes, individuals meeting the preset conditions are eliminated to obtain the initial population individual gene sequence of the k+1th iteration;

[0040] If the fourth judgment result is no, the parent-child population after screening in the kth iteration is used as the initial population individual gene sequence in the k+1th iteration.

[0041] The second aspect of the present application provides a bank risk prediction device, which includes:

[0042] An acquisition unit is used to acquire a basic data set of a multi-objective function for bank risk prediction; the multi-objective function includes an objective function of a bank risk coefficient and an objective function of a bank operating indicator; the basic data set includes a plurality of decision variables and a solution space corresponding to each decision variable; the solution space includes a plurality of sets of simulation solutions for the decision variables; the simulation solution set includes a solution of an objective function of a bank risk coefficient and a solution of an objective function of a bank operating indicator;

[0043] A simulation unit is used to simulate and dynamically iterate the multi-objective function based on the basic data set of the multi-objective function during the iterative process of bank risk prediction, determine the optimal decision variables of the multi-objective function and the optimal simulation solution set corresponding to the optimal decision variables, and output the optimal risk coefficient and the optimal operating indicator corresponding to the optimal simulation solution set; the simulation dynamic iteration represents that the simulation resources used in the kth iteration process are dynamically adjusted based on the simulation results of the k-1th iteration, k≥2.

[0044] The third aspect of the present application provides a computer device, which includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the bank risk prediction method introduced in the first aspect.

[0045] The fourth aspect of the present application provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the bank risk prediction method introduced in the first aspect is implemented.

[0046] Compared with the prior art, this application has the following beneficial effects:

[0047] The present application obtains multiple decision variables of the multi-objective function of bank risk prediction and the solution space corresponding to each decision variable; the multi-objective function includes the objective function of the bank risk coefficient and the objective function of the bank operating indicator; the solution space includes multiple sets of simulation solution sets of the decision variables, and the simulation solution set includes the solution of the objective function of the bank risk coefficient and the solution of the objective function of the bank operating indicator; at the kth iteration, each decision variable and each simulation solution set corresponding to each decision variable are respectively input into the objective function of the bank risk coefficient and the objective function of the bank operating indicator to obtain the simulation result of the kth iteration, and the simulation resources used for the k+1th iteration are dynamically adjusted based on the simulation result obtained in the k+1th iteration, so that each simulation solution set corresponding to each decision variable in the k+1th iteration can obtain a different number of simulations based on the simulation result obtained in the kth iteration, so that the simulation results are closely combined with the allocation of simulation resources, so as to save simulation resources and avoid waste of simulation resources in the iterative process of bank risk prediction. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0049] Figure 1 A flow chart of a bank risk prediction method provided in an embodiment of the present application;

[0050] Figure 2 A schematic diagram of an iterative process provided in an embodiment of the present application;

[0051] Figure 3 A flowchart of S202 provided in an embodiment of the present application;

[0052] Figure 4 A schematic diagram of decision variables and supplementary simulation times indicators provided in an embodiment of the present application;

[0053] Figure 5 A flowchart of S303 provided in an embodiment of the present application;

[0054] Figure 6 A flowchart for determining the number of supplementary simulation times provided in an embodiment of the present application;

[0055] Figure 7 A flowchart of the MOCBA algorithm provided in the embodiment of the present application;

[0056] Figure 8 A schematic diagram of the Gaussian position mutation strategy provided in the embodiment of the present application;

[0057] Fig. 9 A schematic diagram of a uniform crossover strategy provided in an embodiment of the present application;

[0058] Fig.10 A flowchart of a survival selection method provided in an embodiment of the present application;

[0059] Fig.11 A schematic diagram of the sorted parent-child population provided in an embodiment of the present application;

[0060] Fig.12 Schematic diagram of elite retention strategy and binary tournament selection provided for embodiments of the present application;

[0061] Fig.13 A structural diagram of a bank risk prediction device provided in an embodiment of the present application;

[0062] Fig.14 A structural diagram of a simulation unit provided in an embodiment of the present application. DETAILED DESCRIPTION

[0063] As described above, the current simulation process is not closely integrated with the optimization process, and the two processes run almost independently. After research, the inventors have made the simulation results closely integrated with the simulation resource allocation, thereby saving simulation resources and avoiding waste of simulation resources in the iterative process of bank risk prediction.

[0064] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0065] Explanation of relevant terms:

[0066] Multi-objective problem: A multi-objective problem refers to an optimization problem in which there are multiple objective functions to be optimized at the same time, and these objective functions are usually contradictory. This leads to the problem that there is no single optimal solution, so the goal of the multi-objective problem is to find a set of solutions, which is called the Pareto optimal solution set.

[0067] Pareto optimal solution set: In multi-objective optimization, there are usually multiple conflicting optimization objectives, which means that improving one objective may lead to the deterioration of other objectives. The Pareto optimal solution set includes all solutions that cannot be further improved without deteriorating other objectives. In short, the Pareto optimal solution set includes all solutions that have no obvious disadvantages between multiple objectives.

[0068] Non-dominated solution set: The Pareto optimal solution set is a subset of the non-dominated solution set. Because the Pareto optimal solution set is the set of solutions that meet the Pareto optimality condition among all possible solutions, and the non-dominated solution set is the set of all current non-inferior solutions, which may contain some solutions that do not meet the Pareto optimality condition.

[0069] NSGA-II algorithm: NSGA-II is a genetic algorithm for solving multi-objective optimization problems. This algorithm regards the process of solving multi-objective problems as a multi-objective search process, and searches for the Pareto optimal solution set in the problem by simulating natural selection and genetic operations.

