Methods, devices, electronic equipment, and storage media for screening business audiences

Abstract

This application discloses a method, apparatus, electronic device, and storage medium for screening business audiences. The method includes: obtaining model scores for detailed business performance samples from multiple models, and performing adaptive binning on each model score to obtain multiple binning intervals corresponding to each model score; encoding the multiple binning intervals of each model score into corresponding gene loci on a chromosome, with the value of each gene locus indicating whether the corresponding binning interval is selected; constructing a fitness function based on business requirements, the fitness function being used to evaluate the business conversion effect of samples screened by rule clusters composed of selected binning intervals on the chromosome; executing a genetic algorithm evolution process; and applying logical AND and logical OR operations to the target audience data using the rule clusters corresponding to the optimal chromosome to screen out the target business audience.

Description

Technical Field

[0001] This application relates to, but is not limited to, the field of computer technology, and in particular to a method, apparatus, electronic device, and storage medium for screening business users. Background Technology

[0002] In sectors such as internet finance and bank lending, companies typically build multiple business-related algorithmic models to predict and evaluate user behavior. However, business objectives are dynamically adjusted over time, and these objectives often have multi-level and hierarchical characteristics. For example, for the same credit product, a user might only click to view, register, complete information, or obtain a credit line; these user behaviors at different levels constitute a multi-objective dataset. To effectively utilize existing models to adapt to such frequently changing business objectives, the industry generally adopts a multi-model fusion strategy.

[0003] Existing technologies typically employ linear fusion methods, directly assigning different weight values ​​to multiple models and then summing the weighted scores of each model to obtain the final fusion score. The weight values ​​can be manually adjusted based on a small sample of detailed business performance data obtained in the short term. However, when business scenarios are complex and user behavior patterns are non-linear, simple linear weighted combinations struggle to accurately fit the true data distribution, leading to insufficient accuracy in business performance prediction. Furthermore, the determination and adjustment of weight values ​​heavily rely on human experience, requiring continuous trial and error, consuming significant human resources, and making it difficult to guarantee optimal results with each adjustment. Summary of the Invention

[0004] In view of this, embodiments of this application provide at least one method, apparatus, electronic device, and storage medium for screening business audiences.

[0005] The technical solution of this application embodiment is implemented as follows: On the one hand, embodiments of this application provide a method for screening business audiences, the method comprising: Obtain model scores for detailed business performance samples from multiple models, and perform adaptive binning on each model score to obtain multiple binning intervals corresponding to each model score; Each model's multiple binning intervals are encoded as corresponding gene loci on a chromosome, and the value of each gene locus indicates whether the corresponding binning interval is selected. A fitness function is constructed based on business requirements. The fitness function is used to evaluate the business conversion effect of samples selected by the rule clusters composed of selected binning intervals on the chromosome. The genetic algorithm evolution process is executed, which includes population initialization, selection, crossover, mutation and iteration operations to obtain at least one optimal chromosome whose fitness satisfies preset conditions; Apply logical AND and logical OR operations to the target population data using the rule clusters corresponding to the optimal chromosome to filter out the target business population.

[0006] On the other hand, embodiments of this application provide a device for screening business audiences, the method comprising: The processing module is used to obtain the model scores of multiple models for detailed business performance samples, and to perform adaptive binning processing on each model score to obtain multiple binning intervals corresponding to each model score. Each model's multiple binning intervals are encoded as corresponding gene loci on a chromosome, and the value of each gene locus indicates whether the corresponding binning interval is selected. A fitness function is constructed based on business requirements. The fitness function is used to evaluate the business conversion effect of samples selected by the rule clusters composed of selected binning intervals on the chromosome. An execution module is used to execute the genetic algorithm evolution process, which includes population initialization, selection, crossover, mutation and iteration operations to obtain at least one optimal chromosome whose fitness satisfies preset conditions; Apply logical AND and logical OR operations to the target population data using the rule clusters corresponding to the optimal chromosome to filter out the target business population.

[0007] In another aspect, embodiments of this application provide a computer device, including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements some or all of the steps in the above-mentioned method for screening business groups.

[0008] In another aspect, embodiments of this application provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements some or all of the steps in the above-described method for screening business groups.

[0009] In another aspect, embodiments of this application provide a computer program including computer-readable code. When the computer-readable code is run in a computer device, the processor in the computer device executes some or all of the steps in the above-mentioned screening method for business groups.

[0010] In another aspect, embodiments of this application provide a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program. When the computer program is read and executed by a computer, it implements some or all of the steps in the above-mentioned method for screening business groups.

[0011] This application embodiment achieves the automatic and efficient combination of multiple heterogeneous models in scenarios where business objectives frequently change by adaptively binning multiple models and encoding them as chromosome gene loci, and then using a genetic algorithm to evolve the optimal rule cluster. This enables the accurate screening of high-potential business conversion groups.

[0012] It should be understood that the above general description and the following detailed description are merely exemplary and explanatory, and are not intended to limit the technical solutions of this application. Attached Figure Description

[0013] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with this application and, together with the specification, serve to explain the technical solutions of this application.

[0014] Figure 1 A schematic diagram illustrating the implementation process of a method for screening business users provided in this application embodiment; Figure 2 A schematic diagram illustrating the structural composition of a screening device for business audiences provided in an embodiment of this application; Figure 3 This is a schematic diagram of the hardware entity of a computer device provided in an embodiment of this application. Detailed Implementation

[0015] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application are further described in detail below with reference to the accompanying drawings and embodiments. The described embodiments should not be regarded as limitations on this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0016] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0017] The terms “first / second / third” are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that “first / second / third” may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0018] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used herein is for descriptive purposes only and is not intended to be limiting of this application.

[0019] This application provides a method for screening business users, which can be executed by the processor of a computer device. The computer device can refer to a server, laptop, tablet, desktop computer, smart TV, set-top box, mobile device (such as a mobile phone, portable video player, personal digital assistant, dedicated messaging device, portable gaming device), or other similar computer equipment. Figure 1 This is a schematic diagram illustrating the implementation process of a method for screening business users provided in an embodiment of this application, as shown below. Figure 1 As shown, the method includes: Step 101: Obtain the model scores of multiple models for detailed business performance samples, and perform adaptive binning on each model score to obtain multiple binning intervals corresponding to each model score.

[0020] In this embodiment, the model score is a numerical result output by the system after predicting the detailed business performance samples, used to quantify the probability or tendency of a sample belonging to a certain category. The detailed business performance samples are historical data sets collected by the system, recording users' specific feedback behaviors on business activities. Adaptive binning is the process by which the system automatically divides the model score into several continuous, non-overlapping numerical intervals based on the actual numerical distribution of the model score. The binning interval is a specific numerical range obtained after adaptive binning.

