Object stratification method, device, equipment and medium based on artificial intelligence

By constructing and verifying the positive sample hit rate of the preset model, the problem of unstable model output in the prediction of high-quality increase in the object is solved, the accuracy of population hierarchy and rationality of task management is achieved, and the model is improved Stability and effect.

CN114511033BActive Publication Date: 2024-11-26CHINA PING AN LIFE INSURANCE CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210140982.9
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-02-16
Publication Date
2024-11-26
Estimated Expiration
2042-02-16

AI Technical Summary

Technical Problem

In the prediction scenario of high-quality object recruitment subjects, due to the large time interval between the training time of the model and the actual use time, the model output accuracy is unstable, which affects the reasonable hierarchy of the object and task management.

Method used

By obtaining the historical staffing data of the object, building a sample set and training a preset model, calculating the positive sample hit rate of each level, verifying the rationality of the hit rate of each layer to be evaluated, determining the target model, and predicting the population to be stratified The positive sample hit rate is used and stratified according to the predicted probability and hit rate to achieve hierarchical division of the population.

Benefits of technology

It improves the stability of model output and the accuracy of population hierarchy, ensures the rationality of task allocation and management, and enhances the effect of model training.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114511033B_ABST
    Figure CN114511033B_ABST
Patent Text Reader

Abstract

The present invention relates to the field of artificial intelligence, and provides an object stratification method, device, equipment and medium based on artificial intelligence, which can determine at least one layer to be evaluated according to the positive sample hit rate of each layer, realize the division of the initial level, facilitate subsequent evaluation and adjustment, verify the rationality of the positive sample hit rate of each layer to be evaluated in turn, facilitate subsequent correction to ensure the rationality of the trained model stratification, and perform stratification processing on the stratified personnel according to the positive sample hit rate of the stratified population, the prediction probability corresponding to each stratified person and the positive sample hit rate of each stratified layer to obtain a stratification result, realize the hierarchical division of the population, and facilitate subsequent targeted management and task allocation. The present invention also relates to blockchain technology, and the stratification result can be stored on a blockchain node.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to an object stratification method, device, equipment and medium based on artificial intelligence. Background Art

[0002] In actual application scenarios, most of the time, the training time of the model and the time it is actually put into use are different, and there is usually a large time difference between the two.

[0003] For example, in the prediction scenario of high-quality recruitment subjects (such as insurance agents, etc.), the proportion of high-quality recruitment subjects fluctuates between 10% and 20%, and the proportion of positive samples in the predicted population fluctuates greatly. It can be seen that due to the large time interval between the training time of the model and the actual use time, the output accuracy of the model is unstable, which is extremely unfavorable for promotion and application in application scenarios.

[0004] The above results will lead to unstable accuracy in the prediction of the main body of high-quality recruitment for the objects, and thus make it impossible to reasonably stratify the objects, which will ultimately lead to unreasonable tasks allocation and management of the objects and affect the overall task completion of the object group. Summary of the invention

[0005] The embodiments of the present invention provide an artificial intelligence-based object stratification method, device, equipment and medium, which aim to solve the problem of low accuracy in predicting the subject of high-quality recruitment of objects and inability to stratify objects.

[0006] In a first aspect, an embodiment of the present invention provides an object stratification method based on artificial intelligence, which includes:

[0007] Obtain historical recruitment data of the object, and use the historical recruitment data to construct a sample set, split the sample set to obtain a training set and a validation set;

[0008] Acquire a preset model with classification function, and train the preset model using the training set to obtain an intermediate model;

[0009] Inputting the validation set into the intermediate model to obtain the predicted probability of each sample in the validation set;

[0010] Sort each sample in the validation set in descending order of the predicted probability to obtain a sample sequence;

[0011] Splitting the sample sequence from the beginning of the column to the end of the column according to a preset number of parts to obtain at least one sub-sample set;

[0012] Calculate the overall positive sample hit rate of the verification set, and calculate the positive sample hit rate of each sub-sample set;

[0013] Obtaining the positive sample hit rate of each level of the intermediate model, and determining at least one level to be evaluated according to the positive sample hit rate of each level;

[0014] Verifying the rationality of the positive sample hit rate of each layer to be evaluated according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set;

[0015] When the rationality of the positive sample hit rate of each layer to be evaluated passes the verification, the intermediate model is determined as the target model;

[0016] Obtaining a population to be stratified, and predicting the positive sample hit rate of the population to be stratified;

[0017] Obtaining the data to be processed corresponding to each person to be stratified in the population to be stratified, and inputting the data to be processed into the target model to obtain the predicted probability corresponding to each person to be stratified;

[0018] The persons to be stratified are stratified according to the positive sample hit rate of the persons to be stratified, the prediction probability corresponding to each person to be stratified and the positive sample hit rate of each layer to be evaluated to obtain a stratification result.

[0019] According to a preferred embodiment of the present invention, the calculating of the overall positive sample hit rate of the validation set and the calculating of the positive sample hit rate of each sub-sample set include:

[0020] Obtain the total number of samples in the validation set;

[0021] Obtain the number of positive samples in the validation set;

[0022] Calculate the quotient of the number of positive samples in the validation set and the total number of samples in the validation set, and use it as the overall positive sample hit rate;

[0023] Get the total number of samples in each subsample set;

[0024] Get the number of positive samples in each sub-sample set;

[0025] The quotient of the number of positive samples in each sub-sample set and the total number of samples in each sub-sample set is calculated and used as the positive sample hit rate of each sub-sample set.

[0026] According to a preferred embodiment of the present invention, determining at least one layer to be evaluated according to the positive sample hit rate of each layer includes:

[0027] Sort each level in descending order according to the positive sample hit rate of each level;

[0028] The layers except the last layer are determined as the at least one layer to be evaluated.

[0029] According to a preferred embodiment of the present invention, the checking the rationality of the positive sample hit rate of each to-be-evaluated layer according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set includes:

[0030] Determine the subsample set contained in each layer to be evaluated;

[0031] Calculate the sum of the positive sample hit rates of the subsample sets contained in each layer to be evaluated, and use it as the positive sample hit rate of each layer to be evaluated;

[0032] Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the overall positive sample hit rate to obtain the positive sample lift of each layer to be evaluated;

[0033] Obtain the maximum value of the positive sample hit rate of the subsample set contained in each layer to be evaluated, and use it as the first threshold of each layer to be evaluated;

[0034] Determine a database corresponding to the historical recruitment data of the object, and obtain from the database a minimum value of the positive sample hit rate corresponding to the data contained in each preset period within a preset time range;

[0035] Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the minimum value, and use it as the maximum positive sample hit rate of each layer to be evaluated;

[0036] When it is detected that the positive sample lift and the maximum positive sample hit rate of the first to-be-evaluated layer are both less than or equal to the first threshold of the first to-be-evaluated layer, it is determined that the rationality of the positive sample hit rate of the first to-be-evaluated layer has passed the verification; or

[0037] When it is detected that the positive sample lift and the maximum positive sample hit rate of the second layer to be evaluated are not both less than or equal to the first threshold of the second layer to be evaluated, it is determined that the rationality of the positive sample hit rate of the second layer to be evaluated has failed the verification.