[0070] Random problem: A random problem refers to a problem in which there is randomness or uncertainty, that is, the result of the problem is affected by random factors. The random problem studied in this invention refers to a problem in which there are random variables that conform to a certain probability distribution in the objective function and constraints of the problem model.

[0071] Monte Carlo simulation: Monte Carlo simulation is a technique for simulating complex systems or problems through random sampling and probability statistics. Its core idea is to estimate the solution or result of a problem by generating a large number of random samples. In Monte Carlo simulation, random numbers are used to represent uncertainty factors, and then the simulation is run many times to calculate the statistical characteristics of the results. This method is often used to solve problems in mathematics, physics, finance and other fields, especially when the problem involves complex probability distribution.

[0072] Simulation resources: The problem involved in this application is a multi-objective stochastic optimization problem. The randomness of the problem means that after obtaining a set of feasible solutions, it is necessary to simulate the objective function in order to find the values ​​of different objective functions corresponding to each individual. For Monte Carlo simulation, theoretically, the optimal effect can be achieved when the number of simulations reaches infinity, but this approach is obviously unrealistic. In order to improve the simulation efficiency, the total number of simulations should be specified before the simulation is performed. The number of simulations allocated to each individual is the simulation resources occupied by the individual.

[0073] Simulation resource allocation: The purpose of simulation resource allocation is to ensure that the target selects the optimal solution with the highest probability by reasonably allocating the number of simulations when the total number of simulations is fixed, thereby optimizing system performance. Currently, the main simulation resource allocation methods include theoretical optimal allocation method, average allocation method, variance ratio method, simulation budget optimal allocation method and multi-objective simulation budget optimal allocation method (MOCBA algorithm).

[0074] MOCBA algorithm: As a simulation resource allocation algorithm, MOCBA is generally used in combination with a multi-objective evolutionary algorithm. It can adjust the allocation of computing resources according to the performance of the best solution, and can obtain a better simulation resource allocation plan within a smaller number of iterations. While controlling the cost of computing resources, it can greatly improve the exploration and expansion of solutions by multi-objective optimization methods.

[0075] This application adopts multi-objective optimization methods, such as non-dominated sorting genetic algorithm (NSGA), to effectively deal with bank credit risk while considering multiple objectives. These methods can find the Pareto optimal solution set and provide multiple optimal solutions to help banks better balance various objectives and manage credit risk more comprehensively. Therefore, the use of multi-objective methods helps to overcome the limitations of single-objective models and simple weighting, and improve the effectiveness and flexibility of bank credit risk management. Based on the multi-objective optimization method, this application considers the random factors in the model, combines MOCBA with the NSGA-Ⅱ algorithm, and simultaneously encodes the supplementary simulation number indicator and the basic decision variables, so that the simulation resource allocation is updated with the iterative process of the algorithm, and optimizes the collaborative optimization ability of the algorithm. See now Figure 1 The flowchart of the bank risk prediction method shown in the figure illustrates the bank risk prediction method of the present application:

[0076] S101. Obtain a basic data set of a multi-objective function for bank risk prediction.

[0077] The multi-objective function includes an objective function of the bank risk coefficient and an objective function of the bank operating index. The basic data set includes multiple decision variables and a solution space corresponding to each decision variable; the decision variable can represent the user's investment ratio; the solution space includes multiple sets of simulation solution sets of the decision variables; the simulation solution set includes a solution of an objective function of a bank risk coefficient and a solution of an objective function of a bank operating index; the solutions in the simulation solution set are not risk coefficients and operating indexes, and the risk coefficient can be obtained by inputting the solution of the objective function of the bank risk coefficient in the simulation solution set into the objective function of the bank risk coefficient, and the operating index can be obtained by inputting the solution of the objective function of the bank operating index into the objective function of the bank operating index. The operating index of this application can refer to economic benefits.

[0078] S102. During the iterative process of bank risk prediction, the multi-objective function is simulated and dynamically iterated based on the basic data set of the multi-objective function to determine the optimal decision variables of the multi-objective function and the optimal simulation solution set corresponding to the optimal decision variables, and the optimal risk coefficient and optimal operating indicator corresponding to the optimal simulation solution set are output.

[0079] The simulation dynamic iteration characterization states that the simulation resources used in the kth iteration process are dynamically adjusted based on the simulation results of the k-1th iteration, k ≥ 2. The simulation dynamic iteration can ensure that the simulation resource allocation and simulation results are closely integrated, making the simulation results the basis for simulation resource allocation.

[0080] For example, if some simulation results of the k-1th iteration are not ideal or are not what the user wants, the simulation resource allocation ratio of the simulation solution set of the decision variables corresponding to such simulation results in the kth iteration can be reduced.

[0081] Figure 2 For a schematic diagram of the iterative process, see Figure 2 Detailed description of S102:

[0082] S201. Set the value of k to 2.

[0083] S202: Generate a supplementary population individual gene sequence of the kth iteration based on the simulation result of the k-1th iteration.

[0084] The supplemented population individual gene sequence includes supplemented genes; the supplemented genes include decision variables and the number of simulations after supplementation corresponding to the decision variables; the simulation results are risk coefficients and operating indicators corresponding to the simulation solution set of the decision variables.

[0085] Figure 3 This is a flow chart of S202, now refer to Figure 3 Detailed description of S202:

[0086] S301, obtaining a supplementary simulation times index corresponding to a decision variable in a basic data set of the kth iteration.