[0021] The system first obtains the prediction scores of multiple pre-trained algorithm models for detailed business performance samples; these scores are the model scores. For each model's scores, the system analyzes their numerical distribution characteristics and automatically performs adaptive binning based on these characteristics. The core of adaptive binning is that the system does not use fixed numerical intervals, but dynamically determines the bin boundaries based on the data distribution. For example, it may use equal-frequency binning to ensure that each bin contains approximately the same number of samples, or it may perform specific binning based on business rules. Through this process, the system transforms the originally continuous model scores, which could vary greatly, into several discrete bin intervals with clear numerical boundaries, thus laying the foundation for subsequent encoding and optimization.

[0022] Step 102: Encode the multiple binning intervals of each model into corresponding gene loci on the chromosome. The value of each gene locus indicates whether the corresponding binning interval is selected.

[0023] In this embodiment, a chromosome is a data structure used in genetic algorithms to represent a candidate solution, typically consisting of a string of gene bits. A gene bit is a basic unit on a chromosome, and each gene bit corresponds to a specific decision variable. Encoding refers to the process by which the system transforms the decision variables in the actual problem into gene bits on the chromosome. Selection means that the binning interval corresponding to the gene bit is included in the rule combination of the current candidate solution.

[0024] After completing adaptive binning for all models, the system maps the multiple binning intervals corresponding to each model to different gene loci on a chromosome. Specifically, the system assigns a set of contiguous gene loci to each model, with the number of gene loci in the set exactly equal to the number of binning intervals for that model, and each gene locus uniquely corresponds to one binning interval. The value of each gene locus is defined as binary, i.e., 0 or 1. A value of 1 indicates that the binning interval corresponding to that gene locus is selected, meaning that the numerical range represented by that binning interval will be used as part of the filtering criteria; a value of 0 indicates that the binning interval is not selected and is ignored in the rule combination. In this way, the system transforms a complex set of filtering criteria consisting of multiple model binning intervals into a structured binary chromosome that can be manipulated by the genetic algorithm.

[0025] Step 103: Construct a fitness function based on business requirements. The fitness function is used to evaluate the business conversion effect of samples selected by the rule clusters composed of selected binning intervals on the chromosome.

[0026] In this embodiment, business requirements are the specific goals that the business activities served by the system expect to achieve, such as increasing click-through rate or credit granting rate. The fitness function is a mathematical expression used in genetic algorithms to evaluate the superiority or inferiority of chromosomes; its calculation result is called fitness. A rule cluster is a set of selection rules jointly defined by all selected binning intervals on a chromosome. Business conversion effect refers to the performance of samples selected by the rule cluster on business metrics, such as the rate of click occurrence.

[0027] Based on the specific business needs of current business activities, the system constructs a fitness function to quantitatively evaluate the merits of chromosomes. The core function of this function is to assess the performance of the rule cluster represented by a chromosome on actual data. The system applies the rule cluster corresponding to the chromosome to filter detailed samples of business performance, obtaining a subset of samples that meet the criteria. Then, the system statistically analyzes the performance of these samples on key business metrics, such as click-through rate, registration rate, completion rate, or credit granting rate, or linear or non-linear combinations of these metrics. The fitness function calculates a value based on these statistical results; the higher the value, the better the business conversion effect of the samples selected by the rule cluster corresponding to the chromosome, meaning it better aligns with the current business objectives.

[0028] Step 104: Execute the genetic algorithm evolution process, which includes population initialization, selection, crossover, mutation and iteration operations to obtain at least one optimal chromosome whose fitness satisfies the preset conditions.

[0029] In this embodiment, the genetic algorithm evolution process is a computational method that simulates the biological evolution mechanism in nature and finds the optimal solution to a problem through iterative optimization. Population initialization involves the system randomly generating a set of chromosomes as the starting point for evolution. Selection involves the system selecting superior individuals from the current population based on the fitness of chromosomes for reproduction. Crossover involves the system exchanging some gene positions between two selected chromosomes according to certain rules to generate new offspring chromosomes. Mutation involves the system randomly changing the value of a gene position on a chromosome with a very small probability to increase population diversity. Iteration is the process by which the system repeatedly performs selection, crossover, and mutation operations to continuously generate new generations of the population. Preset conditions are termination conditions set by the system to determine whether a chromosome is sufficiently good, such as reaching the maximum number of iterations or a fitness threshold.

[0030] After determining the chromosome encoding method and fitness function, the system begins the genetic algorithm evolution process. The system first initializes the population by randomly generating a batch of binary chromosomes. Then, the system enters an iterative loop. In each generation, the system performs a selection operation, choosing chromosomes with high fitness from the current population as parents based on calculated fitness values. Next, the system performs a crossover operation on the selected parent chromosomes, simulating biological gene recombination to generate new offspring chromosomes. Afterward, the system performs a mutation operation on the offspring chromosomes, randomly flipping 0s or 1s at gene positions with a very low probability, introducing new gene combinations. Through selection, crossover, and mutation operations, the system forms a new generation of chromosomes. The system continues to iterate the above process until at least one optimal chromosome with a fitness value satisfying the preset conditions appears in the population.

[0031] Step 105: Apply logical AND and logical OR operations to the target population data using the rule clusters corresponding to the optimal chromosome to filter out the target business population.

[0032] In this embodiment, the target audience data is a set of new user data for which the system will perform business forecasting. The logical AND operation is a Boolean operation rule where the system determines the result to be true only when multiple conditions are met simultaneously. The logical OR operation is a Boolean operation rule where the system determines the result to be true as long as one condition is met. The target business audience is the user group that the system ultimately selects as having a high conversion rate for business activities.

[0033] After obtaining the optimal chromosome, the system extracts all selected binning intervals to form a complete rule cluster. For the target population data to be predicted, the system first scores it using relevant models and maps each model score to a corresponding binning interval according to the binning boundaries. Then, the system applies this rule cluster for filtering. During the filtering process, the system performs a logical OR operation on multiple selected binning intervals under the same model; that is, as long as the model score of a sample falls within any one of the selected intervals, it is considered to meet the conditions of that model. Next, the system performs a logical AND operation on the conditions of all different models; that is, a sample must simultaneously meet the conditions of all models to be finally selected. Through this series of logical operations, the system accurately filters out users who meet all the filtering conditions from the target population data; these users are the target business population.