[0038] According to a preferred embodiment of the present invention, the method further comprises:

[0039] When the rationality of the positive sample hit rate of each layer to be evaluated has not passed the verification, the rationality of each layer to be evaluated is traversed in order from high to low in terms of the positive sample hit rate of each layer to be evaluated;

[0040] When the rationality of the traversed layer to be evaluated fails to pass the verification, the positive sample hit rate of the traversed layer to be evaluated is continuously reduced according to a preset multiple until the positive sample hit rate of the traversed layer to be evaluated passes the verification;

[0041] Continue traversing from the traversed layer to be evaluated and adjust according to the preset multiple until the rationality of the positive sample hit rate of each layer to be evaluated is verified, and then stop traversing.

[0042] According to a preferred embodiment of the present invention, the predicting the positive sample hit rate of the to-be-stratified population includes:

[0043] Get the preset timing function;

[0044] Determine the positive sample hit rate corresponding to the data contained in each preset period within the preset time range and the time of the corresponding point as factors of the time series function, and construct a trend graph, wherein the trend graph is used to reflect the mapping relationship between time and the positive sample hit rate;

[0045] Obtain the current time corresponding to the population to be stratified;

[0046] Based on the position of the current time in the trend graph, the positive sample hit rate of the to-be-stratified population is determined.

[0047] According to a preferred embodiment of the present invention, the stratification process is performed on the persons to be stratified according to the positive sample hit rate of the persons to be stratified, the prediction probability corresponding to each person to be stratified and the positive sample hit rate of each layer to be evaluated, and the stratification results obtained include:

[0048] Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the positive sample hit rate of the population to be stratified, and obtain the positive sample lift of the population to be stratified in each layer to be evaluated;

[0049] Calculate the product of the overall positive sample hit rate and the positive sample lift of the to-be-stratified population in each to-be-evaluated layer to obtain the sum of the positive sample hit rates of the to-be-stratified population in each to-be-evaluated layer;

[0050] Determine the number of the subsample sets corresponding to each layer to be evaluated of the population to be stratified according to the sum of the positive sample hit rates of the population to be stratified in each layer to be evaluated;

[0051] Calculate the quotient of the number of the sub-sample sets corresponding to each layer to be evaluated of the population to be stratified and the preset number of copies to obtain the proportion of the number of people in each layer to be evaluated of the population to be stratified;

[0052] According to the order of the predicted probability corresponding to each person to be stratified from high to low, the persons to be stratified are sorted to obtain a target sequence;

[0053] The target sequence is split according to the proportion of the population to be stratified in each layer to be evaluated to obtain the stratification result.

[0054] In a second aspect, an embodiment of the present invention provides an object stratification device based on artificial intelligence, comprising:

[0055] An acquisition unit, used to acquire historical recruitment data of an object, construct a sample set using the historical recruitment data, and split the sample set to obtain a training set and a validation set;

[0056] A training unit, used to obtain a preset model with classification function, and train the preset model using the training set to obtain an intermediate model;

[0057] An input unit, used to input the validation set into the intermediate model to obtain the predicted probability of each sample in the validation set;

[0058] A sorting unit, used to sort each sample in the validation set in descending order of the predicted probability to obtain a sample sequence;

[0059] A splitting unit, used to split the sample sequence from the beginning of the column to the end of the column according to a preset number of parts to obtain at least one sub-sample set;

[0060] A calculation unit, used to calculate the overall positive sample hit rate of the verification set and the positive sample hit rate of each sub-sample set;

[0061] A determination unit, used to obtain the positive sample hit rate of each level of the intermediate model, and determine at least one layer to be evaluated according to the positive sample hit rate of each level;

[0062] A verification unit, used to verify the rationality of the positive sample hit rate of each layer to be evaluated according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set;

[0063] The determining unit is further configured to determine the intermediate model as the target model when the rationality of the positive sample hit rate of each layer to be evaluated passes the verification;

[0064] A prediction unit, used to obtain a population to be stratified and predict a positive sample hit rate of the population to be stratified;

[0065] The input unit is further used to obtain the to-be-processed data corresponding to each person to be stratified in the to-be-stratified population, and input the to-be-processed data into the target model to obtain the predicted probability corresponding to each person to be stratified;

[0066] The stratification unit is used to stratify the persons to be stratified according to the positive sample hit rate of the persons to be stratified, the prediction probability corresponding to each person to be stratified and the positive sample hit rate of each layer to be evaluated, so as to obtain a stratification result.

[0067] In a third aspect, an embodiment of the present invention further provides a computer device, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the computer program, the artificial intelligence-based object stratification method described in the first aspect is implemented.

[0068] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the processor executes the artificial intelligence-based object stratification method described in the first aspect above.

[0069] The embodiment of the present invention provides an object stratification method, device, equipment and medium based on artificial intelligence, which can obtain historical recruitment data of the object, and use the historical recruitment data to build a sample set, split the sample set to obtain a training set and a verification set, obtain a preset model with classification function, and use the training set to train the preset model to obtain an intermediate model, input the verification set into the intermediate model, obtain the predicted probability of each sample in the verification set, sort each sample in the verification set in order from high to low according to the predicted probability, obtain a sample sequence, split the sample sequence from the head of the column to the end of the column according to a preset number of copies, obtain at least one sub-sample set, calculate the overall positive sample hit rate of the verification set, and calculate the positive sample hit rate of each sub-sample set, obtain the positive sample hit rate of each level of the intermediate model, and determine at least one layer to be evaluated according to the positive sample hit rate of each level, so as to realize the division of the initial level, which is convenient for subsequent Evaluation and adjustment are carried out, and the rationality of the positive sample hit rate of each layer to be evaluated is verified according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set, and the rationality of the positive sample hit rate of each layer to be evaluated is verified in turn, so as to facilitate subsequent corrections to ensure the rationality of the trained model stratification. When the rationality of the positive sample hit rate of each layer to be evaluated passes the verification, the intermediate model is determined as the target model, the population to be stratified is obtained, and the positive sample hit rate of the population to be stratified is predicted, and the data to be processed corresponding to each person to be stratified in the population to be stratified is obtained, and the data to be processed is input into the target model to obtain the predicted probability corresponding to each person to be stratified, and the people to be stratified are stratified according to the positive sample hit rate of the population to be stratified, the predicted probability corresponding to each person to be stratified and the positive sample hit rate of each layer to be evaluated to obtain a stratified result, thereby realizing the hierarchical division of the population, which is convenient for subsequent targeted management and task allocation. BRIEF DESCRIPTION OF THE DRAWINGS

[0070] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying any creative work.

[0071] Figure 1 A schematic diagram of a process of an object stratification method based on artificial intelligence provided by an embodiment of the present invention;

[0072] Figure 2 A schematic block diagram of an object stratification device based on artificial intelligence provided by an embodiment of the present invention;

[0073] Figure 3 A schematic block diagram of a computer device provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0074] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0075] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprises" indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0076] It should also be understood that the terms used in this specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0077] It should be further understood that the term "and / or" used in the present description and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0078] See also Figure 1 , which is a flow chart of an artificial intelligence-based object stratification method provided in an embodiment of the present invention.

[0079] S10, obtaining historical recruitment data of the object, and using the historical recruitment data to construct a sample set, splitting the sample set to obtain a training set and a validation set.

