Business object recognition model acquisition method and apparatus, and computer device

Abstract

The application relates to a business object recognition model acquisition method and device and a computer device, and is applied to the field of financial technology. The method comprises the following steps: acquiring at least one object description data of a sample business object; each object description data has a corresponding business characteristic attribute; each business characteristic attribute is taken as an independent variable to construct an equation population; the fitness of each equation individual is determined; a business object recognition model is trained in the equation population according to the fitness of each equation individual; object description data of a to-be-recognized business object is input into the business object recognition model to obtain prediction category information of the to-be-recognized business object, and the to-be-recognized business object is determined as a target business object according to the prediction category information of the to-be-recognized business object. The method can improve the efficiency of recognizing whether a business object is a target business object.

Description

Technical Field

[0001] This application relates to the field of financial technology, and in particular to a method, apparatus, computer device, storage medium, and computer program product for obtaining a business object identification model. Background Technology

[0002] The selection of underlying assets and other business targets in loan asset securitization is often a crucial factor in determining whether the issuance can achieve the expected business objectives. Errors in the selection of underlying assets will have a significant impact on the economic rights and interests of the original equity owner (i.e., the bank), and may even lead to serious business risks.

[0003] In existing loan asset securitization issuance processes, the original rights holder's business personnel typically select a list of securitizable underlying assets from the entire asset pool based on simple business filtering rules. For example, they might directly select assets with overdue days exceeding a specified threshold and loan asset types meeting a specified value as the underlying assets for securitization. However, such simple screening rules cannot effectively assess the asset quality of the loan assets.

[0004] Therefore, the relevant technologies suffer from poor performance in identifying whether a business object is the target business object. Summary of the Invention

[0005] Based on this, it is necessary to provide a method, apparatus, computer device, computer-readable storage medium, and computer program product for obtaining a business object identification model that can improve the effectiveness of identifying whether a business object is a target business object, in order to address the above-mentioned technical problems.

[0006] Firstly, this application provides a method for obtaining a business object identification model, the method comprising:

[0007] Obtain at least one object description data of a sample business object; each of the object description data has a corresponding business feature attribute;

[0008] Each of the aforementioned business feature attributes is used as an independent variable to construct an equation population; each equation individual in the equation population is used to characterize the different mapping relationships between the object description data of the sample business object and the predicted category information of the sample business object; the predicted category information is used to describe the predicted probability that the sample business object is the target business object.

[0009] Determine the fitness of each individual equation; the fitness is determined based on the difference between the predicted category information output by the corresponding individual equation for the sample business object and the actual category information of the sample business object;

[0010] Based on the fitness of each individual equation, a business object recognition model is trained in the equation population.

[0011] The object description data of the business object to be identified is input into the business object identification model to obtain the predicted category information of the business object to be identified, and the business object to be identified is determined to be the target business object based on the predicted category information of the business object to be identified.

[0012] In one embodiment, generating a business object identification model in the equation population according to the fitness of each equation individual includes: performing genetic algorithm optimization on each equation individual in the equation population according to the fitness of each equation individual to obtain a target equation individual; and using the equation represented by the target equation individual as the business object identification model.

[0013] In one embodiment, the step of constructing an equation population by using each of the business feature attributes as independent variables includes: combining the independent variable symbols corresponding to each of the business feature attributes with at least one preset operator to obtain at least two combined equations; the number of symbols in each of the combined equations meets a preset number threshold; encoding each of the combined equations to obtain individual equations; and adding each individual equation to the equation population.

[0014] In one embodiment, determining the fitness corresponding to each of the equation individuals includes: for any equation individual among the equation individuals, substituting the object description data of the sample business object into the equation represented by the equation individual to obtain the predicted category information output by the equation individual for the sample business object; and determining the fitness corresponding to the equation individual based on the difference between the predicted category information output by the equation individual and the actual category information of the sample business object.

[0015] In one embodiment, the equation population is an Nth generation equation population, where N is greater than or equal to 1. The step of optimizing each equation individual in the equation population using a genetic algorithm according to the fitness of each equation individual to obtain a target equation individual includes: if all fitness values ​​are less than a preset fitness threshold, then generating an N+1th generation equation population based on each equation individual in the Nth generation equation population, and returning to the step of determining the fitness of each equation individual; if there is a target fitness value among the fitness values ​​that is greater than or equal to the fitness threshold, then the equation individual corresponding to the target fitness value is taken as the target equation individual.

[0016] In one embodiment, generating the (N+1)th generation equation population based on each equation individual in the Nth generation equation population includes: obtaining the accumulated fitness value corresponding to each equation individual in the Nth generation equation population; selecting the (N+1)th generation equation individual from each equation individual in the Nth generation equation population; the fitness value corresponding to the (N+1)th generation equation individual satisfies a preset fitness range; the fitness range is determined based on the accumulated value corresponding to each equation individual; and performing evolutionary operator operations on each (N+1)th generation equation individual to generate the (N+1)th generation equation population.

[0017] In one embodiment, the step of performing evolutionary operator operations on each of the (N+1)th generation equation individuals to generate the (N+1)th generation equation population includes: in response to the (N+1)th generation equation individuals satisfying a preset mutation condition, obtaining the individual code to be mutated corresponding to the (N+1)th generation equation individual; changing the value of at least one coding bit in the individual code to be mutated to obtain the mutated individual code; and adding the equation individual represented by the mutated individual code and the (N+1)th generation equation individuals to the (N+1)th generation equation population.