[0034] This application embodiment achieves the automatic and efficient combination of multiple heterogeneous models in scenarios where business objectives frequently change by adaptively binning multiple models and encoding them as chromosome gene loci, and then using a genetic algorithm to evolve the optimal rule cluster. This enables the accurate screening of high-potential business conversion groups.

[0035] Optionally, step 101 includes: Step 1011: Divide each model into a preset number of bin intervals according to the equal frequency binning method, and each bin interval contains an equal number of samples.

[0036] In this embodiment, the model score is a numerical value calculated by the system for each sample using an algorithmic model, used to predict or evaluate the sample's performance on a certain target, such as a floating-point number between 0 and 1. Equal-frequency binning is a data discretization method where the system sorts a series of values ​​in ascending order and divides them into several intervals, ensuring that each interval contains approximately the same number of values. A binning interval is a range of values ​​obtained after equal-frequency binning, such as [0.9, 1.0] or [0.75, 0.82]. The preset quantity is an integer pre-set by the system to specify how many binning intervals the model score should be divided into, such as 10 or 20 equal divisions. A sample is a single data point that the system analyzes, such as a complete customer information record.

[0037] The system first obtains the scoring results of each model for a batch of analyzed samples, i.e., the model score corresponding to each sample. Next, for each model, the system sorts the model scores of all its samples. Then, the system divides the sorted model score sequence into 10 bins of equal frequency according to a preset number, such as 10. This equal-frequency binning method ensures that each bin contains the same number of samples; for example, if there are 1000 samples in total, each bin will contain 100 samples.

[0038] Step 1012: When the number of value types in the model score is less than the preset number, divide the bins into bins according to the actual number of value types in the model score, with each value corresponding to one bin.

[0039] In this embodiment, the number of possible values ​​for a model score refers to the total number of different values ​​that the model score can take. For example, if a model score has only six possible values: 0, 1, 2, 3, 4, and 5, then its number of possible values ​​is six. The actual number of possible values ​​is a specific number obtained by the system after counting all the different values ​​that the model score can take.

[0040] The system first determines the number of possible values ​​for a model score. If the system determines that the number of possible values ​​for a model score is less than a preset number (e.g., the preset number is 10, but the model score only has 5 different values), the system will abandon the equal-frequency binning method. Instead, the system will directly divide the bins according to the actual number of possible values ​​for the model score, creating an independent binning interval for each different value. For example, if the model score has 5 possible values ​​(0, 1, 2, 3, 4), the system will create 5 binning intervals, corresponding to samples with values ​​of 0, 1, 2, 3, and 4 respectively.

[0041] Step 1013: Record the fractional boundary information of each bin interval, wherein the fractional boundary information includes a lower limit value and an upper limit value.

[0042] In this embodiment, the fractional boundary information is data used by the system to describe the specific numerical range of a binning interval. The lower limit is the minimum allowed model score value within a binning interval recorded by the system. The upper limit is the maximum allowed model score value within a binning interval recorded by the system.

[0043] After completing the binning interval division, the system extracts and records the boundary information for each interval. This information includes the lower and upper limits of the interval. For example, for a binning interval obtained from equal-frequency binning, its lower limit might be 0.75 and its upper limit might be 0.82; the system will record this pair of values. For binning intervals divided according to the actual value type, such as an interval with a value of 3, both its lower and upper limits are 3, and the system will also record this boundary information.

[0044] This application embodiment achieves the effect of uniformly transforming model scores with different distributions and value ranges into structured binning information that can be used for subsequent genetic algorithm encoding by adaptively binning multiple model scores, discretizing continuous model score values ​​into a finite number of binning intervals, and recording the score boundary information of each interval.

[0045] Optionally, step 102 includes: Step 1021: Assign an independent gene locus on the chromosome to each bin interval of each model. The gene loci corresponding to all bin intervals of all models are arranged in sequence to form a complete chromosome.

[0046] In this embodiment, the model score is a numerical value calculated by the system for each sample using a model, representing the probability or degree to which the sample conforms to a certain target. Binning intervals are several consecutive numerical segments into which the system divides all possible score ranges of a model according to equal frequency or business rules; for example, dividing the score range [0,1] into ten intervals such as [0.9,1.0] and [0.8,0.9]. A chromosome is a data structure representing a candidate solution in a genetic algorithm. A gene locus is the basic unit constituting a chromosome, with each gene locus representing a specific attribute or feature.

[0047] The system identifies multiple models participating in the fusion and obtains a model score for each model. For each model, the system adaptively bins its score range, resulting in several bin intervals. For each bin interval of each model, the system assigns a unique, independent location on a chromosome; this location is a gene locus. The system then arranges the gene loci corresponding to all bin intervals of all models in a pre-defined order, thus assembling a complete chromosome. This chromosome encodes all bin interval information for all models.

[0048] Step 1022: Set the binary value of each gene position to either a first state or a second state. The first state indicates that the corresponding binning interval is selected, and the second state indicates that the corresponding binning interval is not selected.

[0049] In this embodiment, the binary value is one of two possible values ​​assigned by the system to the gene bit, such as 0 or 1. The first state is a specific value in the binary value, such as the value 1. The second state is another specific value in the binary value, such as the value 0.

[0050] After constructing the chromosome, the system assigns a binary value to each gene locus on the chromosome. The system sets this binary value to a first state, such as 1, indicating that the binning interval corresponding to that gene locus is selected as a condition for subsequent data filtering. The system sets this binary value to a second state, such as 0, indicating that the binning interval corresponding to that gene locus is not selected and will be ignored in subsequent data filtering. In this way, the combination of values ​​for all gene loci on a chromosome completely defines a set of filtering rules consisting of multiple binning intervals divided by multiple models.

[0051] This application embodiment, by assigning independent gene loci on chromosomes and setting binary states to each binning interval of each model, can efficiently encode the binning information of multiple models into a chromosome structure that can be processed by genetic algorithms, thereby achieving the effect of transforming the complex multi-model fusion problem into a genetic algorithm optimization problem.

[0052] Optionally, step 103 includes: Step 1031: Obtain all selected binning intervals on the chromosome, construct the logic or conditions within the same model based on the selected binning intervals of each model, and combine the logic or conditions between different models into a rule cluster by performing a logical AND operation.

[0053] In this embodiment, a chromosome is a data structure representing a complete solution in a genetic algorithm, composed of multiple gene loci. A binning interval is the score range corresponding to each bin after binning; for example, the score range [0.9, 1.0] for model A is a binning interval. A logical OR condition is a Boolean operation condition, meaning that the entire condition is true if any one of the multiple sub-conditions is satisfied. A logical AND operation is also a Boolean operation, meaning that the entire operation result is true only if all sub-conditions are satisfied. A rule cluster is a combination of multiple screening conditions used to filter specific population groups from the dataset.