[0080] In this embodiment, the historical recruitment data of the object may include, but is not limited to, a combination of one or more of the following data:

[0081] The subject's gender, age, length of service, historical performance, education level, length of employment, and the number of new employees at the preset time.

[0082] In this embodiment, the objects may include, but are not limited to: insurance agents, sales personnel, etc.

[0083] In this embodiment, the sample set may be split according to a preset ratio.

[0084] The preset ratio may be user-configured, such as 7 / 3, 6 / 4, etc. The present invention does not limit the value of the preset ratio.

[0085] S11, obtaining a preset model with classification function, and using the training set to train the preset model to obtain an intermediate model.

[0086] In this embodiment, the preset model may include, but is not limited to, a binary classification model, such as a LightGBM model.

[0087] In structured data scenarios, the LightGBM model has strong learning ability and high model accuracy.

[0088] In this embodiment, when the preset model is used to predict whether an agent can recruit new staff in the next three months, the historical recruitment data corresponding to each agent in the training set is determined as the input data of the preset model, whether each agent will recruit new staff in the next three months is determined as the training target, and the preset model is trained to obtain the intermediate model.

[0089] The output of the intermediate model is the predicted probability of whether to recruit new employees.

[0090] S12, inputting the validation set into the intermediate model to obtain the predicted probability of each sample in the validation set.

[0091] Continuing with the above example, the predicted probability is the probability that the agent corresponding to each sample in the validation set will be able to recruit new members in the next three months.

[0092] S13, sorting each sample in the verification set in descending order of the prediction probability to obtain a sample sequence.

[0093] It can be understood that the higher the probability, the greater the possibility of achieving recruitment. Each sample in the verification set is sorted in descending order according to the predicted probability, so that targeted assistance can be provided for subsequent personnel stratification.

[0094] S14, splitting the sample sequence from the beginning of the column to the end of the column according to a preset number of parts to obtain at least one sub-sample set.

[0095] In this embodiment, the preset number of copies can be configured according to the actual sample size.

[0096] For example, when the sample size is large enough, the preset number of copies may be 1,000 copies; when the sample size is not large enough, the preset number of copies may be 100 copies.

[0097] By further example, when the sample sequence contains 100,000 samples and the preset number of copies is 100, the sample sequence is split from the beginning of the column to the end of the column, and 100 sub-sample sets will be obtained, each sub-sample set contains 1,000 samples, and, in the order of splitting, the higher the predicted probability of each sample in the sub-sample set ranked first, and so on, the lower the predicted probability of each sample in the sub-sample set ranked last.

[0098] S15, calculating the overall positive sample hit rate of the verification set, and calculating the positive sample hit rate of each sub-sample set.

[0099] In at least one embodiment of the present invention, the calculating of the overall positive sample hit rate of the validation set and the calculating of the positive sample hit rate of each sub-sample set comprises:

[0100] Obtain the total number of samples in the validation set;

[0101] Obtain the number of positive samples in the validation set;

[0102] Calculate the quotient of the number of positive samples in the validation set and the total number of samples in the validation set, and use it as the overall positive sample hit rate;

[0103] Get the total number of samples in each subsample set;

[0104] Get the number of positive samples in each sub-sample set;

[0105] The quotient of the number of positive samples in each sub-sample set and the total number of samples in each sub-sample set is calculated and used as the positive sample hit rate of each sub-sample set.

[0106] In this embodiment, the proportion of positive samples in the set is determined as the positive sample hit rate.

[0107] For example, when predicting whether an agent can recruit new members in the next three months, the agent who can recruit new members in the next three months is the positive sample, and the proportion of positive samples in the overall samples is the positive sample hit rate.

[0108] S16, obtaining the positive sample hit rate of each level of the intermediate model, and determining at least one level to be evaluated according to the positive sample hit rate of each level.

[0109] In at least one embodiment of the present invention, determining at least one layer to be evaluated according to the positive sample hit rate of each layer includes:

[0110] Sort each level in descending order according to the positive sample hit rate of each level;

[0111] The layers except the last layer are determined as the at least one layer to be evaluated.

[0112] It can be understood that since the positive sample hit rate level set at each level in the intermediate model is multiplied by its sample proportion and then summed up to equal the overall positive sample hit rate, it is necessary to retain a level hit rate freedom.

[0113] For example, in the agent high-quality recruitment subject prediction scenario, the intermediate model is used to predict whether the agent in service is a high-quality recruitment subject (i.e., whether there will be recruitment in the next three months) at the beginning of each quarter. The intermediate model has three levels: A, B, and C. After initialization, the proportion of high-quality recruitment subjects at level A is P. A , the proportion of high-quality recruits at level B is P B , the C level is the level with reserved hit rate freedom.

[0114] For example, if the proportion of high-quality recruiting subjects in the current subject (overall positive sample hit rate) is P, then after initialization, P A =1.5P,P B =1.2P.

[0115] In this embodiment, the positive sample hit rate of each to-be-evaluated layer in the at least one to-be-evaluated layer may be evaluated and adjusted according to subsequent processing to determine the rationality of the positive sample hit rate of each level.

[0116] This embodiment stratifies samples based on the positive sample hit rate and predicted probability to weaken the impact of time differences on samples, thereby improving the stability of samples and ensuring the effect of subsequent model training.

[0117] S17, verifying the rationality of the positive sample hit rate of each to-be-evaluated layer according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set.

[0118] In at least one embodiment of the present invention, the checking the rationality of the positive sample hit rate of each to-be-evaluated layer according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set includes:

[0119] Determine the subsample set contained in each layer to be evaluated;

[0120] Calculate the sum of the positive sample hit rates of the subsample sets contained in each layer to be evaluated, and use it as the positive sample hit rate of each layer to be evaluated;

[0121] Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the overall positive sample hit rate to obtain the positive sample lift of each layer to be evaluated;

[0122] Obtain the maximum value of the positive sample hit rate of the subsample set contained in each layer to be evaluated, and use it as the first threshold of each layer to be evaluated;

[0123] Determine a database corresponding to the historical recruitment data of the object, and obtain from the database a minimum value of the positive sample hit rate corresponding to the data contained in each preset period within a preset time range;

[0124] Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the minimum value, and use it as the maximum positive sample hit rate of each layer to be evaluated;

[0125] When it is detected that the positive sample lift and the maximum positive sample hit rate of the first to-be-evaluated layer are both less than or equal to the first threshold of the first to-be-evaluated layer, it is determined that the rationality of the positive sample hit rate of the first to-be-evaluated layer has passed the verification; or

[0126] When it is detected that the positive sample lift and the maximum positive sample hit rate of the second layer to be evaluated are not both less than or equal to the first threshold of the second layer to be evaluated, it is determined that the rationality of the positive sample hit rate of the second layer to be evaluated has failed the verification.

[0127] In this embodiment, the subsample set included in each layer to be evaluated can be determined according to the actual stratification method of the intermediate model. For example, A =1.5P,P B =1.2P ratio to determine the subsample set contained in each layer to be evaluated.

[0128] In this embodiment, the preset time range can be custom configured, such as two years.

[0129] In this embodiment, the preset period may also be custom configured, such as every month.

[0130] For example, you can obtain the minimum monthly positive sample hit rate for the past two years.