[0087] The supplementary simulation times index of the decision variable of the kth iteration can be set according to the simulation results of the k-1th iteration. For example, if the risk coefficient is relatively low and the operation index is relatively high in the simulation results of the simulation solution set of a certain decision variable in the k-1th iteration, then the supplementary simulation times index of the simulation solution set of this decision variable can be set relatively high in the kth iteration.

[0088] S302, encoding the supplementary simulation times index corresponding to the decision variables in the basic data set of the kth iteration and the decision variables in the basic data set of the kth iteration to obtain the initial population individual gene sequence of the kth iteration.

[0089] The initial population individual gene sequence includes initial genes; the initial genes include decision variables and supplementary simulation times indicators corresponding to the decision variables.

[0090] The decision variables are encoded together with the supplementary simulation number indicator to obtain the gene sequence of the individual population in a combined form, ensuring that the simulation number is involved in the optimization process and better retaining the historical information of the evolutionary algorithm during the iteration process.

[0091] like Figure 4 The diagram of decision variables and supplementary simulation times is shown in Figure 1. The first x bits of the code represent the decision variables, using a mixed coding mode of integer coding and real number coding. The last bit represents the supplementary simulation times index, using integer coding. This coding method can better handle numerical and discrete problems and improve the applicability of the algorithm.

[0092] Specifically, the present application generates uniformly distributed random real numbers or integers according to the upper and lower bounds of the corresponding decision variables based on the first x bits of the code, and then screens them according to the constraints, retaining only the solutions that satisfy the constraints. The simulation times indicator is supplemented, and the first value is a random integer with a probability distribution between the boundaries [s0, s1]. The random distribution can be a random distribution in the problem, such as the banking risk default rate.

[0093] S303: Determine a supplementary population individual gene sequence for the kth iteration based on the initial population individual gene sequence for the kth iteration and the simulation result for the k-1th iteration.

[0094] Figure 5 This is the flow chart of S303. Figure 5 Detailed description of S303:

[0095] S401, determining the number of supplementary simulations for decision variables in the initial population individual gene sequence of the kth iteration based on the supplementary simulation number index corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration.

[0096] It is determined whether the simulation result of the k-1th iteration is a non-dominated solution set to obtain a first determination result.

[0097] If the first judgment result is yes, then determine whether the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration is greater than the first preset threshold to obtain the second judgment result; the first preset threshold is the upper limit of the first supplementary simulation boundary value.

[0098] If the second judgment result is yes, the number of supplementary simulations of the decision variables in the initial gene sequence of the population individuals in the kth iteration is set as the upper limit of the first supplementary simulation boundary value.

[0099] If the second judgment result is no, then the number of supplementary simulations of the decision variables in the initial population individual gene sequence of the kth iteration is set as the number of supplementary simulations corresponding to the decision variables in the basic data set of the kth iteration.

[0100] If the first judgment result is no, then determine whether the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration is greater than the second preset threshold to obtain a third judgment result; the second preset threshold is the upper limit of the second supplementary simulation boundary value.

[0101] If the third judgment result is yes, the number of supplementary simulations of the decision variables in the initial gene sequence of the population individuals in the kth iteration is set as the upper limit of the second supplementary simulation boundary value.

[0102] If the third judgment result is no, then the number of supplementary simulations of the decision variable in the initial population individual gene sequence of the kth iteration is set to the number of supplementary simulations corresponding to the decision variable in the basic data set of the kth iteration; the lower limit of the first supplementary simulation boundary value is equal to the upper limit of the second supplementary simulation boundary value.

[0103] Figure 6 For a flowchart to determine the number of additional simulations, see Figure 6 An example is given to illustrate the method of supplementing the number of simulations:

[0104] First, determine whether each individual is able to constitute a non-dominated solution set. If the individual does not constitute a non-dominated solution set, set the supplementary simulation boundary value to [s0, s1], otherwise set the boundary value to [s1, s2]. When the supplementary simulation number index of the individual is not within the boundary value, it will be adaptively decoded to the closest boundary value. For example, if the individual constitutes a non-dominated solution set, and the encoding value of the supplementary simulation number index is s, and s>s2, then the value is automatically converted to s2 first. s0 represents the lower limit of the second supplementary simulation boundary value, s1 represents the upper limit of the second supplementary simulation boundary value, and s2 represents the upper limit of the first supplementary simulation boundary value.

[0105] After determining the number of supplementary simulations, the simulation resource ratio of the current simulation solution set is given through normalization processing, and then the number of supplementary simulations for each individual is obtained. Finally, each individual is simulated through Monte Carlo simulation.

[0106] This application also limits the number of supplementary simulations. When the cumulative number of simulations input is greater than the set threshold, the individual will no longer increase the number of simulations through simulation resource allocation. The reason for limiting the number of supplementary simulations is that the survival selection method of this application includes an elite retention strategy, and there is a high probability that elite solutions that survive for multiple generations will remain in the later stages of iteration. The existence of these solutions will take up a lot of simulation resources.

[0107] S402, using the MOCBA algorithm to allocate simulation resources of the multi-objective function based on the simulation results of the k-1th iteration, and obtaining the number of simulations corresponding to the decision variables in the initial population individual gene sequence of the kth iteration.

[0108] In the first iteration, the MOCBA algorithm will make a unified initial setting for the number of simulations of each simulation solution set of each decision variable in the first iteration, and set the number of simulations of each simulation solution set of each decision variable in the first iteration to Δ.

[0109] Figure 7 This is the flowchart of the MOCBA algorithm. Figure 7 Detailed description of the MOCBA algorithm:

[0110] S701, calculating the mean and variance of each simulation result of the k-1th iteration.