[0054] The system first analyzes a chromosome, identifying all genes with a value of 1 at each genomic locus. These loci correspond to the selected binning intervals. Next, for each model, the system connects all selected binning intervals within that model using a logical OR condition, constructing the model's internal filtering criteria. This means that a sample satisfies the model's filtering criteria as long as it falls within any selected binning interval. Then, the system combines the logical OR conditions of different models using a logical AND operation to form a final rule cluster. This means that a sample must simultaneously satisfy the filtering criteria of all models to be selected by this rule cluster.

[0055] Step 1032: Apply the rule cluster to the business performance detail samples and filter out the sample subset that conforms to the rule cluster.

[0056] In this embodiment, a rule cluster is a combination of multiple model filtering conditions. The business performance detail sample is a batch of historical data whose business performance is known, such as samples recording whether each customer has clicked, registered, or engaged in other behaviors. The sample subset is a selection of samples chosen from the business performance detail sample.

[0057] The system applies the generated rule clusters to the pre-prepared sample data of business performance details. The system examines each sample in the data one by one to determine if it simultaneously satisfies the logic or conditions of all models within the rule cluster. Only those samples that fully satisfy all conditions are selected by the system to form a smaller subset of samples.

[0058] Step 1033: Count the number of business conversion behaviors in the sample subset, whereby the business conversion behaviors include at least one of click behavior, registration behavior, completion behavior, and credit authorization behavior.

[0059] In this embodiment, the sample subset is a selected set of samples that meet the criteria. Business conversion behavior refers to specific actions that users take during business activities and that have business value, such as clicking an advertisement, registering an account, completing an application, or obtaining a credit line.

[0060] The system analyzes the obtained sample subset. It counts the number of times each sample in the subset performs a specified business conversion action. For example, the system can count how many samples in the subset clicked, how many registered, and so on. The system can count the number of one or more business conversion actions.

[0061] Step 1034: Calculate the fitness value based on the number of business conversion behaviors in the statistics. The fitness value is at least one of click-through rate, registration rate, completion rate, and credit granting rate or a linear combination thereof.

[0062] In this embodiment, the fitness value is a quantitative indicator used in genetic algorithms to evaluate the quality of a chromosome. Click-through rate (CTR) refers to the proportion of samples that have performed a click to the total number of samples in a subset. Registration rate, completion rate, and credit granting rate are similar concepts, referring to the proportion of samples performing the corresponding action to a specific base. Linear combination refers to combining multiple ratio values ​​into a single value through a weighted summation.

[0063] The system calculates one or more ratios, such as click-through rate (CTR) and registration rate, based on the statistically obtained number of business conversion behaviors and the total number of samples in the sample subset. Then, the system uses these ratios to perform calculations according to a preset fitness function. If the fitness function is simply CTR, then the fitness value equals the calculated CTR. If the fitness function is a linear combination of CTR and confidence rate, such as 0.6 CTR + 0.4 confidence rate, the system will calculate the final fitness value according to this formula. This fitness value will be used in the subsequent individual selection process of the genetic algorithm.

[0064] This application embodiment encodes the binning intervals of the model into chromosome gene loci and performs logical combination, then calculates the fitness value by combining business performance data, and quantitatively evaluates the rule clusters. This achieves the effect of automatically and efficiently searching for the optimal combination of model binning interval rules, thereby realizing the accurate screening of the target audience.

[0065] Optionally, step 1034 includes: Step 10341: Calculate the total number of users in the sample subset, calculate the number of users in the sample subset who clicked, and divide the number of users who clicked by the total number of users to obtain the click-through rate.

[0066] In this embodiment, the sample subset is a dataset filtered by the system based on chromosomal gene loci. Users in this dataset are those whose scores meet the condition that all gene loci in the chromosome are 1. The total number of users is the total number of users in this sample subset. Click behavior refers to a specific action performed by a user on a target product during business activities, such as clicking to view a page. Click-through rate (CTR) is an indicator that measures the frequency of user click behavior within the sample subset.

[0067] The system obtains a sample subset filtered by chromosomal gene loci. It then counts the total number of users within this subset. Next, the system identifies and counts the number of users within this subset who clicked on a page. This click behavior is predefined based on business requirements, such as a user clicking on a credit product page. Finally, the system performs a division operation, using the number of users who clicked as the divisor and the total number of users as the divisor. The result is the click-through rate.

[0068] Step 10342, or, calculate the number of users who have engaged in credit granting behavior in the sample subset, and divide the number of users who have engaged in credit granting behavior by the total number of users to obtain the credit granting rate.

[0069] In this embodiment, credit granting refers to the act of a user being granted a credit limit by the system or financial institution after completing a credit product application. The credit granting rate is an indicator that measures the frequency of credit granting actions by users within a sample subset.

[0070] After obtaining a sample subset, the system identifies and counts the number of users within that subset who have engaged in credit granting behavior. This credit granting behavior refers to a user completing the application process and ultimately receiving a credit limit. Next, the system uses the number of users who engaged in credit granting behavior as the dividend and the total number of users as the divisor, performing a division operation. The result is the credit granting rate.

[0071] Step 10343, or, multiply the click-through rate by the second weight to obtain the first product, multiply the credit rate by the third weight to obtain the second product, and add the first product and the second product to obtain the combined fitness value.

[0072] In this embodiment, the second weight is a pre-set numerical coefficient for click-through rate (CTR) to adjust its importance in the combined fitness value. The first product is the result of multiplying the CTR by the second weight. The third weight is a pre-set numerical coefficient for credit rate (KR) to adjust its importance in the combined fitness value. The second product is the result of multiplying the KR by the third weight. The combined fitness value is a comprehensive evaluation metric used to quantify the overall performance of the rule combination represented by the chromosome across multiple business objectives.

[0073] The system obtains the calculated click-through rate and trust rate. The click-through rate is multiplied by a preset second weight to obtain the first product. The trust rate is multiplied by a preset third weight to obtain the second product. The system then performs an addition operation, adding the first and second products together. The result is the fitness value of the combination corresponding to that chromosome.

[0074] This application embodiment encodes multi-model binning information into chromosome gene loci and performs genetic algorithm evolution. Then, it combines click-through rate, credit rate, or a weighted combination thereof as a fitness function to evaluate the chromosomes. This achieves the technical effect of automatically selecting the optimal combination of multi-model rules, thereby enabling efficient and accurate positioning of target audiences and improving business conversion results in scenarios where business objectives change frequently.