[0131] Continuing with the above example, evaluate P A With P B The rationality of the numerical value is as follows:

[0132] (1) Positive sample improvement at level A It cannot be greater than the maximum level that the minimum granularity of the population can reach, that is, α A ≤α1;

[0133] (2) The positive sample lift α of level A A The maximum possible value α A-max When the hit rate of the positive samples of the whole population fluctuates, condition (1) can still be met. For example, the fluctuation range of the hit rate of the positive samples of the whole population is [P min ,P max ], where P min is the minimum value of the monthly positive sample hit rate in the past two years, P max The maximum value of the monthly positive sample hit rate in the past two years must meet the following requirements:

[0134] (3) Positive sample improvement at level B Since the levels are divided in descending order of positive sample hit rate, if i is a positive integer, that is, level A consists of the first n sub-sample sets, then α must be satisfied B ≤α n+1 At the same time, if the fluctuation range of the positive sample hit rate of the whole population is considered, then α B The maximum possible value

[0135] Through the above implementation, the rationality of the positive sample hit rate of each layer to be evaluated is verified in turn, which is convenient for subsequent correction to ensure the rationality of the training model stratification.

[0136] S18, when the rationality of the positive sample hit rate of each layer to be evaluated passes the verification, the intermediate model is determined as the target model.

[0137] In this embodiment, when the positive sample lift degree of each layer to be evaluated and the corresponding maximum positive sample hit rate are less than or equal to the corresponding first threshold, it is determined that the rationality of the positive sample hit rate of each layer to be evaluated has passed the verification, and the intermediate model is determined as the target model.

[0138] When the positive sample lift of each to-be-evaluated layer and the corresponding maximum positive sample hit rate are not both less than or equal to the corresponding first threshold, it is determined that the rationality of the positive sample hit rate of each to-be-evaluated layer has not passed the verification.

[0139] In at least one embodiment of the present invention, the method further comprises:

[0140] When the rationality of the positive sample hit rate of each layer to be evaluated has not passed the verification, the rationality of each layer to be evaluated is traversed in order from high to low in terms of the positive sample hit rate of each layer to be evaluated;

[0141] When the rationality of the traversed layer to be evaluated fails to pass the verification, the positive sample hit rate of the traversed layer to be evaluated is continuously reduced according to a preset multiple until the positive sample hit rate of the traversed layer to be evaluated passes the verification;

[0142] Continue traversing from the traversed layer to be evaluated and adjust according to the preset multiple until the rationality of the positive sample hit rate of each layer to be evaluated is verified, and then stop traversing.

[0143] The preset multiple can be 0.9 times

[0144] Continuing with the above example, when it is determined that the rationality of the positive sample hit rate of each layer to be evaluated has not passed the verification, first determine whether the positive sample hit rate of layer A is appropriate. If not, adjust layer A and multiply the positive sample hit rate of layer A by 0.9; then determine whether the positive sample hit rate of layer A is appropriate, until the positive sample hit rate of layer A is appropriate; then further determine whether the positive sample hit rate of layer B is appropriate. If not, adjust layer B and multiply the positive sample hit rate of layer B by 0.9, until the positive sample hit rate of layer B is appropriate.

[0145] Through the above implementation, when it is verified that the hierarchical configuration of the intermediate model is unreasonable, automatic adjustment can be performed to ensure the accuracy of model prediction.

[0146] S19, obtaining a population to be stratified, and predicting the positive sample hit rate of the population to be stratified.

[0147] In this embodiment, the population to be stratified can be determined by relevant staff, such as project leaders, assessors, and the like.

[0148] Specifically, the predicting the positive sample hit rate of the to-be-stratified population includes:

[0149] Get the preset timing function;

[0150] Determine the positive sample hit rate corresponding to the data contained in each preset period within the preset time range and the time of the corresponding point as factors of the time series function, and construct a trend graph, wherein the trend graph is used to reflect the mapping relationship between time and the positive sample hit rate;

[0151] Obtain the current time corresponding to the population to be stratified;

[0152] Based on the position of the current time in the trend graph, the positive sample hit rate of the to-be-stratified population is determined.

[0153] The time series function may also be a time series model, such as a prophet model.

[0154] The trend graph is obtained by learning the regularity of the time series function. The trend graph can reflect the relationship between time and the positive sample hit rate, so as to predict the positive sample hit rate of the crowd.

[0155] For example, in the recruitment prediction scenario, we first need to determine the recruitment of the employed population at the beginning of each month in the three months after the statistical month, so as to calculate the monthly proportion of the main population of high-quality recruitment in the whole system, that is, the proportion of positive samples, and then use the time series model prophet to predict the recruitment of the current employed population in the next three months, which is the positive sample hit rate P′.

[0156] Through the above implementation, the positive sample hit rate of the population to be stratified can be determined based on time series prediction to obtain the overall positive sample level of the population to be stratified.

[0157] S20, obtaining the data to be processed corresponding to each person to be stratified in the population to be stratified, and inputting the data to be processed into the target model to obtain the predicted probability corresponding to each person to be stratified.

[0158] In the above implementation, the predicted probability of each person to be stratified is predicted by the target model finally obtained, so as to be used for subsequent further data analysis.

[0159] S21, stratifying the persons to be stratified according to the positive sample hit rate of the persons to be stratified, the prediction probability corresponding to each person to be stratified and the positive sample hit rate of each layer to be evaluated to obtain a stratification result.

[0160] Specifically, the stratification process is performed on the persons to be stratified according to the positive sample hit rate of the population to be stratified, the prediction probability corresponding to each person to be stratified, and the positive sample hit rate of each layer to be evaluated, and the stratification results obtained include:

[0161] Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the positive sample hit rate of the population to be stratified, and obtain the positive sample lift of the population to be stratified in each layer to be evaluated;

[0162] Calculate the product of the overall positive sample hit rate and the positive sample lift of the to-be-stratified population in each to-be-evaluated layer to obtain the sum of the positive sample hit rates of the to-be-stratified population in each to-be-evaluated layer;

[0163] Determine the number of the subsample sets corresponding to each layer to be evaluated of the population to be stratified according to the sum of the positive sample hit rates of the population to be stratified in each layer to be evaluated;

[0164] Calculate the quotient of the number of the sub-sample sets corresponding to each layer to be evaluated of the population to be stratified and the preset number of copies to obtain the proportion of the number of people in each layer to be evaluated of the population to be stratified;

[0165] According to the order of the predicted probability corresponding to each person to be stratified from high to low, the persons to be stratified are sorted to obtain a target sequence;

[0166] The target sequence is split according to the proportion of the population to be stratified in each layer to be evaluated to obtain the stratification result.

[0167] Continuing with the above example, when the positive sample hit rate of the to-be-stratified population is P′, and the value of the preset number is 1000:

[0168] 1. Calculate the positive sample lift of the population to be stratified in each layer to be evaluated:

[0169] 2. Determine the proportion of people in Tier A So that n satisfies

[0170] 3. Determine the proportion of people in Tier B So that m satisfies

[0171] 4. Based on the trained target model, predict the probability of each person to be stratified being a high-quality agent in the next three months, and sort them from high to low according to the probability;

[0172] 5. Divide the samples before RA into level A, remove the samples of level A from the samples of before RA+RB, and get level B.

[0173] Through the above implementation, it is possible to achieve hierarchical division of the population, which is convenient for subsequent targeted management and task allocation.