[0111] S702, judging whether each simulation result of the k-1th iteration constitutes a non-dominated solution set, if so, proceeding to S703; if not, proceeding to S705.

[0112] S703 . If a non-dominated solution set is formed, then the simulation resource ratio of the simulation solution set in the non-dominated solution set is calculated, and then the process goes to S704 .

[0113] S704. Determine the simulation resource allocation ratio of each simulation solution set based on the proportion of simulation resources in the simulation solution set in the non-dominated solution set, the mean and variance of each simulation result of the k-1th iteration, and multiply the simulation resource allocation ratio by Δ to obtain the number of simulations corresponding to the decision variables in the initial population individual gene sequence of the kth iteration.

[0114] S705 . If it does not constitute a non-dominated solution set, then calculate the simulation resource ratio of the simulation solution set in the dominated solution set, and proceed to S706 .

[0115] S706. Determine the simulation resource allocation ratio of each simulation solution set based on the proportion of simulation resources in the simulation solution set in the dominant solution set, the mean and variance of each simulation result of the k-1th iteration, and multiply the simulation resource allocation ratio by Δ to obtain the number of simulations corresponding to the decision variables in the initial population individual gene sequence of the kth iteration.

[0116] S403, based on the number of simulations corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the number of supplementary simulations of the decision variables in the initial population individual gene sequence of the kth iteration, determine the supplementary population individual gene sequence of the kth iteration. The number of simulations after supplementation corresponding to the decision variables is determined by adding the number of supplementary simulations corresponding to the decision variables and the number of simulations.

[0117] S203 , performing simulation based on the supplemented population individual gene sequences of the k-th iteration and the multi-objective function to determine the parent population of the k-th iteration.

[0118] The parent population includes parent individuals; the parent individuals include decision variables and simulation results.

[0119] The gene sequences of the supplemented population individuals of the kth iteration are input into the multi-objective function to obtain the parent population of the kth iteration.

[0120] S204: Determine whether the parent population of the kth iteration meets the iteration end condition.

[0121] S205. If it meets the requirements, the decision variables of the parent individuals in the parent population of the kth iteration are taken as the optimal decision variables, the simulation solution set of the decision variables of the parent individuals in the parent population of the kth iteration is taken as the optimal simulation solution set, and the risk coefficient and operating indicator corresponding to the optimal simulation solution set are output as the optimal risk coefficient and the optimal operating indicator.

[0122] The iteration end condition is that the number of iterations reaches a preset number of iterations or the parent population contains a decision variable and a simulation result.

[0123] S206. If not, determine the initial population individual gene sequence of the k+1th iteration based on the parent population of the kth iteration, add 1 to the value of k, and return to the step of generating the supplementary population individual gene sequence of the kth iteration based on the simulation result of the k-1th iteration.

[0124] Determining the initial population individual gene sequence of the k+1th iteration based on the parent population of the kth iteration in S206 specifically includes:

[0125] By inputting a parent population with N individuals, a child population with the same number of N individuals can be generated, and the two populations can be merged into a parent-child population with 2N individuals.

[0126] The crossover mutator includes three steps: crossover, mutation, and merging parent and child populations. The crossover step randomly selects part of the codes of two individuals in the parent population and exchanges them to generate new child individuals, and the mutation step randomly changes some codes of the individuals to generate new child individuals. The purpose of both steps is to ensure that the species does not fall into a local optimal solution. The crossover step of this application adopts a uniform crossover strategy, and the mutation step adopts a Gaussian position mutation strategy.

[0127] Figure 8 This is a schematic diagram of the Gaussian position mutation strategy. Figure 8 Explain the Gaussian position mutation strategy:

[0128] Introduce random mutation factors into the population to increase diversity and avoid falling into the local optimal solution. Since the chromosome variable may exceed the range after mutation, it is also necessary to judge the boundary conditions: if the variable value after mutation exceeds the upper and lower bounds, the variable value is limited between the upper and lower bounds. If it is greater than the upper bound, it is limited to UB(i), and if it is less than the lower bound, it is limited to LB(i). The calculation process is as follows:

[0129]

[0130] v=u*randn*(UB(i)-LB(i)).

[0131] in, is the value of the decision variable numbered i of the jth individual in the tth generation, u is the scaling factor, and v represents the amount of variation. randn is a random variable that conforms to the normal distribution with a mean of 0 and a standard deviation of 1.

[0132] Fig. 9 This is a schematic diagram of the uniform crossover strategy. Fig. 9 Explanation of the uniform crossover strategy:

[0133] A random point is used for crossover operation, that is, a random number α is generated, and then a part of the second chromosome is mixed with the corresponding part of the first chromosome to generate two new chromosomes. The method is as follows:

[0134]

[0135] Among them, α is a uniformly distributed random number in [0,1], α and b represent the two parent chromosomes for crossover, and t represents the number of iterations.

[0136] The parent population of the k-th iteration is subjected to crossover mutation to determine the child population of the k-th iteration.

[0137] The parent population of the k-th iteration and the child population of the k-th iteration are merged to obtain the parent-child population of the k-th iteration.

[0138] The parent-child population of the k-th iteration is selected by using a survival selection method to determine the initial population individual gene sequence of the k+1-th iteration.

[0139] The survival selection method of the present application combines the risk preference of the banking industry, directly eliminates individuals whose risk value is higher than the set threshold, and adopts a binary tournament selection strategy for other individuals. Selecting a parent-child population with 2N individuals generated by inputting a crossover mutator can eliminate N individuals, and screen out N surviving individuals as the new parent population to participate in subsequent iterations.