[0075] Optionally, step 104 includes: Step 1041: Randomly generate an initial population containing multiple chromosomes, with the gene loci on each chromosome randomly set to either the first state or the second state.

[0076] In this embodiment, the initial population is a set of candidate solutions randomly generated by the system in the genetic algorithm to begin evolutionary computation. Each candidate solution is a chromosome. A chromosome is an encoding structure used in the genetic algorithm to represent a complete solution, consisting of a series of gene bits. A gene bit is the basic unit on a chromosome, representing a decision variable or attribute in the solution. The first state and the second state are two possible values ​​of the gene bit, representing whether the attribute represented by that gene bit is selected or ignored, respectively.

[0077] The system initializes the evolutionary process of the genetic algorithm. It randomly generates an initial population containing multiple chromosomes. For each gene locus on each chromosome, the system randomly assigns its value to either a first state or a second state; for example, the first state can be represented by the value 1, and the second state by the value 0. This randomization ensures diversity of solutions in the initial population, providing a broad starting point for subsequent evolutionary searches.

[0078] Step 1042: Calculate the fitness value of each chromosome in the population, and select the chromosome with the highest fitness value as the parent chromosome.

[0079] In this embodiment, the fitness value is a numerical value obtained by the system after quantitatively evaluating the merits of the solution represented by each chromosome according to a preset fitness function. The parent chromosome is the chromosome selected by the system from the current population based on its fitness value to generate the next generation solution.

[0080] After generating the initial population, the system calculates the fitness value for each chromosome in the population. This fitness value reflects the quality of the rule combination represented by that chromosome in solving a specific problem. Next, the system sorts all chromosomes according to their fitness values ​​and selects the chromosomes with the highest fitness values ​​as parent chromosomes. These selected parent chromosomes represent the better solutions in the current population and will be used to generate potentially better solutions for the next generation.

[0081] Step 1043: Perform single-point or multi-point crossover on the parent chromosome to generate the offspring chromosome.

[0082] In the embodiments of this application, single-point crossover involves the system randomly selecting a crossover point on the gene sequence of the parent chromosome and then exchanging all gene loci after that point to generate two offspring chromosomes. Multi-point crossover involves the system randomly selecting multiple crossover points on the gene sequence of the parent chromosome and then exchanging gene segments between these crossover points to generate offspring chromosomes. Offspring chromosomes are new chromosomes generated by the system through crossover operations on the parent chromosomes; they inherit some of the genetic characteristics of the parent chromosomes.

[0083] The system performs a crossover operation on selected parent chromosomes to generate new offspring chromosomes. The system can perform single-point or multi-point crossover. In single-point crossover, the system randomly selects a gene locus as a boundary point and then exchanges all gene loci after that boundary point between the two parent chromosomes, thus generating two new offspring chromosomes. In multi-point crossover, the system randomly selects multiple gene loci as boundary points and then exchanges gene segments between these boundary points between the two parent chromosomes. Through crossover operations, the system can combine desirable genes from parent chromosomes to explore new solution spaces.

[0084] Step 1044: Perform a state-flipping mutation operation on the gene loci of the offspring chromosome with a preset mutation probability to obtain a new generation population.

[0085] In this embodiment, the preset mutation probability is a pre-set value used to control the likelihood of state flipping at each gene locus. State flipping refers to the operation of changing the current value of a gene locus from a first state to a second state, or from a second state to a first state. The new generation population is a new set of chromosomes generated by the system based on the current population through a series of operations such as selection, crossover, and mutation, for the next round of evolution.

[0086] The system performs mutation operations on the generated offspring chromosomes. For each gene locus on each offspring chromosome, the system determines whether to flip the state based on a preset mutation probability. If the randomly generated value is less than the preset mutation probability, the system flips the state of that gene locus, for example, changing the first state to the second state, or vice versa. The mutation operation introduces randomness, helping the system escape local optima and explore a wider solution space. After the mutation operation, all offspring chromosomes together constitute the new generation population.

[0087] Step 1045: Repeat the selection, crossover, and mutation operations until the preset number of iterations or fitness value converges, and output at least one optimal chromosome with the highest fitness value.

[0088] In this embodiment, the preset iteration number is the maximum number of evolutionary rounds set before the system begins evolution, used to control the algorithm's running time. Fitness convergence means that the fitness values ​​of all chromosomes in the population no longer change significantly after multiple generations of evolution, indicating that the algorithm has found a relatively stable optimal solution region. The optimal chromosome is one or more chromosomes with the highest fitness values ​​selected by the system from the final population after evolution, representing the best solution to the problem.

[0089] The system repeatedly performs the selection in step 1042, the crossover in step 1043, and the mutation in step 1044, forming an iterative evolutionary loop. The system continues this loop until a preset termination condition is met. The termination condition can be reaching a preset number of iterations, such as 100 generations; or observing convergence of fitness values, i.e., the fitness values ​​of all chromosomes in the population change very little over multiple generations. When any termination condition is met, the system stops evolving. Finally, the system outputs at least one optimal chromosome with the highest fitness value from the final population; these chromosomes represent the best rule combination obtained through genetic algorithm optimization.

[0090] This application embodiment uses genetic algorithm operations such as random initialization of the population, selection based on fitness values, crossover, and mutation, and iterative evolution until convergence. The system can automatically and efficiently search for the optimal rule combination from a large number of model score bin combinations, achieving the effect of quickly adapting to changes in business objectives and improving business performance without manual intervention.

[0091] Optionally, step 105 includes: Step 1051: Obtain the selected binning intervals for each model segment in the optimal chromosome. For the same model segment, perform a logical OR operation on all selected binning intervals to obtain the screening conditions for that model segment.

[0092] In this embodiment, the optimal chromosome is the individual chromosome with the highest fitness or that meets preset conditions, determined by the fitness function after the genetic algorithm evolution. It represents a set of optimal model score binning interval combination rules. The model score is the predicted score output by the algorithm model for each sample, used to measure the probability that the sample conforms to the model's objective. Binning intervals are several continuous and non-overlapping score ranges divided after sorting the model scores from largest to smallest, such as equal-frequency binning or division according to business rules. The logical OR operation is a Boolean operation used to combine multiple conditions; as long as any one of the conditions is met, the overall result is true.