[0018] In one embodiment, the step of performing evolutionary operator operations on each of the (N+1)th generation equation individuals to generate the (N+1)th generation equation population includes: in response to at least two of the (N+1)th generation equation individuals satisfying a preset crossover condition, obtaining the crossover individual codes corresponding to at least two of the (N+1)th generation equation individuals; determining the code segments in each crossover individual code based on the values ​​corresponding to at least two code bits in each crossover individual code; exchanging the code segments between each crossover individual code to obtain the crossover individual code; and adding the equation individual represented by the crossover individual code and the (N+1)th generation equation individuals to the (N+1)th generation equation population.

[0019] Secondly, this application also provides a device for obtaining a business object identification model, the device comprising:

[0020] The acquisition module is used to acquire at least one object description data of a sample business object; each of the object description data has a corresponding business feature attribute.

[0021] A construction module is used to construct an equation population by taking each of the business feature attributes as independent variables; each equation individual in the equation population is used to represent different mapping relationships between the object description data of the sample business object and the prediction category information of the sample business object; the prediction category information is used to describe the prediction probability that the sample business object is the target business object.

[0022] A determination module is used to determine the fitness of each of the equation individuals; the fitness is determined based on the difference between the predicted class information output by the corresponding equation individual for the sample business object and the actual class information of the sample business object;

[0023] The optimization module is used to train a business object recognition model in the equation population according to the fitness of each individual equation.

[0024] The identification module is used to input the object description data of the business object to be identified into the business object identification model, obtain the predicted category information of the business object to be identified, and determine the business object to be identified as the target business object based on the predicted category information of the business object to be identified.

[0025] Thirdly, this application also provides a computer device. The computer device includes a memory and a processor, the memory storing a computer program that, when executed by the processor, implements the steps of the method described above.

[0026] Fourthly, this application also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program thereon, which, when executed by a processor, implements the steps of the above-described method.

[0027] Fifthly, this application also provides a computer program product. The computer program product includes a computer program that, when executed by a processor, implements the steps of the above-described method.

[0028] The aforementioned method, apparatus, computer equipment, storage medium, and computer program product for acquiring business object identification models involve acquiring at least one object description data of a sample business object; each object description data has a corresponding business feature attribute; and using each business feature attribute as an independent variable to construct an equation population; each equation individual in the equation population is used to characterize different mapping relationships between the object description data of the sample business object and the predicted category information of the sample business object; the predicted category information is used to describe the predicted probability that the sample business object is the target business object; and the fitness corresponding to each equation individual is determined; the fitness is based on the predicted category information output by the corresponding equation individual for the sample business object and the predicted category information of the sample business object. The differences between the actual types of business objects are determined; according to the fitness of each individual in the equation, a business object identification model is trained in the equation population; the object description data of the business object to be identified is input into the business object identification model to obtain the predicted type information of the business object to be identified, and based on the predicted type information of the business object to be identified, the business object to be identified is determined to be the target business object; in this way, the idea of ​​genetic algorithm can be used to quickly and effectively establish a high-precision fitting model of the identification result of the business object and at least one business feature attribute associated with the business object, i.e., the business object identification model, thereby improving the identification effect of whether the business object is the target business object (e.g., whether the asset is the underlying asset of securitization). Attached Figure Description

[0029] Figure 1 This is an application environment diagram of a method for obtaining a business object identification model in one embodiment;

[0030] Figure 2 This is a flowchart illustrating a method for obtaining a business object identification model in one embodiment;

[0031] Figure 3 This is a schematic diagram of a securitization underlying asset determination system in one embodiment;

[0032] Figure 4 This is a schematic diagram illustrating the encoding of an equation in one embodiment;

[0033] Figure 5 This is a schematic diagram of the training process of a business object recognition model in one embodiment;

[0034] Figure 6 This is a flowchart illustrating a method for obtaining a business object identification model in another embodiment;

[0035] Figure 7 This is a structural block diagram of a device for acquiring a business object identification model in one embodiment;

[0036] Figure 8This is an internal structural diagram of a computer device in one embodiment. Detailed Implementation

[0037] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0038] It should be noted that the method, apparatus, computer equipment, storage medium and computer program product for obtaining a business object identification model disclosed in this application can be applied to the financial technology field, and can also be used in any field other than the financial technology field.

[0039] The method for obtaining the business object identification model provided in this application embodiment can be applied to, for example, Figure 1 In the application environment shown, electronic device 102 communicates with server 104 via a network. A data storage system can store the data that server 104 needs to process. The data storage system can be integrated onto server 104, or it can be located in the cloud or on another network server. Electronic device 102 acquires at least one object description data of a sample business object; each object description data has a corresponding business feature attribute; electronic device 102 uses each business feature attribute as an independent variable to construct an equation population; each equation individual in the equation population is used to characterize different mapping relationships between the object description data of the sample business object and the predicted category information of the sample business object; the predicted category information is used to describe the predicted probability that the sample business object is the target business object; electronic device 102 determines the fitness corresponding to each equation individual; the fitness is determined based on the difference between the predicted category information output by the corresponding equation individual for the sample business object and the actual category information of the sample business object; electronic device 102 trains a business object recognition model in the equation population according to the fitness corresponding to each equation individual; electronic device 102 inputs the object description data of the business object to be identified into the business object recognition model to obtain the predicted category information of the business object to be identified, and determines that the business object to be identified is the target business object based on the predicted category information of the business object to be identified. The electronic device 102 can be, but is not limited to, various personal computers, laptops, smartphones, tablets, IoT devices, and portable wearable devices. IoT devices can include smart speakers, smart TVs, smart air conditioners, and smart in-vehicle devices. Portable wearable devices can include smartwatches, smart bracelets, and head-mounted devices. The server 104 can be implemented using a standalone server or a server cluster consisting of multiple servers.