[0140] Fig.10 This is a flowchart of the survival selection method. Fig.10 Explain the survival selection method:

[0141] 1) Fast non-dominated sorting and congestion calculation

[0142] The purpose of these two steps is to sort and compare the individuals in the population, so that the excellent individuals can be selected to survive when the elite retention strategy and binary tournament selection are adopted. The purpose of this step is to quickly sort the parent-child population with 2N individuals. The sorted parent-child population individuals are divided into several levels and Z is used. i Identification, such as Fig.11 The schematic diagram of the sorted parent-child generation population is shown in Figure 1. When making a selection, the smaller the subscript of the level, the more suitable the individual is for survival. When performing the elite retention strategy and the binary tournament selection, the more suitable individuals for survival will be selected to survive.

[0143] 2) Eliminate high-risk individuals

[0144] For the banking industry, it is impossible to pursue high profits without considering risks, and the risk preference is relatively conservative. Therefore, this application adds this step to the survival selection method to eliminate high-risk hidden individuals. By determining a high-risk threshold, it is used to determine whether an individual is a high-risk hidden danger. If it is, it is directly eliminated. This threshold can be determined based on business needs, regulatory requirements, or expert experience.

[0145] 3) Elite retention strategy

[0146] like Fig.12 Schematic diagram of the elite retention strategy and binary tournament selection shown, where P t and Q t are the parent population and the offspring population, R t is the parent-child population with 2N individuals after the parent-child population is merged, Z iIt represents the stratification of the parent-child generation population after fast non-dominated sorting. In order to ensure that the best individuals can survive, the elite retention strategy is used to directly retain the best individuals of αN to enter the next generation, where α is the elite retention ratio.

[0147] 4) Binary Tournament Selection

[0148] After the elite individuals are selected, binary tournament selection is continued among the remaining (2-α)N individuals to select the remaining (1-α)N individuals. In this way, elite individuals have a higher chance of being selected, which can effectively retain valuable genetic information and improve algorithm performance and search efficiency. The specific implementation is as follows: Fig.12 shown.

[0149] This application adopts binary tournament selection combined with elite retention strategy as the selection operator to maintain the diversity and robustness of the population and improve the performance and search efficiency of the algorithm.

[0150] Perform a fast non-dominated sort on the parent-child population of the k-th iteration to obtain a sorted parent-child population of the k-th iteration.

[0151] The crowding degree of each individual in the sorted parent-child population of the k-th iteration is calculated, and the individuals in the sorted parent-child population of the k-th iteration are screened based on the crowding degree of each individual to determine the screened parent-child population of the k-th iteration.

[0152] It is determined whether there are individuals meeting preset conditions in the parent-child generation population after screening in the k-th iteration, and a fourth determination result is obtained.

[0153] If the fourth judgment result is yes, individuals meeting the preset conditions are eliminated to obtain the initial population individual gene sequence of the k+1th iteration.

[0154] If the fourth judgment result is no, the parent-child population after screening in the kth iteration is used as the initial population individual gene sequence in the k+1th iteration.

[0155] There are some problems and limitations in the existing technology for solving multi-objective problems by converting them into single-objective problems. The single-objective model only considers one specific goal, such as minimizing credit risk, while ignoring other goals that may be equally important, such as maximizing returns or meeting regulatory compliance requirements. This may lead to insufficient trade-offs between multiple objectives and failure to fully consider the overall strategic needs of the bank. Simplifying a multi-objective problem into a single-objective problem and weighting each objective usually requires subjective weighing of the importance of each objective. This weighting method may not be reasonable because different objectives may influence each other, and this trade-off may change over time and market conditions. The multiple objectives of this application are taken into account in the risk management of the bank, which can better meet the overall strategic needs of the bank than a single objective, thereby overcoming the technical problems brought about by the existing technology.

[0156] In the prior art, the simulation process is not closely integrated with the optimization process. The two algorithms are almost run independently. The optimization process is only responsible for continuing to iterate the simulation solution fed back by the simulation process. The simulation process only uses the current generation population solution iterated by the optimization process to allocate simulation resources through the simulation resource allocation algorithm (EA, PTV, OCBA, MOCBA, etc.). The number of simulations is not taken into account in the iteration of the multi-objective evolutionary algorithm, resulting in a waste of simulation resources. The quality of the solution obtained by optimizing in the same time is low. However, the present application combines the simulation resource allocation strategy with the multi-objective evolutionary algorithm, introduces the concept of supplementary simulation times, so that the simulation resource allocation considers the historical iteration information of the individual population, and is no longer determined solely by the quality of each solution in the contemporary population. The improved supplementary simulation optimization strategy is embedded in the excellent evolutionary algorithm, and an adaptive decoding strategy is further proposed. Compared with existing technologies, the quality of the solution can be improved in the same simulation time.

[0157] The existing technology does not take into account the situation that when the elite retention strategy is adopted, simulation resources are occupied by elite individuals for a long time, resulting in a waste of simulation resources. This application sets a simulation threshold to prevent individuals selected by the elite retention strategy from occupying too many simulation resources for a long time. By allocating more simulation resources to non-elite solutions, more solutions for decision makers to choose can be found in the same time compared to existing technologies, thereby improving the distribution of solutions.

[0158] The prior art does not take into account the particularity of risk preference under bank risk management. The present invention solves this problem by eliminating high-risk individuals. The present application takes into account the particularity of risk preference under bank risk management and excludes individuals with high risk hidden dangers in the survival selection method.