[0174] This embodiment predicts the positive sample ratio of future samples and implements a threshold segmentation scheme for the population based on the concept of positive sample hit rate improvement, thereby improving the stability of the model output.

[0175] In this embodiment, after obtaining the stratification results, targeted task allocation and management can be performed on the people at each level, so that the people at each level can better play their advantages.

[0176] It should be noted that in order to further ensure the security of the data and prevent the data from being maliciously tampered with, the stratification results can be stored on the blockchain node.

[0177] It can be seen from the above technical scheme that the present invention obtains the historical recruitment data of the object, and uses the historical recruitment data to build a sample set, splits the sample set to obtain a training set and a verification set, obtains a preset model with classification function, and uses the training set to train the preset model to obtain an intermediate model, inputs the verification set into the intermediate model, obtains the predicted probability of each sample in the verification set, sorts each sample in the verification set in order from high to low according to the predicted probability, obtains a sample sequence, splits the sample sequence from the head of the column to the end of the column according to a preset number of copies, obtains at least one sub-sample set, calculates the overall positive sample hit rate of the verification set, and calculates the positive sample hit rate of each sub-sample set, obtains the positive sample hit rate of each level of the intermediate model, and determines at least one layer to be evaluated according to the positive sample hit rate of each level, realizes the initial level division, and is convenient for subsequent evaluation and adjustment, according to the The rationality of the positive sample hit rate of each to-be-evaluated layer is verified by using the overall positive sample hit rate and the positive sample hit rate of each sub-sample set, and the rationality of the positive sample hit rate of each to-be-evaluated layer is verified in turn to facilitate subsequent corrections to ensure the rationality of the trained model stratification. When the rationality of the positive sample hit rate of each to-be-evaluated layer passes the verification, the intermediate model is determined as the target model, the to-be-stratified population is obtained, and the positive sample hit rate of the to-be-stratified population is predicted. The processed data corresponding to each to-be-stratified person in the to-be-stratified population is obtained, and the processed data is input into the target model to obtain the predicted probability corresponding to each to-be-stratified person. The to-be-stratified persons are stratified according to the positive sample hit rate of the to-be-stratified population, the predicted probability corresponding to each to-be-stratified person and the positive sample hit rate of each to-be-stratified layer to obtain a stratified result, thereby realizing the hierarchical division of the population and facilitating subsequent targeted management and task allocation.

[0178] The embodiment of the present invention further provides an artificial intelligence-based object stratification device, which is used to execute any embodiment of the aforementioned artificial intelligence-based object stratification method. Figure 2 , Figure 2 It is a schematic block diagram of an artificial intelligence-based object stratification device provided in an embodiment of the present invention.

[0179] like Figure 2 As shown, the artificial intelligence-based object stratification device 100 includes: an acquisition unit 101, a training unit 102, an input unit 103, a sorting unit 104, a splitting unit 105, a calculation unit 106, a determination unit 107, a verification unit 108, a prediction unit 109, and a stratification unit 110.

[0180] The acquisition unit 101 acquires historical recruitment data of the object, and uses the historical recruitment data to construct a sample set, and splits the sample set to obtain a training set and a validation set.

[0181] In this embodiment, the historical recruitment data of the object may include, but is not limited to, a combination of one or more of the following data:

[0182] The subject's gender, age, length of service, historical performance, education level, length of employment, and the number of new employees at the preset time.

[0183] In this embodiment, the objects may include, but are not limited to: insurance agents, sales personnel, etc.

[0184] In this embodiment, the sample set may be split according to a preset ratio.

[0185] The preset ratio may be user-configured, such as 7 / 3, 6 / 4, etc. The present invention does not limit the value of the preset ratio.

[0186] The training unit 102 obtains a preset model with a classification function, and uses the training set to train the preset model to obtain an intermediate model.

[0187] In this embodiment, the preset model may include, but is not limited to, a binary classification model, such as a LightGBM model.

[0188] In structured data scenarios, the LightGBM model has strong learning ability and high model accuracy.

[0189] In this embodiment, when the preset model is used to predict whether an agent can recruit new staff in the next three months, the historical recruitment data corresponding to each agent in the training set is determined as the input data of the preset model, whether each agent will recruit new staff in the next three months is determined as the training target, and the preset model is trained to obtain the intermediate model.

[0190] The output of the intermediate model is the predicted probability of whether to recruit new employees.

[0191] The input unit 103 inputs the validation set into the intermediate model to obtain the predicted probability of each sample in the validation set.

[0192] Continuing with the above example, the predicted probability is the probability that the agent corresponding to each sample in the validation set will be able to recruit new members in the next three months.

[0193] The sorting unit 104 sorts each sample in the validation set in descending order of the prediction probability to obtain a sample sequence.

[0194] It can be understood that the higher the probability, the greater the possibility of achieving recruitment. Each sample in the verification set is sorted in descending order according to the predicted probability, so that targeted assistance can be provided for subsequent personnel stratification.

[0195] The splitting unit 105 splits the sample sequence from the beginning of the column to the end of the column according to a preset number of parts to obtain at least one sub-sample set.

[0196] In this embodiment, the preset number of copies can be configured according to the actual sample size.

[0197] For example, when the sample size is large enough, the preset number of copies may be 1,000 copies; when the sample size is not large enough, the preset number of copies may be 100 copies.

[0198] By further example, when the sample sequence contains 100,000 samples and the preset number of copies is 100, the sample sequence is split from the beginning of the column to the end of the column, and 100 sub-sample sets will be obtained, each sub-sample set contains 1,000 samples, and, in the order of splitting, the higher the predicted probability of each sample in the sub-sample set ranked first, and so on, the lower the predicted probability of each sample in the sub-sample set ranked last.

[0199] The calculation unit 106 calculates the overall positive sample hit rate of the verification set and calculates the positive sample hit rate of each sub-sample set.

[0200] In at least one embodiment of the present invention, the calculation unit 106 calculates the overall positive sample hit rate of the validation set and calculates the positive sample hit rate of each sub-sample set, including:

[0201] Obtain the total number of samples in the validation set;

[0202] Obtain the number of positive samples in the validation set;

[0203] Calculate the quotient of the number of positive samples in the validation set and the total number of samples in the validation set, and use it as the overall positive sample hit rate;

[0204] Get the total number of samples in each subsample set;

[0205] Get the number of positive samples in each sub-sample set;

[0206] The quotient of the number of positive samples in each sub-sample set and the total number of samples in each sub-sample set is calculated and used as the positive sample hit rate of each sub-sample set.

[0207] In this embodiment, the proportion of positive samples in the set is determined as the positive sample hit rate.

[0208] For example, when predicting whether an agent can recruit new members in the next three months, the agent who can recruit new members in the next three months is the positive sample, and the proportion of positive samples in the overall samples is the positive sample hit rate.

[0209] The determination unit 107 obtains the positive sample hit rate of each level of the intermediate model, and determines at least one layer to be evaluated according to the positive sample hit rate of each level.

[0210] In at least one embodiment of the present invention, the determining unit 107 determines at least one layer to be evaluated according to the positive sample hit rate of each layer, including:

[0211] Sort each level in descending order according to the positive sample hit rate of each level;

[0212] The layers except the last layer are determined as the at least one layer to be evaluated.