[0093] The system first analyzes the optimal chromosome, which consists of multiple gene loci. Each gene locus corresponds to a binning interval for a model score, and a gene locus of 1 indicates that the binning interval is selected. For the same model score, the system extracts all selected binning intervals, which are the score ranges where all gene loci are 1 for that model score. Subsequently, the system performs a logical OR operation on these selected binning intervals. This means that as long as a sample's model score falls within any of the selected binning intervals, it is considered to meet the selection criteria for that model score, thus forming the final selection criteria for that model score.

[0094] Step 1052: Perform a logical AND operation on the filtering conditions of different models to form a single rule cluster.

[0095] In this embodiment, different model scores refer to the scores output by multiple prediction models trained from different sources or algorithms, such as the score of model A, the score of model B, etc. A screening condition is a score range condition set for a single model score, used to determine whether a sample passes the model's screening. The logical AND operation is a Boolean operation used to combine multiple conditions; all conditions must be satisfied simultaneously for the overall result to be true. A single rule cluster is a complete set of rules derived from an optimal chromosome, containing multiple model score screening conditions.

[0096] The system performs a logical AND operation on the screening conditions of these different model subclasses. This means that a sample must simultaneously meet the screening conditions of all model subclasses to be considered as conforming to the requirements of that rule cluster. In this way, the system combines the screening conditions of multiple model subclasses into a logical AND relationship, forming a complete and indivisible set of rules, i.e., a single rule cluster.

[0097] Step 1053: Apply a single rule cluster to the target population data to filter out the first target population that simultaneously meets all model screening conditions.

[0098] In this embodiment, the target audience data is a dataset of potential customers that the system needs to filter using rule clusters, but which have not yet been processed by the business. The first target audience is a subset selected from the target audience data after applying a single rule cluster, wherein the samples in this subset simultaneously meet the filtering conditions of all model scores in the rule cluster.

[0099] The system applies the generated single rule cluster to the target audience data. The system iterates through each sample in the target audience data, checking whether each model score of the sample falls within the corresponding model score filtering criteria. Only when all model scores of a sample meet its corresponding filtering criteria is the sample selected. Ultimately, the set of all selected samples constitutes the first target audience, which consists of potential customers who best match the business objectives of that rule cluster, strictly filtered according to that single rule cluster.

[0100] Step 1054: When there are multiple optimal chromosomes, perform a logical OR operation on the first target population selected by the single rule cluster corresponding to each optimal chromosome, and merge them to obtain the second target population.

[0101] In this embodiment, multiple optimal chromosomes refer to the genetic algorithm that, after evolution, may generate more than one chromosome individual with high fitness, each representing a superior combination of rules. The logical OR operation is used to merge the screening results of multiple rule clusters; a sample is included in the result set as long as it meets the conditions of any one of the rule clusters. The second target population is a broader set of people containing multiple potential business targets, obtained by merging the screening results of the rule clusters corresponding to multiple optimal chromosomes.

[0102] When the genetic algorithm generates multiple optimal chromosomes, the system executes steps 1051 to 1053 for each optimal chromosome, resulting in multiple distinct first target populations. Each first target population corresponds to a different rule cluster. To expand business coverage, the system performs a logical OR operation to merge these first target populations. This means that as long as a sample belongs to any first target population, it will be included in the final set. The merged result is the second target population, representing a more comprehensive potential customer group that the system can locate under the guidance of multiple optimal rule clusters.

[0103] This application embodiment parses the optimal chromosomes generated by the genetic algorithm into model screening conditions, combines them into rule clusters using logical OR and AND operations, and then merges the results of multiple rule clusters to achieve the effect of accurately screening and expanding the target business population.

[0104] Based on the foregoing embodiments, this application provides a screening device for business audiences. The device includes various units and modules included in each unit, which can be implemented by a processor in a computer device; of course, it can also be implemented by specific logic circuits. In the implementation process, the processor can be a central processing unit (CPU), a microprocessor unit (MPU), a digital signal processor (DSP), or a field programmable gate array (FPGA), etc.

[0105] Figure 2 This application provides a schematic diagram of the composition of a business audience screening device, as shown in the embodiments of this application. Figure 2 As shown, the screening device 20 for the target audience includes: The processing module 201 is used to obtain the model scores of multiple models for detailed business performance samples, and to perform adaptive binning processing on each model score to obtain multiple binning intervals corresponding to each model score. Each model's multiple binning intervals are encoded as corresponding gene loci on a chromosome, and the value of each gene locus indicates whether the corresponding binning interval is selected. A fitness function is constructed based on business requirements. The fitness function is used to evaluate the business conversion effect of samples selected by the rule clusters composed of selected binning intervals on the chromosome. The execution module 202 is used to execute the genetic algorithm evolution process, which includes population initialization, selection, crossover, mutation and iteration operations to obtain at least one optimal chromosome whose fitness satisfies preset conditions; Apply logical AND and logical OR operations to the target population data using the rule clusters corresponding to the optimal chromosome to filter out the target business population.

[0106] Optionally, the processing module 201 is further configured to: Each model is divided into a preset number of bins using an equal-frequency binning method, and each bin contains an equal number of samples. When the number of possible values ​​for the model score is less than the preset number, the binning intervals are divided according to the actual number of possible values ​​for the model score, with each value corresponding to one binning interval. Record the fractional boundary information for each bin interval, including a lower limit and an upper limit.

[0107] Optionally, the processing module 201 is further configured to: Each binning interval of each model is assigned an independent gene locus on a chromosome, and the gene loci corresponding to all binning intervals of all models are arranged sequentially to form a complete chromosome. The binary value of each gene position is set to either a first state or a second state. The first state indicates that the corresponding binning interval is selected, and the second state indicates that the corresponding binning interval is not selected.

[0108] Optionally, the processing module 201 is further configured to: Obtain all selected binning intervals on the chromosome, construct the logic or conditions within the same model based on the selected binning intervals of each model, and combine the logic or conditions between different models into a rule cluster by performing a logical AND operation. The rule clusters are applied to the business performance detail samples to filter out a subset of samples that conform to the rule clusters; The number of business conversion behaviors in the sample subset is counted, and the business conversion behaviors include at least one of click behavior, registration behavior, completion behavior and credit authorization behavior; The fitness value is calculated based on the number of business conversion behaviors counted. The fitness value is at least one of click-through rate, registration rate, completion rate, and credit granting rate, or a linear combination thereof.

[0109] Optionally, the processing module 201 is further configured to: Calculate the total number of users in the sample subset, calculate the number of users in the sample subset who clicked, and divide the number of users who clicked by the total number of users to obtain the click-through rate; Alternatively, the number of users who have engaged in credit granting behavior in the sample subset can be calculated, and the credit granting rate can be obtained by dividing the number of users who have engaged in credit granting behavior by the total number of users. Alternatively, the click-through rate can be multiplied by the second weight to obtain a first product, the credit rate can be multiplied by the third weight to obtain a second product, and the first product and the second product can be added together to obtain a combined fitness value.