[0159] Fig.13 This is a structural diagram of the bank risk prediction device, see Fig.13Explanation of the bank risk prediction device:

[0160] The acquisition unit is used to acquire a basic data set of a multi-objective function for bank risk prediction. The multi-objective function includes an objective function of a bank risk coefficient and an objective function of a bank operating index; the basic data set includes a plurality of decision variables and a solution space corresponding to each decision variable; the solution space includes a plurality of sets of simulation solutions for the decision variables; the simulation solution set includes a solution of an objective function of a bank risk coefficient and a solution of an objective function of a bank operating index.

[0161] The simulation unit is used to simulate and dynamically iterate the multi-objective function based on the basic data set of the multi-objective function during the iterative process of bank risk prediction, determine the optimal decision variables of the multi-objective function and the optimal simulation solution set corresponding to the optimal decision variables, and output the optimal risk coefficient and the optimal operating indicator corresponding to the optimal simulation solution set. The simulation dynamic iteration characterizes that the simulation resources used in the kth iteration process are dynamically adjusted based on the simulation results of the k-1th iteration, k≥2.

[0162] Fig.14 This is the structural diagram of the simulation unit. Fig.14 The structure of the simulation unit is described below:

[0163] The simulation unit includes a setting subunit, a generation subunit, an iteration subunit and a judgment subunit. The functions of each subunit in the simulation unit are described in detail below:

[0164] The subunit is set to set the value of k to 2.

[0165] A generation subunit is used to generate a supplementary population individual gene sequence of the kth iteration based on the simulation result of the k-1th iteration; the supplementary population individual gene sequence includes a supplementary gene; the supplementary gene includes a decision variable and the number of simulations after the supplementation corresponding to the decision variable; the simulation result is a risk coefficient and an operation indicator corresponding to a set of simulation solutions of the decision variable.

[0166] The generation subunit includes obtaining the primary module, encoding the primary module and determining the primary module. The function of each module in the generation subunit is now described:

[0167] The first-level module is obtained to obtain the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration.

[0168] The encoding first-level module is used to encode the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration and the decision variable in the basic data set of the kth iteration to obtain the initial population individual gene sequence of the kth iteration; the initial population individual gene sequence includes the initial gene; the initial gene includes the decision variable and the supplementary simulation number index corresponding to the decision variable.

[0169] A first-level module is determined, which is used to determine a supplementary population individual gene sequence of the kth iteration based on the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration.

[0170] The first-level module includes a second-level module for supplementing simulation times, a second-level module for allocating simulation times, and a second-level module for accumulating. The function of each second-level module in the first-level module is now described:

[0171] The secondary module for supplementing simulation times is used to determine the supplementary simulation times of the decision variables in the initial population individual gene sequence of the kth iteration based on the supplementary simulation times index corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration.

[0172] It is determined whether the simulation result of the k-1th iteration is a non-dominated solution set to obtain a first determination result.

[0173] If the first judgment result is yes, then determine whether the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration is greater than the first preset threshold to obtain the second judgment result; the first preset threshold is the upper limit of the first supplementary simulation boundary value.

[0174] If the second judgment result is yes, the number of supplementary simulations of the decision variables in the initial gene sequence of the population individuals in the kth iteration is set as the upper limit of the first supplementary simulation boundary value.

[0175] If the second judgment result is no, then the number of supplementary simulations of the decision variables in the initial population individual gene sequence of the kth iteration is set as the number of supplementary simulations corresponding to the decision variables in the basic data set of the kth iteration.

[0176] If the first judgment result is no, then determine whether the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration is greater than the second preset threshold to obtain a third judgment result; the second preset threshold is the upper limit of the second supplementary simulation boundary value.

[0177] If the third judgment result is yes, the number of supplementary simulations of the decision variables in the initial gene sequence of the population individuals in the kth iteration is set as the upper limit of the second supplementary simulation boundary value.

[0178] If the third judgment result is no, then the number of supplementary simulations of the decision variable in the initial population individual gene sequence of the kth iteration is set to the number of supplementary simulations corresponding to the decision variable in the basic data set of the kth iteration; the lower limit of the first supplementary simulation boundary value is equal to the upper limit of the second supplementary simulation boundary value.

[0179] The secondary module for allocating simulation times is used to allocate simulation resources of multi-objective functions based on the simulation results of the k-1th iteration using the MOCBA algorithm, and obtain the simulation times corresponding to the decision variables in the initial population individual gene sequence of the kth iteration.

[0180] The cumulative secondary module is used to determine the supplemented population individual gene sequence of the kth iteration based on the number of simulations corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the number of supplemented simulations of the decision variables in the initial population individual gene sequence of the kth iteration; the number of simulations after supplementation corresponding to the decision variables is determined by adding the number of supplemented simulations corresponding to the decision variables and the number of simulations.

[0181] The iterative subunit is used to perform simulation based on the supplemented population individual gene sequences of the k-th iteration and the multi-objective function to determine the parent population of the k-th iteration; the parent population includes parent individuals; the parent individuals include decision variables and simulation results.

[0182] A judgment subunit is used to judge whether the parent population of the kth iteration meets the iteration end condition. If so, the decision variables of the parent individuals in the parent population of the kth iteration are used as the optimal decision variables, the simulation solution set of the decision variables of the parent individuals in the parent population of the kth iteration is used as the optimal simulation solution set, and the risk coefficient and operation index corresponding to the optimal simulation solution set are output as the optimal risk coefficient and the optimal operation index; the iteration end condition is that the number of iterations reaches the preset number of iterations or the parent population contains a decision variable and a simulation result; if not, the initial population individual gene sequence of the k+1th iteration is determined based on the parent population of the kth iteration, and returned to the generation subunit.