[0213] It can be understood that since the positive sample hit rate level set at each level in the intermediate model is multiplied by its sample proportion and then summed up to equal the overall positive sample hit rate, it is necessary to retain a level hit rate freedom.

[0214] For example, in the agent high-quality recruitment subject prediction scenario, the intermediate model is used to predict whether the agent in service is a high-quality recruitment subject (i.e., whether there will be recruitment in the next three months) at the beginning of each quarter. The intermediate model has three levels: A, B, and C. After initialization, the proportion of high-quality recruitment subjects at level A is P. A , the proportion of high-quality recruits at level B is P B , the C level is the level with reserved hit rate freedom.

[0215] For example, if the proportion of high-quality recruiting subjects in the current subject (overall positive sample hit rate) is P, then after initialization, P A =1.5P,P B =1.2P.

[0216] In this embodiment, the positive sample hit rate of each to-be-evaluated layer in the at least one to-be-evaluated layer may be evaluated and adjusted according to subsequent processing to determine the rationality of the positive sample hit rate of each level.

[0217] This embodiment stratifies samples based on the positive sample hit rate and predicted probability to weaken the impact of time differences on samples, thereby improving the stability of samples and ensuring the effect of subsequent model training.

[0218] The verification unit 108 verifies the rationality of the positive sample hit rate of each to-be-evaluated layer according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set.

[0219] In at least one embodiment of the present invention, the verification unit 108 verifies the rationality of the positive sample hit rate of each to-be-evaluated layer according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set, including:

[0220] Determine the subsample set contained in each layer to be evaluated;

[0221] Calculate the sum of the positive sample hit rates of the subsample sets contained in each layer to be evaluated, and use it as the positive sample hit rate of each layer to be evaluated;

[0222] Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the overall positive sample hit rate to obtain the positive sample lift of each layer to be evaluated;

[0223] Obtain the maximum value of the positive sample hit rate of the subsample set contained in each layer to be evaluated, and use it as the first threshold of each layer to be evaluated;

[0224] Determine a database corresponding to the historical recruitment data of the object, and obtain from the database a minimum value of the positive sample hit rate corresponding to the data contained in each preset period within a preset time range;

[0225] Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the minimum value, and use it as the maximum positive sample hit rate of each layer to be evaluated;

[0226] When it is detected that the positive sample lift and the maximum positive sample hit rate of the first to-be-evaluated layer are both less than or equal to the first threshold of the first to-be-evaluated layer, it is determined that the rationality of the positive sample hit rate of the first to-be-evaluated layer has passed the verification; or

[0227] When it is detected that the positive sample lift and the maximum positive sample hit rate of the second layer to be evaluated are not both less than or equal to the first threshold of the second layer to be evaluated, it is determined that the rationality of the positive sample hit rate of the second layer to be evaluated has failed the verification.

[0228] In this embodiment, the subsample set included in each layer to be evaluated can be determined according to the actual stratification method of the intermediate model. For example, A =1.5P,P B =1.2P ratio to determine the subsample set contained in each layer to be evaluated.

[0229] In this embodiment, the preset time range can be custom configured, such as two years.

[0230] In this embodiment, the preset period may also be custom configured, such as every month.

[0231] For example, you can obtain the minimum monthly positive sample hit rate for the past two years.

[0232] Continuing with the above example, evaluate P A With P B The rationality of the numerical value is as follows:

[0233] (4) Positive sample improvement at level A It cannot be greater than the maximum level that the minimum granularity of the population can reach, that is, α A ≤α1;

[0234] (5) Positive sample improvement α at level A A The maximum possible value α A-max When the hit rate of the positive samples of the whole population fluctuates, condition (1) can still be met. For example, the fluctuation range of the hit rate of the positive samples of the whole population is [P min ,P max ], where P min is the minimum value of the monthly positive sample hit rate in the past two years, P max The maximum value of the monthly positive sample hit rate in the past two years must meet the following requirements:

[0235] (6) Positive sample improvement at level B Since the levels are divided in descending order of positive sample hit rate, if i is a positive integer, that is, level A consists of the first n sub-sample sets, then α must be satisfied B ≤α n+1 At the same time, if the fluctuation range of the positive sample hit rate of the whole population is considered, then α B The maximum possible value

[0236] Through the above implementation, the rationality of the positive sample hit rate of each layer to be evaluated is verified in turn, which is convenient for subsequent correction to ensure the rationality of the training model stratification.

[0237] When the rationality of the positive sample hit rate of each layer to be evaluated passes the verification, the determining unit 107 determines the intermediate model as the target model.

[0238] In this embodiment, when the positive sample lift degree of each layer to be evaluated and the corresponding maximum positive sample hit rate are less than or equal to the corresponding first threshold, it is determined that the rationality of the positive sample hit rate of each layer to be evaluated has passed the verification, and the intermediate model is determined as the target model.

[0239] When the positive sample lift of each to-be-evaluated layer and the corresponding maximum positive sample hit rate are not both less than or equal to the corresponding first threshold, it is determined that the rationality of the positive sample hit rate of each to-be-evaluated layer has not passed the verification.

[0240] In at least one embodiment of the present invention, when the rationality of the positive sample hit rate of each to-be-evaluated layer has not passed the verification, the rationality of each to-be-evaluated layer is traversed in order from high to low in terms of the positive sample hit rate of each to-be-evaluated layer;

[0241] When the rationality of the traversed layer to be evaluated fails to pass the verification, the positive sample hit rate of the traversed layer to be evaluated is continuously reduced according to a preset multiple until the positive sample hit rate of the traversed layer to be evaluated passes the verification;

[0242] Continue traversing from the traversed layer to be evaluated and adjust according to the preset multiple until the rationality of the positive sample hit rate of each layer to be evaluated is verified, and then stop traversing.

[0243] The preset multiple can be 0.9 times

[0244] Continuing with the above example, when it is determined that the rationality of the positive sample hit rate of each layer to be evaluated has not passed the verification, first determine whether the positive sample hit rate of layer A is appropriate. If not, adjust layer A and multiply the positive sample hit rate of layer A by 0.9; then determine whether the positive sample hit rate of layer A is appropriate, until the positive sample hit rate of layer A is appropriate; then further determine whether the positive sample hit rate of layer B is appropriate. If not, adjust layer B and multiply the positive sample hit rate of layer B by 0.9, until the positive sample hit rate of layer B is appropriate.

[0245] Through the above implementation, when it is verified that the hierarchical configuration of the intermediate model is unreasonable, automatic adjustment can be performed to ensure the accuracy of model prediction.

[0246] The prediction unit 109 obtains the population to be stratified and predicts the positive sample hit rate of the population to be stratified.

[0247] In this embodiment, the population to be stratified can be determined by relevant staff, such as project leaders, assessors, and the like.

[0248] Specifically, the prediction unit 109 predicts the positive sample hit rate of the to-be-stratified population including:

[0249] Get the preset timing function;

[0250] Determine the positive sample hit rate corresponding to the data contained in each preset period within the preset time range and the time of the corresponding point as factors of the time series function, and construct a trend graph, wherein the trend graph is used to reflect the mapping relationship between time and the positive sample hit rate;

[0251] Obtain the current time corresponding to the population to be stratified;

[0252] Based on the position of the current time in the trend graph, the positive sample hit rate of the to-be-stratified population is determined.