[0110] Optionally, the execution module 202 is further configured to: An initial population containing multiple chromosomes is randomly generated, and the gene loci on each chromosome are randomly set to either the first state or the second state. Calculate the fitness value of each chromosome in the population, and select the chromosome with the highest fitness value as the parent chromosome based on the fitness value. Perform single-point or multi-point crossover operations on the parent chromosomes to generate offspring chromosomes; The gene loci of the offspring chromosome are subjected to a state-flipping mutation operation with a preset mutation probability to obtain a new generation population. Repeat the selection, crossover, and mutation operations until the preset number of iterations or fitness value converges, and output at least one optimal chromosome with the highest fitness value.

[0111] Optionally, the execution module 202 is further configured to: Obtain the selected binning intervals for each model segment in the optimal chromosome. For the same model segment, perform a logical OR operation on all selected binning intervals to obtain the filtering conditions for that model segment. Perform a logical AND operation on the filtering conditions of different models to form a single rule cluster; Apply a single rule cluster to the target population data to filter out the first target population that simultaneously meets all model screening conditions; When multiple optimal chromosomes exist, the first target population selected by the single rule cluster corresponding to each optimal chromosome is logically ORed to obtain the second target population.

[0112] This application embodiment achieves the automatic and efficient combination of multiple heterogeneous models in scenarios where business objectives frequently change by adaptively binning multiple models and encoding them as chromosome gene loci, and then using a genetic algorithm to evolve the optimal rule cluster. This enables the accurate screening of high-potential business conversion groups.

[0113] The descriptions of the apparatus embodiments above are similar to those of the method embodiments above, and have similar beneficial effects. In some embodiments, the functions or modules included in the apparatus provided in this application can be used to perform the methods described in the method embodiments above. For technical details not disclosed in the apparatus embodiments of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0114] It should be noted that, in the embodiments of this application, if the above-mentioned method for screening business users is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiments of this application, or the part that contributes to the related technology, can be embodied in the form of a software product. This software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, mobile hard drives, read-only memory (ROM), magnetic disks, or optical disks. Thus, the embodiments of this application are not limited to any specific hardware, software, or firmware, or any combination of hardware, software, and firmware.

[0115] This application provides a computer device including a memory and a processor. The memory stores a computer program that can run on the processor. When the processor executes the program, it implements some or all of the steps in the above-described method.

[0116] This application provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements some or all of the steps in the above-described method. The computer-readable storage medium can be transient or non-transient.

[0117] This application provides a computer program including computer-readable code, wherein when the computer-readable code is executed in a computer device, a processor in the computer device performs some or all of the steps in the above-described method.

[0118] This application provides a computer program product, which includes a non-transitory computer-readable storage medium storing a computer program. When the computer program is read and executed by a computer, it implements some or all of the steps in the above-described method. This computer program product can be implemented specifically through hardware, software, or a combination thereof. In some embodiments, the computer program product is specifically embodied as a computer storage medium; in other embodiments, the computer program product is specifically embodied as a software product, such as a software development kit (SDK), etc.

[0119] It should be noted that the descriptions of the various embodiments above tend to emphasize the differences between them, while their similarities or commonalities can be referred to interchangeably. The descriptions of the above embodiments of the device, storage medium, computer program, and computer program product are similar to the descriptions of the above method embodiments and have similar beneficial effects. For technical details not disclosed in the embodiments of the device, storage medium, computer program, and computer program product of this application, please refer to the descriptions of the method embodiments of this application for understanding.

[0120] It should be noted that, Figure 3 This is a schematic diagram of a hardware entity of a computer device in an embodiment of this application, such as... Figure 3 As shown, the hardware entity of the computer device 700 includes: one or more processors 701, a communication interface 702, and a memory 703, wherein: Processor 701 typically controls the overall operation of computer device 700.

[0121] Communication interface 702 enables computer devices to communicate with other terminals or servers over a network.

[0122] The memory 703 is configured to store instructions and applications executable by the processor 701, and can also cache data to be processed or already processed (e.g., image data, audio data, voice communication data, and video communication data) in the processor 701 and various modules in the computer device 700. It can be implemented using flash memory or random access memory (RAM). Data transfer between the processor 701, the communication interface 702, and the memory 703 can be performed via bus 704. Only one processor is shown in the figure; each processor 700 includes one or more cores.

[0123] It should be noted that the computer device may include multiple processors 701, and each processor 701 can interact with each other through aggregated communication methods such as all-to-all, all-gather, or all-reduce. The processors 701 may be central processing units (CPUs), graphics processing units (GPUs), embedded neural network processing units (NPUs), tensor processing units (TPUs), data processing units (DPUs), accelerated processing units (APUs), floating-point processing units (FPUs), or application-specific integrated circuits (ASICs). The processors may also be single-core or multi-core processors. The processor may consist of a CPU and hardware chips. The hardware chips may be ASICs, PLDs, or combinations thereof. The PLDs may be complex programmable logic devices (CPLDs), FPGAs, generic array logic (GALs), or any combination thereof. The processor can also be implemented using logic devices with built-in processing logic, such as FPGAs or digital signal processors (DSPs).

[0124] The communication interface 702 can be a wired interface or a wireless interface, used to communicate with other modules or devices. The wired interface can be an Ethernet interface, a local interconnect network (LIN), etc., and the wireless interface can be a cellular network interface or a wireless LAN interface, etc.

[0125] Memory 703 can be non-volatile memory, such as read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. Memory 703 can also be volatile memory, which can be random access memory (RAM) used as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synclink dynamic random access memory (SLDRAM), and direct rambus RAM (DRRAM), direct rambus DRAM (DRDRAM), and rambus DRAM.

[0126] The 704 bus can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus can be divided into address bus, data bus, control bus, etc.

[0127] It should be understood that the phrase "one embodiment" or "an embodiment" throughout the specification means that a specific feature, structure, or characteristic related to the embodiment is included in at least one embodiment of this application. Therefore, "in one embodiment" or "in an embodiment" appearing throughout the specification does not necessarily refer to the same embodiment. Furthermore, these specific features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. It should be understood that in the various embodiments of this application, the sequence numbers of the above steps / processes do not imply a sequential order of execution; the execution order of each step / process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application. The sequence numbers of the above embodiments of this application are merely descriptive and do not represent the superiority or inferiority of the embodiments.