[0040] In one embodiment, such as Figure 2As shown, a method for obtaining a business object identification model is provided, which can be applied to... Figure 1 Taking an electronic device as an example, the explanation includes the following steps:

[0041] Step 202: Obtain at least one object description data of the sample business object; each object description data has a corresponding business feature attribute.

[0042] In this context, the business object can refer to the object whose business conditions need to be determined. In practical applications, the business object can be a financial product, service, resource, etc. For example, the business object can be a financial asset.

[0043] Among them, object description data can refer to data used to describe various aspects of business objects, and each object description data has corresponding business characteristic attributes.

[0044] Among these, business characteristic attributes refer to the characteristic attributes used to determine whether a business object meets the business conditions. Taking a financial business asset as an example, when it is necessary to determine whether a financial business asset can be securitized, at least one business characteristic attribute associated with the business object can be a core influencing factor such as the number of overdue days of the financial business asset, loan type, customer qualifications, and repayment rate.

[0045] For example, the object description data of financial business asset A may include overdue days of 10, loan type of commercial loan, and repayment rate of 80%; then the business characteristic attributes of financial business asset A are overdue days, loan type, and repayment rate.

[0046] In practical implementation, when it is necessary to determine whether a business object to be identified is a target business object that meets the business conditions, the electronic device can obtain at least one object description data of a sample business object; each object description data has a corresponding business characteristic attribute. For example, when it is necessary to determine whether a business asset to be identified is a securitizable business asset, the electronic device can sample from several known securitizable target business assets to obtain a sample business asset. The electronic device obtains the object description data of the sample business asset, and each object description data has a corresponding business characteristic attribute.

[0047] Step 204: Use each business characteristic attribute as an independent variable to construct an equation population.

[0048] In this context, each equation individual in the equation population is used to characterize different mapping relationships between the object description data of the sample business object and the predicted category information of the sample business object.

[0049] For example, when an electronic device inputs the object description data of a sample business object into any one of the equation individuals, that equation individual will preset the input object description data as the prediction category information of the sample business object, and use the prediction category information to describe the predicted probability that the sample business object is the target business object.

[0050] The identification result is used to describe whether the business object is the target business object.

[0051] In practice, after obtaining at least one business characteristic attribute associated with a business object, the electronic device can use each business characteristic attribute as an independent variable to construct an equation population.

[0052] Specifically, electronic devices can form equations composed of independent variable symbols corresponding to each business characteristic attribute, dependent variable symbols corresponding to the recognition results, and operation symbols (such as +, -, *, / , cos, sin, tan, cot, (), exponentiation, etc.).

[0053] For example, an electronic device can combine the independent variable symbol 'x' corresponding to the number of overdue days, the independent variable symbol 'y' corresponding to the loan type, the independent variable symbol 'u' corresponding to the customer's creditworthiness, the independent variable symbol 'v' corresponding to the repayment rate, the dependent variable symbol 'γ' corresponding to the recognition result, and operators (such as +, -, *, / , cos, sin, tan, cot, (), exponentiation, etc.) to obtain equation expressions, such as γ = x + uy / v, γ = cosx + uy / v, γ = x + u*y / v, etc. Then, the electronic device encodes each equation expression into an individual equation to construct an equation population.

[0054] Step 206: Determine the fitness of each equation individual; the fitness is determined based on the difference between the predicted category information output by the corresponding equation individual for the sample business object and the actual category information of the sample business object.

[0055] In practice, after the electronic device constructs the equation population, it can determine the fitness of each equation individual.

[0056] Specifically, the electronic device can substitute the object description data of the sample business object into the equation represented by the equation individual to obtain the predicted class information output by the equation individual for the sample business object. Then, the electronic device can use a preset fitness function to determine the fitness of the equation individual based on the difference between the predicted class information and the actual class information of the sample business object. In practical applications, the electronic device can use a probability value to represent the actual class information of the sample business object. The actual class information of the sample business object can be set to 1.

[0057] Step 208: Train a business object recognition model in the equation population according to the fitness of each individual equation.

[0058] In practical implementation, electronic devices can utilize genetic algorithms to optimize the individual equations in the equation population based on their fitness. Specifically, the electronic devices can perform a series of genetic operator operations such as selection, elimination, crossover, and mutation on each individual equation according to its fitness, until a offspring equation with a fitness that meets preset conditions is obtained, which serves as the target equation individual.

[0059] In practical applications, electronic devices perform parameter setting operations, specifically including the probabilities of evolutionary operators (mutation, crossover, etc.), population size Pop (the number of individuals participating in training), maximum number of iterations T, and termination conditions (the number of iterations reaches T or the fitness value of the best individual in the population is less than a certain fitness threshold). Each parameter is set according to different scenarios. Among these, the probabilities of evolutionary operators and the population size form the basis of the evolutionary computation search process, while the setting of the maximum number of iterations and the termination condition determines when the evolutionary computation search process ends. Since evolutionary computation is a continuously optimizing iterative process, it does not utilize information such as function gradients to measure the position of individuals in the solution space, making it impossible to use traditional methods to determine the convergence of the algorithm or whether it needs to be terminated. Therefore, a common approach is to pre-define a maximum number of evolutionary generations or a number of generations after which the algorithm terminates when the fitness value of the solution does not show significant improvement. Setting the termination parameter in evolutionary computation is also a crucial step affecting algorithm performance. Too small an evolutionary generation will lead to exiting the search before finding the global optimum, while too large an evolutionary generation will result in a large number of invalid searches, incurring too much time cost and reducing algorithm efficiency.