[0253] The time series function may also be a time series model, such as a prophet model.

[0254] The trend graph is obtained by learning the regularity of the time series function. The trend graph can reflect the relationship between time and the positive sample hit rate, so as to predict the positive sample hit rate of the crowd.

[0255] For example, in the recruitment prediction scenario, we first need to determine the recruitment of the employed population at the beginning of each month in the three months after the statistical month, so as to calculate the monthly proportion of the main population of high-quality recruitment in the whole system, that is, the proportion of positive samples, and then use the time series model prophet to predict the recruitment of the current employed population in the next three months, which is the positive sample hit rate P′.

[0256] Through the above implementation, the positive sample hit rate of the population to be stratified can be determined based on time series prediction to obtain the overall positive sample level of the population to be stratified.

[0257] The input unit 103 obtains the data to be processed corresponding to each person to be stratified in the population to be stratified, and inputs the data to be processed into the target model to obtain the predicted probability corresponding to each person to be stratified.

[0258] In the above implementation, the predicted probability of each person to be stratified is predicted by the target model finally obtained, so as to be used for subsequent further data analysis.

[0259] The stratification unit 110 stratifies the persons to be stratified according to the positive sample hit rate of the persons to be stratified, the prediction probability corresponding to each person to be stratified and the positive sample hit rate of each layer to be evaluated to obtain a stratification result.

[0260] Specifically, the stratification unit 110 stratifies the persons to be stratified according to the positive sample hit rate of the population to be stratified, the prediction probability corresponding to each person to be stratified, and the positive sample hit rate of each layer to be evaluated, and the stratification results include:

[0261] Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the positive sample hit rate of the population to be stratified, and obtain the positive sample lift of the population to be stratified in each layer to be evaluated;

[0262] Calculate the product of the overall positive sample hit rate and the positive sample lift of the to-be-stratified population in each to-be-evaluated layer to obtain the sum of the positive sample hit rates of the to-be-stratified population in each to-be-evaluated layer;

[0263] Determine the number of the subsample sets corresponding to each layer to be evaluated of the population to be stratified according to the sum of the positive sample hit rates of the population to be stratified in each layer to be evaluated;

[0264] Calculate the quotient of the number of the sub-sample sets corresponding to each layer to be evaluated of the population to be stratified and the preset number of copies to obtain the proportion of the number of people in each layer to be evaluated of the population to be stratified;

[0265] According to the order of the predicted probability corresponding to each person to be stratified from high to low, the persons to be stratified are sorted to obtain a target sequence;

[0266] The target sequence is split according to the proportion of the population to be stratified in each layer to be evaluated to obtain the stratification result.

[0267] Continuing with the above example, when the positive sample hit rate of the to-be-stratified population is P′, and the value of the preset number is 1000:

[0268] 1. Calculate the positive sample lift of the population to be stratified in each layer to be evaluated:

[0269] 2. Determine the proportion of people in Tier A So that n satisfies

[0270] 3. Determine the proportion of people in Tier B So that m satisfies

[0271] 4. Based on the trained target model, predict the probability of each person to be stratified being a high-quality agent in the next three months, and sort them from high to low according to the probability;

[0272] 5. Divide the samples before RA into level A, remove the samples of level A from the samples of before RA+RB, and get level B.

[0273] Through the above implementation, it is possible to achieve hierarchical division of the population, which is convenient for subsequent targeted management and task allocation.

[0274] This embodiment predicts the positive sample ratio of future samples and implements a threshold segmentation scheme for the population based on the concept of positive sample hit rate improvement, thereby improving the stability of the model output.

[0275] In this embodiment, after obtaining the stratification results, targeted task allocation and management can be performed on the people at each level, so that the people at each level can better play their advantages.

[0276] It should be noted that in order to further ensure the security of the data and prevent the data from being maliciously tampered with, the stratification results can be stored on the blockchain node.

[0277] It can be seen from the above technical scheme that the present invention obtains the historical recruitment data of the object, and uses the historical recruitment data to build a sample set, splits the sample set to obtain a training set and a verification set, obtains a preset model with classification function, and uses the training set to train the preset model to obtain an intermediate model, inputs the verification set into the intermediate model, obtains the predicted probability of each sample in the verification set, sorts each sample in the verification set in order from high to low according to the predicted probability, obtains a sample sequence, splits the sample sequence from the head of the column to the end of the column according to a preset number of copies, obtains at least one sub-sample set, calculates the overall positive sample hit rate of the verification set, and calculates the positive sample hit rate of each sub-sample set, obtains the positive sample hit rate of each level of the intermediate model, and determines at least one layer to be evaluated according to the positive sample hit rate of each level, realizes the initial level division, and is convenient for subsequent evaluation and adjustment, according to the The rationality of the positive sample hit rate of each to-be-evaluated layer is verified by using the overall positive sample hit rate and the positive sample hit rate of each sub-sample set, and the rationality of the positive sample hit rate of each to-be-evaluated layer is verified in turn to facilitate subsequent corrections to ensure the rationality of the trained model stratification. When the rationality of the positive sample hit rate of each to-be-evaluated layer passes the verification, the intermediate model is determined as the target model, the to-be-stratified population is obtained, and the positive sample hit rate of the to-be-stratified population is predicted. The processed data corresponding to each to-be-stratified person in the to-be-stratified population is obtained, and the processed data is input into the target model to obtain the predicted probability corresponding to each to-be-stratified person. The to-be-stratified persons are stratified according to the positive sample hit rate of the to-be-stratified population, the predicted probability corresponding to each to-be-stratified person and the positive sample hit rate of each to-be-stratified layer to obtain a stratified result, thereby realizing the hierarchical division of the population and facilitating subsequent targeted management and task allocation.

[0278] The above-mentioned object stratification device based on artificial intelligence can be implemented in the form of a computer program, which can be used in Figure 3 Runs on the computer device shown.

[0279] See also Figure 3, Figure 3 5 is a schematic block diagram of a computer device provided in an embodiment of the present invention. The computer device 500 is a server, which can be an independent server or a server cluster composed of multiple servers. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0280] Among them, Artificial Intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0281] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0282] See also Figure 3 The computer device 500 includes a processor 502 , a memory and a network interface 505 connected via a system bus 501 , wherein the memory may include a storage medium 503 and an internal memory 504 .

[0283] The storage medium 503 may store an operating system 5031 and a computer program 5032. When the computer program 5032 is executed, the processor 502 may execute an object layering method based on artificial intelligence.

[0284] The processor 502 is used to provide computing and control capabilities to support the operation of the entire computer device 500 .

[0285] The internal memory 504 provides an environment for the operation of the computer program 5032 in the storage medium 503. When the computer program 5032 is executed by the processor 502, the processor 502 can execute an object stratification method based on artificial intelligence.

[0286] The network interface 505 is used for network communication, such as providing data information transmission, etc. Those skilled in the art will understand that Figure 3The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the computer device 500 to which the solution of the present invention is applied. The specific computer device 500 may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0287] The processor 502 is used to run a computer program 5032 stored in the memory to implement the artificial intelligence-based object stratification method disclosed in the embodiment of the present invention.