[0128] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0129] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods, such as: multiple units or components can be combined, or integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the various components shown or discussed can be through some interfaces, and the indirect coupling or communication connection between devices or units can be electrical, mechanical, or other forms.

[0130] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units. They may be located in one place or distributed across multiple network units. Some or all of the units may be selected to achieve the purpose of this embodiment according to actual needs.

[0131] In addition, each functional unit in the various embodiments of this application can be integrated into one processing unit, or each unit can be a separate unit, or two or more units can be integrated into one unit; the integrated unit can be implemented in hardware or in the form of hardware plus software functional units.

[0132] Those skilled in the art will understand that all or part of the steps of the above method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it performs the above method embodiments. The aforementioned storage medium includes various media that can store program code, such as mobile storage devices, read-only memory (ROM), magnetic disks, or optical disks.

[0133] Alternatively, if the integrated units described above are implemented as software functional modules and sold or used as independent products, they can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, or the part that contributes to related technologies, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as mobile storage devices, ROM, magnetic disks, or optical disks.

[0134] The above description is merely an embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for screening business audiences, characterized in that, The method includes: Obtain model scores for detailed business performance samples from multiple models, and perform adaptive binning on each model score to obtain multiple binning intervals corresponding to each model score; Each model's multiple binning intervals are encoded as corresponding gene loci on a chromosome, and the value of each gene locus indicates whether the corresponding binning interval is selected. A fitness function is constructed based on business requirements. The fitness function is used to evaluate the business conversion effect of samples selected by the rule clusters composed of selected binning intervals on the chromosome. The genetic algorithm evolution process is executed, which includes population initialization, selection, crossover, mutation and iteration operations to obtain at least one optimal chromosome whose fitness satisfies preset conditions; Apply logical AND and logical OR operations to the target population data using the rule clusters corresponding to the optimal chromosome to filter out the target business population.

2. The method according to claim 1, characterized in that, The adaptive binning process for each model score yields multiple bin intervals corresponding to each model score, including: Each model is divided into a preset number of bins using an equal-frequency binning method, and each bin contains an equal number of samples. When the number of possible values ​​for the model score is less than the preset number, the binning intervals are divided according to the actual number of possible values ​​for the model score, with each value corresponding to one binning interval. Record the fractional boundary information for each bin interval, including a lower limit and an upper limit.

3. The method according to claim 1, characterized in that, The process of encoding the multiple binning intervals of each model into corresponding gene loci on the chromosome includes: Each binning interval of each model is assigned an independent gene locus on a chromosome, and the gene loci corresponding to all binning intervals of all models are arranged sequentially to form a complete chromosome. The binary value of each gene position is set to either a first state or a second state. The first state indicates that the corresponding binning interval is selected, and the second state indicates that the corresponding binning interval is not selected.

4. The method according to claim 1, characterized in that, The construction of the fitness function based on business requirements includes: Obtain all selected binning intervals on the chromosome, construct the logic or conditions within the same model based on the selected binning intervals of each model, and combine the logic or conditions between different models into a rule cluster by performing a logical AND operation. The rule clusters are applied to the business performance detail samples to filter out a subset of samples that conform to the rule clusters; The number of business conversion behaviors in the sample subset is counted, and the business conversion behaviors include at least one of click behavior, registration behavior, completion behavior and credit authorization behavior; The fitness value is calculated based on the number of business conversion behaviors counted. The fitness value is at least one of click-through rate, registration rate, completion rate, and credit granting rate, or a linear combination thereof.

5. The method according to claim 4, characterized in that, The calculation of the fitness value based on the statistical number of business conversion behaviors includes: Calculate the total number of users in the sample subset, calculate the number of users in the sample subset who clicked, and divide the number of users who clicked by the total number of users to obtain the click-through rate; Alternatively, the number of users who have engaged in credit granting behavior in the sample subset can be calculated, and the credit granting rate can be obtained by dividing the number of users who have engaged in credit granting behavior by the total number of users. Alternatively, the click-through rate can be multiplied by the second weight to obtain a first product, the credit rate can be multiplied by the third weight to obtain a second product, and the first product and the second product can be added together to obtain a combined fitness value.

6. The method according to claim 1, characterized in that, The genetic algorithm evolution process includes: An initial population containing multiple chromosomes is randomly generated, and the gene loci on each chromosome are randomly set to either the first state or the second state. Calculate the fitness value of each chromosome in the population, and select the chromosome with the highest fitness value as the parent chromosome based on the fitness value. Perform single-point or multi-point crossover operations on the parent chromosomes to generate offspring chromosomes; The gene loci of the offspring chromosome are subjected to a state-flipping mutation operation with a preset mutation probability to obtain a new generation population. Repeat the selection, crossover, and mutation operations until the preset number of iterations or fitness value converges, and output at least one optimal chromosome with the highest fitness value.

7. The method according to claim 1, characterized in that, The process of applying logical AND and logical OR operations to the target population data using the rule clusters corresponding to the optimal chromosome to filter out the target business population includes: Obtain the selected binning intervals for each model segment in the optimal chromosome. For the same model segment, perform a logical OR operation on all selected binning intervals to obtain the filtering conditions for that model segment. Perform a logical AND operation on the filtering conditions of different models to form a single rule cluster; Apply a single rule cluster to the target population data to filter out the first target population that simultaneously meets all model screening conditions; When multiple optimal chromosomes exist, the first target population selected by the single rule cluster corresponding to each optimal chromosome is logically ORed to obtain the second target population.

8. A device for screening business audiences, characterized in that, The device includes: The processing module is used to obtain the model scores of multiple models for detailed business performance samples, and to perform adaptive binning processing on each model score to obtain multiple binning intervals corresponding to each model score. Each model's multiple binning intervals are encoded as corresponding gene loci on a chromosome, and the value of each gene locus indicates whether the corresponding binning interval is selected. A fitness function is constructed based on business requirements. The fitness function is used to evaluate the business conversion effect of samples selected by the rule clusters composed of selected binning intervals on the chromosome. An execution module is used to execute the genetic algorithm evolution process, which includes population initialization, selection, crossover, mutation and iteration operations to obtain at least one optimal chromosome whose fitness satisfies preset conditions; Apply logical AND and logical OR operations to the target population data using the rule clusters corresponding to the optimal chromosome to filter out the target business population.

9. A computer device comprising a memory and a processor, the memory storing a computer program executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the screening method for the business population as 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 executed by a processor, the computer program implements the steps of the screening method for the business population as described in any one of claims 1 to 7.

Images