[0183] The judgment subunit includes a judgment primary module, a crossover mutation primary module, a merging primary module, and an individual selection primary module. The function of each primary module in the judgment subunit is described below:

[0184] The first-level judgment module is used to judge whether the parent population of the k-th iteration meets the iteration end condition. If it meets the condition, the decision variables of the parent individuals in the parent population of the k-th iteration are used as the optimal decision variables, the simulation solution set of the decision variables of the parent individuals in the parent population of the k-th iteration is used as the optimal simulation solution set, and the risk coefficient and operation index corresponding to the optimal simulation solution set are output as the optimal risk coefficient and the optimal operation index. If it does not meet the condition, it will go to the first-level crossover mutation module.

[0185] The crossover mutation primary module is used to perform crossover mutation on the parent population of the k-th iteration to determine the child population of the k-th iteration.

[0186] The first-level merging module is used to merge the parent population of the k-th iteration and the child population of the k-th iteration to obtain the parent-child population of the k-th iteration.

[0187] The individual selection primary module is used to select the parent-child population of the k-th iteration by using the survival selection method to determine the initial population individual gene sequence of the k+1-th iteration.

[0188] The first-level individual selection module includes a sorting second-level module, a screening second-level module and a judgment second-level module. The function of each second-level module in the first-level individual selection module is described below:

[0189] The secondary sorting module is used to perform fast non-dominated sorting on the parent-child population of the k-th iteration to obtain the sorted parent-child population of the k-th iteration.

[0190] The secondary screening module is used to calculate the crowding degree of each individual in the sorted parent-child generation population of the k-th iteration, screen the individuals in the sorted parent-child generation population of the k-th iteration based on the crowding degree of each individual, and determine the screened parent-child generation population of the k-th iteration.

[0191] The secondary judgment module is used to judge whether there are individuals meeting the preset conditions in the parent-child generation population after screening of the k-th iteration, and obtain a fourth judgment result; if the fourth judgment result is yes, the individuals meeting the preset conditions are eliminated to obtain the initial population individual gene sequence of the k+1-th iteration; if the fourth judgment result is no, the parent-child generation population after screening of the k-th iteration is used as the initial population individual gene sequence of the k+1-th iteration.

[0192] A computer device comprises: a memory, a processor and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement a bank risk prediction method.

[0193] A computer-readable storage medium stores a computer program, which implements a bank risk prediction method when executed by a processor.

[0194] It should be noted that each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the device, equipment and medium embodiments, since they are basically similar to the method embodiments, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiments. The device, equipment and medium embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components indicated as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. A person of ordinary skill in the art can understand and implement it without paying creative labor.

[0195] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A bank risk prediction method, characterized in that: The bank risk prediction method comprises: Obtain a basic data set of a multi-objective function for bank risk prediction; the multi-objective function includes an objective function of a bank risk coefficient and an objective function of a bank operating indicator; the basic data set includes a plurality of decision variables and a solution space corresponding to each decision variable; the solution space includes a plurality of sets of simulation solutions for the decision variables; the simulation solution set includes a solution to an objective function of a bank risk coefficient and a solution to an objective function of a bank operating indicator; In the iterative process of bank risk prediction, the multi-objective function is simulated and dynamically iterated based on the basic data set of the multi-objective function to determine the optimal decision variables of the multi-objective function and the optimal simulation solution set corresponding to the optimal decision variables, and the optimal risk coefficient and the optimal operating indicator corresponding to the optimal simulation solution set are output; the simulation dynamic iteration represents that the simulation resources used in the kth iteration process are dynamically adjusted based on the simulation results of the k-1th iteration, k≥2.

2. The bank risk prediction method according to claim 1, characterized in that: In the iterative process of bank risk prediction, the multi-objective function is simulated and dynamically iterated based on the basic data set of the multi-objective function, the optimal decision variables of the multi-objective function and the optimal simulation solution set corresponding to the optimal decision variables are determined, and the optimal risk coefficient and the optimal operating indicator corresponding to the optimal simulation solution set are output, which specifically includes: Let the value of k be 2; Generate a supplemented population individual gene sequence of the kth iteration based on the simulation result of the k-1th iteration; the supplemented population individual gene sequence includes a supplemented gene; the supplemented gene includes a decision variable and the number of simulations after the supplementation corresponding to the decision variable; the simulation result is a risk coefficient and an operation index corresponding to a set of simulation solutions of the decision variable; Performing simulation based on the supplemented population individual gene sequences of the k-th iteration and the multi-objective function to determine the parent population of the k-th iteration; the parent population includes parent individuals; the parent individuals include decision variables and simulation results; Determine whether the parent population of the k-th iteration meets the iteration end condition. If so, use the decision variables of the parent individuals in the parent population of the k-th iteration as the optimal decision variables, use the simulation solution set of the decision variables of the parent individuals in the parent population of the k-th iteration as the optimal simulation solution set, and output the risk coefficient and operation index corresponding to the optimal simulation solution set as the optimal risk coefficient and the optimal operation index; the iteration end condition is that the number of iterations reaches the preset number of iterations or the parent population contains a decision variable and a simulation result; If not, the initial population individual gene sequence of the k+1th iteration is determined based on the parent population of the kth iteration, the value of k is increased by 1, and the step of generating the supplementary population individual gene sequence of the kth iteration based on the simulation result of the k-1th iteration is returned.