[0060] Once the electronic device identifies the target equation, it can reverse the equation according to the encoding rules for the target equation and use the reversed equation as a business object identification model to identify whether the input business object is the target business object.

[0061] Step S210: Input the object description data of the business object to be identified into the business object identification model to obtain the predicted type information of the business object to be identified, and determine the business object to be identified as the target business object based on the predicted type information of the business object to be identified.

[0062] Taking financial business assets as an example, after determining the business object identification model (i.e., the securitization underlying asset determination fitting model), the electronic device maps the object description data of the business object to be identified, such as overdue days (x), loan type (y), customer qualification (u), and repayment rate (v), to the predicted type information γ of the business object to be identified, i.e., the probability that the asset meets the securitization underlying asset determination.

[0063] according to Figure 3 As shown, a securitization underlying asset determination system is established. The asset quality influencing factors such as the number of overdue days (x), loan type (y), customer qualification (u), and repayment rate (v) of the candidate assets are input into the securitization underlying asset determination fitting model, and the probability γ of the asset meeting the securitization underlying asset determination is output. According to the actual business rules, a fitting threshold δ is set. When |γ-1|<=δ, the asset is determined to be a securitization underlying asset.

[0064] In the above method for acquiring the business object identification model, at least one object description data of a sample business object is acquired; each object description data has a corresponding business feature attribute; and each business feature attribute is used as an independent variable to construct an equation population; each equation individual in the equation population is used to characterize different mapping relationships between the object description data of the sample business object and the predicted category information of the sample business object; the predicted category information is used to describe the predicted probability that the sample business object is the target business object; the fitness corresponding to each equation individual is determined; the fitness is based on the predicted category information output by the corresponding equation individual for the sample business object and the actual category information of the sample business object. The differences between them are determined; according to the fitness of each individual in the equation, a business object identification model is trained in the equation population; the object description data of the business object to be identified is input into the target identification model to obtain the predicted type information of the business object to be identified, and the business object to be identified is determined to be the target business object based on the predicted type information of the business object to be identified; in this way, the idea of ​​genetic algorithm can be used to quickly and effectively establish a high-precision fitting model of the identification result of the business object and at least one business feature attribute associated with the business object, that is, the business object identification model, thereby improving the identification effect of whether the business object is the target business object (e.g., whether the asset is the underlying asset of securitization).

[0065] In another embodiment, each business feature attribute is used as an independent variable to construct an equation population, including: combining the independent variable symbols corresponding to each business feature attribute with at least one preset operator to obtain at least two combined equations; the number of symbols in each combined equation meets a preset number threshold; encoding each combined equation to obtain individual equations; and adding each individual equation to the equation population.

[0066] In specific implementation, when constructing an equation population by using various business feature attributes as independent variables, the electronic device can combine the independent variable symbols corresponding to each business feature attribute, the dependent variable symbols corresponding to the recognition results, and at least one preset operator to obtain at least two combined equations. The number of symbols in each combined equation meets a preset threshold, for example, the number of symbols in each combined equation is equal. Then, the electronic device encodes each combined equation to obtain individual equations. Each individual equation is then added to the equation population.

[0067] Taking assets used to determine whether they meet the criteria for securitization as an example, the equations generated by electronic devices are composed of overdue days (x), loan type (y), customer qualification (u), repayment rate (v), and selectable operators (+, -, *, / , cos, sin, tan, cot, (), exponentiation, etc.). Each individual equation has a consistent length (i.e., the total number of operators and variables N contained in the equation), such as (y / u)¹ / ² + xv. To facilitate evolutionary operator operations and reduce the complexity of implementation using computer languages, complex equations can be encoded using binary trees, as follows: Figure 4 As shown, Q represents the square root operator, and the population individual expression "+Q-yuxv" is formed by encoding the binary tree nodes in the order from top to bottom and from left to right.

[0068] The technical solution of this embodiment combines the independent variable symbols corresponding to each business feature attribute, the dependent variable symbols corresponding to the recognition result, and at least one preset operator to quickly initialize a preset number of equation expressions, thereby facilitating the encoding of the equation expressions and constructing an equation population that can be optimized by genetic algorithms.

[0069] In another embodiment, determining the fitness of each equation individual includes: for any equation individual among the equation individuals, substituting the business feature information of the sample business object into the equation represented by the equation individual to obtain the prediction information corresponding to the recognition result output by the equation individual for the sample business object; and determining the fitness of the equation individual based on the difference between the prediction information output by the equation individual and the actual information corresponding to the recognition result of the sample business object.

[0070] Assume f(x) is the result of the population individual equation, that is, the predicted category information corresponding to the identification result output by the individual for the sample business object, which conforms to the probability of the underlying asset of securitization. It is known that the actual information corresponding to the identification result of the sample business object is 1. Theoretically, the smaller the value of f'(x) = |f(x)-1|, the better the individual. In order to adapt to the roulette wheel rule, this function can be further adjusted. Finally, the individual fitness function used in this scheme is: F(x) = 1 / ((|f(x)-1|)2+1). The larger F(x) is, the better the fitness of the individual, and the greater the probability of it being selected into the mating pool for evolutionary operator operation and generating offspring to enter the next generation cycle.

[0071] In another embodiment, the genetic algorithm is used to optimize each equation individual in the equation population according to the fitness of each equation individual to obtain the target equation individual. This includes: if all fitness values ​​are less than a preset fitness threshold, then the N+1th generation equation population is generated based on each equation individual in the Nth generation equation population, and the step of determining the fitness of each equation individual is returned; if there is a target fitness value among the fitness values ​​that is greater than or equal to the fitness threshold, then the equation individual corresponding to the target fitness value is taken as the target equation individual.

[0072] The equation population is the Nth generation equation population, where N is greater than or equal to 1.

[0073] In its implementation, the electronic device optimizes the equation population using a genetic algorithm based on the fitness of each equation individual to obtain the target equation individual. During this process, the electronic device can determine if the fitness of each equation individual is less than a preset fitness threshold. If all fitness values ​​are less than the threshold, the electronic device generates an (N+1)th generation equation population based on the equation individuals in the Nth generation population and returns to the step of determining the fitness of each equation individual, thus re-determining the fitness of each equation individual in the (N+1)th generation quantized model set. In practical applications, the process by which the electronic device determines the fitness of each equation individual in the (N+1)th generation quantized model set is consistent with the method used in determining the fitness of each equation individual in the Nth generation quantized model set, and will not be repeated here. This continues until the electronic device detects a target fitness value greater than or equal to the fitness threshold; in this case, the equation individual corresponding to the target fitness value is taken as the target equation individual.

[0074] In another embodiment, generating an N+1 generation equation population based on the equation individuals in the Nth generation equation population includes: obtaining the accumulated fitness value corresponding to each equation individual in the Nth generation equation population; selecting N+1 generation equation individuals from among the equation individuals in the Nth generation equation population; ensuring that the fitness value corresponding to the N+1 generation equation individual meets a preset fitness range; the fitness range is determined based on the accumulated fitness value corresponding to each equation individual; and performing evolution operator operations on each N+1 generation equation individual to generate the N+1 generation equation population.

[0075] In practice, the electronic device acquires the cumulative fitness values ​​of each individual in the Nth generation equation population; then, it selects individuals from the (N+1)th generation equation population. The electronic device then performs evolutionary operator operations on each (N+1)th generation equation individual to generate the (N+1)th generation equation population.

[0076] Evolutionary operators can include mutation operators, crossover operators, selection operators, and elimination operators.

[0077] Electronic devices employ the wheel selection principle as their selection strategy, which plays a positive role in optimizing subsequent evolutionary outcomes. According to fitness values ​​and the wheel selection principle, individuals with higher fitness are more likely to be replicated in the next generation. During replication, the number of replications is determined based on the wheel selection principle, while the population size remains constant. Assuming the current population size is 4, and the individuals are numbered 1, 2, 3, and 4, the fitness values ​​of each individual and their cumulative fitness values ​​are shown in Table 1 below:

[0078] Table 1. Fitness values ​​and cumulative fitness values ​​of individuals

[0079]

[0080] Generate uniformly distributed random numbers within the interval [0, 1.6]. If the generated random number is 1.2, since 1.2 falls within the interval (0.6, 1.3], select individual number 3, i.e., the N+1th generation equation individual, to enter the mating pool (individuals can be selected repeatedly). Repeat this step until the set number of operands is reached, and use the resulting new population as the parent to participate in the evolutionary operation.

[0081] In another embodiment, evolutionary operator operations are performed on each N+1 generation equation individual to generate an N+1 generation equation population, including: in response to the N+1 generation equation individual satisfying a preset mutation condition, obtaining the individual code to be mutated corresponding to the N+1 generation equation individual; changing the value of at least one coding bit in the individual code to be mutated to obtain the mutated individual code; and adding the equation individual represented by the mutated individual code and the N+1 generation equation individual to the N+1 generation equation population.

[0082] In the specific implementation, during the process of generating the N+1 generation equation population by performing evolution operator operations on each N+1 generation equation individual, the electronic device can obtain the individual code to be mutated corresponding to the N+1 generation equation individual if the N+1 generation equation individual meets the preset mutation conditions; then, the electronic device changes the value of at least one bit in the individual code to be mutated to obtain the mutated individual code; finally, the electronic device adds the equation individual represented by the mutated individual code and the N+1 generation equation individual to the N+1 generation equation population.

[0083] Specifically, the electronic device can perform mutation operator operations on individuals of the N+1 generation equation within a controllable range; the electronic device sets the mutation rate to β, and satisfies β∈(0,1), it iterates through and selects each individual in the existing mating pool as the parent individual, generates a random number ε∈[0,1], if ε<β, then the individual mutates and generates a child individual to enter the next generation cycle, otherwise the individual does not mutate and directly enters the next generation cycle; the mutation point position n is a randomly generated integer, and n∈(1,M), where M is the maximum length of the individual encoding expression, which can be set according to the actual scenario.

[0084] The operation is illustrated in Tables 2 and 3 below (it should be noted that, for ease of understanding, the individuals in the tables are expressed directly using equations, rather than expressions encoded in binary trees), where the single-point mutation point is at position 4, and the multi-point mutation points are at positions 1, 2, and 5:

[0085] Table 2 Single-point mutation operator process

[0086]

[0087] Table 3 Multipoint mutation operator process

[0088]

[0089]

[0090] The technical solution of this embodiment obtains the individual code to be mutated corresponding to the N+1 generation equation individual by responding to the N+1 generation equation individual meeting the preset mutation conditions; and changes the value of at least one coding bit in the individual code to be mutated to obtain the mutated individual code; then adds the equation individual represented by the mutated individual code and the N+1 generation equation individual to the N+1 generation equation population; it can effectively perform mutation operations on each N+1 generation equation individual to obtain the N+1 generation equation population, so that the equation individuals in the N+1 generation equation population can inherit the characteristics of the Nth equation individual, while the number of equation individuals in the N+1 generation equation population meets the preset conditions.