3. The bank risk prediction method according to claim 2, characterized in that: The generating of the k-th iteration supplementary population individual gene sequences based on the k-1-th iteration simulation result specifically includes: Obtain the supplementary simulation times index corresponding to the decision variables in the basic data set of the kth iteration; Encoding the supplementary simulation times index corresponding to the decision variables in the basic data set of the kth iteration and the decision variables in the basic data set of the kth iteration to obtain the initial population individual gene sequence of the kth iteration; the initial population individual gene sequence includes the initial gene; the initial gene includes the decision variables and the supplementary simulation times index corresponding to the decision variables; The supplementary population individual gene sequence of the kth iteration is determined based on the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration.

4. The bank risk prediction method according to claim 3, characterized in that: The step of determining the supplementary population individual gene sequence of the kth iteration based on the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration specifically includes: Determine the number of supplementary simulations for the decision variables in the initial population individual gene sequence of the kth iteration based on the supplementary simulation number index corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration; The MOCBA algorithm is used to allocate simulation resources of the multi-objective function based on the simulation results of the k-1th iteration, and the simulation times corresponding to the decision variables in the initial gene sequence of the population individuals of the kth iteration are obtained; Based on the number of simulations corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the number of supplementary simulations of the decision variables in the initial population individual gene sequence of the kth iteration, the supplemented population individual gene sequence of the kth iteration is determined; the number of simulations after supplementation corresponding to the decision variables is determined by adding the number of supplementary simulations corresponding to the decision variables and the number of simulations.

5. The bank risk prediction method according to claim 4, characterized in that: The method of determining the number of supplementary simulations of the decision variables in the initial population individual gene sequence of the kth iteration based on the supplementary simulation number index corresponding to the decision variables in the initial population individual gene sequence of the kth iteration and the simulation result of the k-1th iteration specifically includes: Determine whether the simulation result of the k-1th iteration is a non-dominated solution set, and obtain a first determination result; If the first judgment result is yes, then determine whether the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration is greater than a first preset threshold value to obtain a second judgment result; the first preset threshold value is the upper limit of the first supplementary simulation boundary value; If the second judgment result is yes, then the number of supplementary simulations of the decision variables in the initial gene sequence of the population individuals of the kth iteration is set as the upper limit of the first supplementary simulation boundary value; If the second judgment result is no, then the number of supplementary simulations of the decision variable in the initial population individual gene sequence of the kth iteration is set to the number of supplementary simulations corresponding to the decision variable in the basic data set of the kth iteration; If the first judgment result is no, then determine whether the supplementary simulation number index corresponding to the decision variable in the basic data set of the kth iteration is greater than the second preset threshold value, and obtain a third judgment result; the second preset threshold value is the upper limit of the second supplementary simulation boundary value; If the third judgment result is yes, then the number of supplementary simulations of the decision variables in the initial gene sequence of the population individuals of the kth iteration is set as the upper limit of the second supplementary simulation boundary value; If the third judgment result is no, then the number of supplementary simulations of the decision variable in the initial population individual gene sequence of the kth iteration is set to the number of supplementary simulations corresponding to the decision variable in the basic data set of the kth iteration; the lower limit of the first supplementary simulation boundary value is equal to the upper limit of the second supplementary simulation boundary value.

6. The bank risk prediction method according to claim 2, characterized in that: The step of determining the initial population individual gene sequence of the k+1th iteration based on the parent population of the kth iteration specifically includes: Performing crossover mutation on the parent population of the k-th iteration to determine the child population of the k-th iteration; Merging the parent population of the k-th iteration and the child population of the k-th iteration to obtain the parent-child population of the k-th iteration; The parent-child population of the k-th iteration is selected by using a survival selection method to determine the initial population individual gene sequence of the k+1-th iteration.

7. The bank risk prediction method according to claim 6, characterized in that: The method of adopting the survival selection method to select the parent-child population of the kth iteration to determine the initial population individual gene sequence of the k+1th iteration specifically includes: Performing a fast non-dominated sort on the parent-child population of the k-th iteration to obtain a sorted parent-child population of the k-th iteration; Calculating the crowding degree of each individual in the sorted parent-child generation population of the k-th iteration, screening the individuals in the sorted parent-child generation population of the k-th iteration based on the crowding degree of each individual, and determining the screened parent-child generation population of the k-th iteration; Determine whether there is an individual meeting a preset condition in the parent-child generation population after screening in the k-th iteration, and obtain a fourth determination result; If the fourth judgment result is yes, individuals meeting the preset conditions are eliminated to obtain the initial population individual gene sequence of the k+1th iteration; If the fourth judgment result is no, the parent-child population after screening in the kth iteration is used as the initial population individual gene sequence in the k+1th iteration.

8. A bank risk prediction device, characterized in that: The bank risk prediction device comprises: An acquisition unit is used to acquire a basic data set of a multi-objective function for bank risk prediction; the multi-objective function includes an objective function of a bank risk coefficient and an objective function of a bank operating indicator; the basic data set includes a plurality of decision variables and a solution space corresponding to each decision variable; the solution space includes a plurality of sets of simulation solutions for the decision variables; the simulation solution set includes a solution of an objective function of a bank risk coefficient and a solution of an objective function of a bank operating indicator; A simulation unit is used to simulate and dynamically iterate the multi-objective function based on the basic data set of the multi-objective function during the iterative process of bank risk prediction, determine the optimal decision variables of the multi-objective function and the optimal simulation solution set corresponding to the optimal decision variables, and output the optimal risk coefficient and the optimal operating indicator corresponding to the optimal simulation solution set; the simulation dynamic iteration represents that the simulation resources used in the kth iteration process are dynamically adjusted based on the simulation results of the k-1th iteration, k≥2.

9. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the bank risk prediction method described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the bank risk prediction method described in any one of claims 1 to 7 is implemented.

Images