[0091] In another embodiment, evolutionary operator operations are performed on each N+1 generation equation individual to generate an N+1 generation equation population, including: in response to at least two N+1 generation equation individuals satisfying a preset crossover condition, obtaining the codes of the individuals to be crossed corresponding to at least two N+1 generation equation individuals; determining the code segments in the codes of each individual to be crossed based on the values ​​corresponding to at least two code bits in the codes of each individual to be crossed; exchanging the code segments between the codes of each individual to be crossed to obtain the crossover individual codes; and adding the equation individuals represented by the crossover individual codes and the N+1 generation equation individuals to the N+1 generation equation population.

[0092] In specific implementation, the electronic device, in response to at least two N+1 generation equation individuals satisfying a preset crossover condition, obtains the codes of the individuals to be crossed corresponding to at least two N+1 generation equation individuals. Then, the electronic device determines the code segment in each individual code based on the values ​​corresponding to at least two code bits in each individual code. Next, the electronic device swaps the code segments between the individual codes to be crossed to obtain the crossover individual code. Finally, the electronic device adds the equation individual represented by the crossover individual code and the N+1 generation equation individuals to the N+1 generation equation population.

[0093] Specifically, the electronic device, based on the crossover operator design concept, randomly selects two individuals from the mating pool as parent individuals, sets the crossover rate to α, and satisfies α∈(0,1). It generates a random number ε∈[0,1]. If ε<α, satisfying the preset crossover condition, the individuals crossover, and the two generated child individuals enter the next generation cycle. Otherwise, the two individuals do not crossover and directly enter the next generation cycle. The operation is illustrated in Table 4 below (it should be noted that, for ease of understanding, the individuals in the table are expressed directly using equations, rather than expressions encoded in a binary tree). In this table, the crossover point for a single-point crossover is between 3 and 4, and the crossover points for a two-point crossover are between 1 and 2, and between 4 and 6.

[0094] Table 4 Crossover Operator Process

[0095]

[0096] The technical solution of this embodiment obtains the codes of at least two N+1 generation equation individuals corresponding to the equation individuals to be crossed in response to at least two N+1 generation equation individuals satisfying a preset crossover condition; and determines the code segments in each equation individual code based on the values ​​corresponding to at least two code bits in each equation individual code; then swaps the code segments between each equation individual code to obtain the crossover individual code; and adds the equation individual represented by the crossover individual code and the N+1 generation equation individuals to the N+1 generation equation population; thereby, it is possible to effectively perform crossover operations on each N+1 generation equation individual to obtain the N+1 generation equation population, so that the equation individuals in the N+1 generation equation population can inherit the characteristics of the Nth equation individual, while the number of equation individuals in the N+1 generation equation population satisfies the preset condition.

[0097] For the convenience of those skilled in the art, Figure 5 A training process for a business object recognition model is provided, such as... Figure 5 The steps are as follows: Step 502, parameter setting; Step 504, population initialization; Step 506, strategy design; Step 508, fitness function determination; Step 510, evolutionary operator operation; Step 512, termination condition judgment; Step 514, output results. It should be noted that the specific limitations of the above steps can be found in the specific limitations of the method for obtaining a business object recognition model described above, and will not be elaborated further here.

[0098] In another embodiment, such as Figure 6 As shown, a method for obtaining a business object identification model is provided, which can be applied to... Figure 1 Taking an electronic device as an example, the explanation includes the following steps:

[0099] Step 610: Obtain at least one object description data of the sample business object; each object description data has a corresponding business feature attribute.

[0100] Step 620: Combine the independent variable symbols corresponding to each business feature attribute with at least one preset operator to obtain at least two combined equations; the number of symbols in each combined equation meets the preset number threshold.

[0101] Step 630: Encode each combined equation to obtain individual equations.

[0102] Step 640: Add each equation individual to the equation population; each equation individual in the equation population is used to represent different mapping relationships between the object description data of the sample business object and the prediction category information of the sample business object; the prediction category information is used to describe the prediction probability that the sample business object is the target business object.

[0103] Step 650: For any equation individual among the equation individuals, substitute the business feature information of the sample business object into the equation represented by any equation individual to obtain the prediction category information corresponding to the recognition result output by any equation individual for the sample business object.

[0104] Step 660: Based on the difference between the predicted category information output by any equation individual and the actual category information of the sample business object, determine the fitness corresponding to any equation individual.

[0105] Step 670: Based on the fitness of each individual equation, perform genetic algorithm optimization on each individual equation in the equation population to obtain the target equation individual.

[0106] Step 680: Use the equation represented by the target equation individual as the business object identification model.

[0107] Step 690: Input the object description data of the business object to be identified into the business object identification model to obtain the predicted type information of the business object to be identified, and determine the business object to be identified as the target business object based on the predicted type information of the business object to be identified.

[0108] It should be noted that the specific limitations of the above steps can be found in the specific limitations of the method for obtaining a business object identification model mentioned above, and will not be repeated here.

[0109] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0110] Based on the same inventive concept, this application also provides a device for acquiring a business object identification model to implement the above-described method for acquiring a business object identification model. The solution provided by this device is similar to the solution described in the above-described method. Therefore, the specific limitations in one or more embodiments of the device for acquiring a business object identification model provided below can be found in the limitations of the method for acquiring a business object identification model described above, and will not be repeated here.

[0111] In one embodiment, such as Figure 7 As shown, a device for obtaining a business object identification model is provided, comprising:

[0112] The acquisition module 710 is used to acquire at least one object description data of a sample business object; each of the object description data has a corresponding business feature attribute.

[0113] The construction module 720 is used to construct an equation population by taking each of the business feature attributes as independent variables; each equation individual in the equation population is used to represent different mapping relationships between the object description data of the sample business object and the prediction category information of the sample business object; the prediction category information is used to describe the prediction probability that the sample business object is the target business object.

[0114] The determination module 730 is used to determine the fitness of each of the equation individuals; the fitness is determined based on the difference between the predicted category information output by the corresponding equation individual for the sample business object and the actual category information of the sample business object;

[0115] Optimization module 740 is used to train a business object recognition model in the equation population according to the fitness of each individual equation;

[0116] The identification module 750 is used to input the object description data of the business object to be identified into the business object identification model, obtain the predicted category information of the business object to be identified, and determine the business object to be identified as the target business object based on the predicted category information of the business object to be identified.

[0117] In one embodiment, the optimization module 740 is specifically used to perform genetic algorithm optimization on each individual of the equation population according to the fitness of each individual of the equation to obtain a target individual of the equation; and to use the equation represented by the target individual of the equation as the business object identification model.

[0118] In one embodiment, the construction module 720 is specifically used to combine the independent variable symbols corresponding to each of the business feature attributes with at least one preset operator to obtain at least two combined equations; the number of symbols in each combined equation meets a preset number threshold; each combined equation is encoded to obtain individual equations; and each individual equation is added to the equation population.

[0119] In one embodiment, the determining module 730 is specifically used to, for any one of the equation individuals, substitute the object description data of the sample business object into the equation represented by the equation individual to obtain the predicted category information output by the equation individual for the sample business object; and determine the fitness corresponding to the equation individual based on the difference between the predicted category information output by the equation individual and the actual category information of the sample business object.

[0120] In one embodiment, the equation population is an Nth generation equation population, where N is greater than or equal to 1. The optimization module 740 is specifically used to generate an N+1th generation equation population based on each equation individual in the Nth generation equation population if all fitness values ​​are less than a preset fitness threshold, and return to the step of determining the fitness corresponding to each equation individual; if there is a target fitness value among the fitness values ​​that is greater than or equal to the fitness threshold, then the equation individual corresponding to the target fitness value is taken as the target equation individual.

[0121] In one embodiment, the optimization module 740 is specifically used to obtain the cumulative fitness value of each equation individual in the Nth generation equation population; select an N+1th generation equation individual from each equation individual in the Nth generation equation population; the fitness value of the N+1th generation equation individual satisfies a preset fitness range; the fitness range is determined based on the cumulative value of each equation individual; and perform evolution operator operations on each N+1th generation equation individual to generate the N+1th generation equation population.

[0122] In one embodiment, the optimization module 740 is specifically configured to, in response to the N+1 generation equation individual satisfying a preset mutation condition, obtain the individual code to be mutated corresponding to the N+1 generation equation individual; change the value of at least one coding bit in the individual code to be mutated to obtain the mutated individual code; and add the equation individual represented by the mutated individual code and the N+1 generation equation individual to the N+1 generation equation population.

[0123] In one embodiment, the optimization module 740 is specifically configured to, in response to at least two of the N+1 generation equation individuals satisfying a preset crossover condition, obtain the crossover codes corresponding to at least two of the N+1 generation equation individuals; determine the code segments in each crossover individual code based on the values ​​corresponding to at least two code bits in each crossover individual code; swap the code segments between each crossover individual code to obtain the crossover individual code; and add the equation individual represented by the crossover individual code and the N+1 generation equation individual to the N+1 generation equation population.

[0124] Each module in the aforementioned business object identification model acquisition device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the operations corresponding to each module.

[0125] In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 8 As shown, the computer device includes a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for acquiring a business object identification model. The display screen can be an LCD screen or an e-ink display screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad mounted on the computer device casing, or an external keyboard, touchpad, or mouse.

[0126] Those skilled in the art will understand that Figure 8 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0127] In one embodiment, a computer device is provided, including a memory and a processor. The memory stores a computer program, which, when executed by the processor, causes the processor to perform the steps of the method for obtaining a business object identification model described above. The steps of the method for obtaining a business object identification model here can be steps from the methods for obtaining a business object identification model described in the various embodiments above.

[0128] In one embodiment, a computer-readable storage medium is provided, storing a computer program that, when executed by a processor, causes the processor to perform the steps of the method for obtaining a business object identification model described above. The steps of the method for obtaining a business object identification model here can be steps from the methods for obtaining a business object identification model described in the various embodiments above.

[0129] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, causes the processor to perform the steps of the above-described method for obtaining a business object identification model. The steps of the method for obtaining a business object identification model here can be steps from the methods for obtaining a business object identification model described in the various embodiments above.

[0130] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties.

[0131] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0132] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0133] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for obtaining a business object identification model, characterized in that, The method includes: Obtain at least one object description data of a sample business object; each of the object description data has a corresponding business feature attribute; Each of the aforementioned business feature attributes is used as an independent variable to construct an equation population; each equation individual in the equation population is used to characterize the different mapping relationships between the object description data of the sample business object and the predicted category information of the sample business object; the predicted category information is used to describe the predicted probability that the sample business object is the target business object. Determine the fitness of each individual equation; the fitness is determined based on the difference between the predicted category information output by the corresponding individual equation for the sample business object and the actual category information of the sample business object; Based on the fitness of each individual equation, a business object recognition model is trained in the equation population. The object description data of the business object to be identified is input into the business object identification model to obtain the predicted category information of the business object to be identified, and the business object to be identified is determined to be the target business object based on the predicted category information of the business object to be identified; the step of generating a business object identification model in the equation population according to the fitness of each equation individual includes: According to the fitness of each individual of the equation, the genetic algorithm is used to optimize each individual of the equation in the equation population to obtain the target equation individual. The genetic algorithm optimization includes selection, elimination, crossover and mutation operations on the equation individuals. The crossover operation is to exchange the coding segments of at least two equation individuals. The mutation operation is to change the value of at least one coding bit in the coding of the equation individual. The equation represented by the target equation individual is used as the business object identification model.

2. The method according to claim 1, characterized in that, The step of constructing an equation population by using each of the aforementioned business feature attributes as independent variables includes: The independent variable symbols corresponding to each of the business feature attributes are combined with at least one preset operator to obtain at least two combined equations; the number of symbols in each of the combined equations meets a preset number threshold. Each of the combined equations is encoded to obtain individual equations. Each individual of the aforementioned equations is added to the equation population.

3. The method according to claim 2, characterized in that, Determining the fitness of each individual in the equations includes: For any one of the equation individuals, substitute the object description data of the sample business object into the equation represented by the equation individual to obtain the prediction category information output by the equation individual for the sample business object. The fitness of any individual equation is determined based on the difference between the predicted category information output by the individual equation and the actual category information of the sample business object.

4. The method according to claim 1, characterized in that, The equation population is the Nth generation equation population, where N is greater than or equal to 1. The step of optimizing each equation individual in the equation population using a genetic algorithm based on the fitness of each equation individual to obtain the target equation individual includes: If all fitness values ​​are less than the preset fitness threshold, then the N+1th generation equation population is generated based on each equation individual in the Nth generation equation population, and the step of determining the fitness value corresponding to each equation individual is returned. If there is a target fitness among the fitness values ​​that is greater than or equal to the fitness threshold, then the equation individual corresponding to the target fitness value is taken as the target equation individual.

5. The method according to any one of claims 1 to 4, characterized in that, The step of generating the (N+1)th generation equation population based on each equation individual in the Nth generation equation population includes: Obtain the cumulative fitness value corresponding to each individual of the equation in the Nth generation equation population; In the Nth generation equation population, select the (N+1)th generation equation individual; the fitness of the (N+1)th generation equation individual satisfies a preset fitness range; the fitness range is determined based on the cumulative value corresponding to each equation individual. Evolutionary operators are performed on each of the N+1 generation equation individuals to generate the N+1 generation equation population.

6. The method according to claim 5, characterized in that, The step of performing evolutionary operator operations on each of the (N+1)th generation equation individuals to generate the (N+1)th generation equation population includes: In response to the N+1 generation equation individual satisfying the preset mutation conditions, the code of the individual to be mutated corresponding to the N+1 generation equation individual is obtained; The value of at least one bit in the code of the individual to be mutated is changed to obtain the code of the mutated individual; The individual representing the equation after mutation and the individual representing the N+1th generation equation are added to the N+1th generation equation population.

7. The method according to claim 6, characterized in that, The step of performing evolutionary operator operations on each of the (N+1)th generation equation individuals to generate the (N+1)th generation equation population includes: In response to at least two of the N+1 generation equation individuals satisfying a preset crossover condition, the codes of the individuals to be crossed corresponding to at least two of the N+1 generation equation individuals are obtained; Based on the values ​​corresponding to at least two coding bits in the coding of each individual to be crossed, determine the coding segment in the coding of each individual to be crossed; Swap the code segments between the codes of the individuals to be crossed to obtain the cross-crossed individual codes; The equation individuals represented by the crossover individual codes and the equation individuals of generation N+1 are added to the equation population of generation N+1.

8. A device for acquiring a business object identification model, characterized in that, The device includes: The acquisition module is used to acquire at least one object description data of a sample business object; each of the object description data has a corresponding business feature attribute. A construction module is used to construct an equation population by taking each of the business feature attributes as independent variables; each equation individual in the equation population is used to represent different mapping relationships between the object description data of the sample business object and the prediction category information of the sample business object; the prediction category information is used to describe the prediction probability that the sample business object is the target business object. A determination module is used to determine the fitness of each of the equation individuals; the fitness is determined based on the difference between the predicted class information output by the corresponding equation individual for the sample business object and the actual class information of the sample business object; The optimization module is used to train a business object recognition model in the equation population according to the fitness of each individual equation. The identification module is used to input the object description data of the business object to be identified into the business object identification model, obtain the predicted category information of the business object to be identified, and determine the business object to be identified as the target business object based on the predicted category information of the business object to be identified. The optimization module is specifically used for: According to the fitness of each individual of the equation, the genetic algorithm is used to optimize each individual of the equation in the equation population to obtain the target equation individual. The genetic algorithm optimization includes selection, elimination, crossover and mutation operations on the equation individuals. The crossover operation is to exchange the coding segments of at least two equation individuals. The mutation operation is to change the value of at least one coding bit in the coding of the equation individual. The equation represented by the target equation individual is used as the business object identification model.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to 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, it implements the steps of the method according to any one of claims 1 to 7.

11. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 7.

Images