[0288] Those skilled in the art will understand that Figure 3 The embodiments of the computer device shown in the figure do not constitute a limitation on the specific composition of the computer device. In other embodiments, the computer device may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently. For example, in some embodiments, the computer device may only include a memory and a processor. In such embodiments, the structure and function of the memory and the processor are the same as those of the embodiment of the present invention. Figure 3 The embodiments shown are consistent and will not be described again here.

[0289] It should be understood that in the embodiment of the present invention, the processor 502 may be a central processing unit (CPU), and the processor 502 may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0290] In another embodiment of the present invention, a computer-readable storage medium is provided. The computer-readable storage medium may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores a computer program, wherein when the computer program is executed by a processor, the object stratification method based on artificial intelligence disclosed in the embodiment of the present invention is implemented.

[0291] The blockchain referred to in this invention is a new application mode of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanism, encryption algorithm, etc. Blockchain is essentially a decentralized database, a string of data blocks generated by cryptographic methods. Each data block contains a batch of network transaction information, which is used to verify the validity of its information (anti-counterfeiting) and generate the next block. Blockchain can include the underlying blockchain platform, platform product service layer, and application service layer.

[0292] It should be noted that the data involved in this case were all obtained legally.

[0293] Those skilled in the art can clearly understand that, for the convenience and simplicity of description, the specific working process of the above-described equipment, devices and units can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here. Those of ordinary skill in the art can appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to the function. Whether these functions are executed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0294] In the several embodiments provided by the present invention, it should be understood that the disclosed equipment, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. Units with the same function may also be combined into one unit. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. In addition, the mutual coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection through some interfaces, devices or units, or may be an electrical, mechanical or other form of connection.

[0295] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiments of the present invention.

[0296] The present invention can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronic devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present invention can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present invention can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0297] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0298] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a storage medium. Based on this understanding, the technical solution of the present invention is essentially or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), disk or optical disk and other media that can store program codes.

[0299] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed by the present invention, and these modifications or replacements should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention shall be based on the protection scope of the claims.

Claims

1. An object stratification method based on artificial intelligence, characterized in that: include: Obtain historical recruitment data of the object, and use the historical recruitment data to construct a sample set, split the sample set to obtain a training set and a validation set; Acquire a preset model with classification function, and train the preset model using the training set to obtain an intermediate model; Inputting the validation set into the intermediate model to obtain the predicted probability of each sample in the validation set; Sort each sample in the validation set in descending order of the predicted probability to obtain a sample sequence; Splitting the sample sequence from the beginning of the column to the end of the column according to a preset number of parts to obtain at least one sub-sample set; Calculate the overall positive sample hit rate of the verification set, and calculate the positive sample hit rate of each sub-sample set; Obtaining the positive sample hit rate of each level of the intermediate model, and determining at least one level to be evaluated according to the positive sample hit rate of each level; Verifying the rationality of the positive sample hit rate of each layer to be evaluated according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set; When the rationality of the positive sample hit rate of each layer to be evaluated passes the verification, the intermediate model is determined as the target model; Obtaining a population to be stratified, and predicting the positive sample hit rate of the population to be stratified; Obtaining the data to be processed corresponding to each person to be stratified in the population to be stratified, and inputting the data to be processed into the target model to obtain the predicted probability corresponding to each person to be stratified; According to the positive sample hit rate of the population to be stratified, the prediction probability corresponding to each person to be stratified and the positive sample hit rate of each layer to be evaluated, the persons to be stratified are stratified to obtain stratification results; The checking of the rationality of the positive sample hit rate of each to-be-evaluated layer according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set includes: Determine the subsample set contained in each layer to be evaluated; Calculate the sum of the positive sample hit rates of the subsample sets contained in each layer to be evaluated, and use it as the positive sample hit rate of each layer to be evaluated; Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the overall positive sample hit rate to obtain the positive sample lift of each layer to be evaluated; Obtain the maximum value of the positive sample hit rate of the subsample set contained in each layer to be evaluated, and use it as the first threshold of each layer to be evaluated; Determine a database corresponding to the historical recruitment data of the object, and obtain from the database a minimum value of the positive sample hit rate corresponding to the data contained in each preset period within a preset time range; Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the minimum value, and use it as the maximum positive sample hit rate of each layer to be evaluated; When it is detected that the positive sample lift and the maximum positive sample hit rate of the first to-be-evaluated layer are both less than or equal to the first threshold of the first to-be-evaluated layer, it is determined that the rationality of the positive sample hit rate of the first to-be-evaluated layer has passed the verification; or When it is detected that the positive sample lift and the maximum positive sample hit rate of the second layer to be evaluated are not both less than or equal to the first threshold of the second layer to be evaluated, it is determined that the rationality of the positive sample hit rate of the second layer to be evaluated has failed the verification.

2. The object stratification method based on artificial intelligence according to claim 1, characterized in that: The calculating of the overall positive sample hit rate of the validation set and the calculating of the positive sample hit rate of each sub-sample set includes: Obtain the total number of samples in the validation set; Obtain the number of positive samples in the validation set; Calculate the quotient of the number of positive samples in the validation set and the total number of samples in the validation set, and use it as the overall positive sample hit rate; Get the total number of samples in each subsample set; Get the number of positive samples in each sub-sample set; The quotient of the number of positive samples in each sub-sample set and the total number of samples in each sub-sample set is calculated and used as the positive sample hit rate of each sub-sample set.

3. The object stratification method based on artificial intelligence according to claim 1, characterized in that: Determining at least one layer to be evaluated according to the positive sample hit rate of each layer includes: Sort each level in descending order according to the positive sample hit rate of each level; The layers except the last layer are determined as the at least one layer to be evaluated.

4. The object stratification method based on artificial intelligence according to claim 1, characterized in that: The method further comprises: When the rationality of the positive sample hit rate of each layer to be evaluated has not passed the verification, the rationality of each layer to be evaluated is traversed in order from high to low in terms of the positive sample hit rate of each layer to be evaluated; When the rationality of the traversed layer to be evaluated fails to pass the verification, the positive sample hit rate of the traversed layer to be evaluated is continuously reduced according to a preset multiple until the positive sample hit rate of the traversed layer to be evaluated passes the verification; Continue traversing from the traversed layer to be evaluated and adjust according to the preset multiple until the rationality of the positive sample hit rate of each layer to be evaluated is verified, and then stop traversing.

5. The object stratification method based on artificial intelligence according to claim 1, characterized in that: The prediction of the positive sample hit rate of the to-be-stratified population includes: Get the preset timing function; Determine the positive sample hit rate corresponding to the data contained in each preset period within the preset time range and the time of the corresponding point as factors of the time series function, and construct a trend graph, wherein the trend graph is used to reflect the mapping relationship between time and the positive sample hit rate; Obtain the current time corresponding to the population to be stratified; Based on the position of the current time in the trend graph, the positive sample hit rate of the to-be-stratified population is determined.

6. The object stratification method based on artificial intelligence according to claim 1, characterized in that: The stratification results obtained by stratifying the personnel to be stratified according to the positive sample hit rate of the population to be stratified, the prediction probability corresponding to each person to be stratified and the positive sample hit rate of each layer to be evaluated include: Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the positive sample hit rate of the population to be stratified, and obtain the positive sample lift of the population to be stratified in each layer to be evaluated; Calculate the product of the overall positive sample hit rate and the positive sample lift of the to-be-stratified population in each to-be-evaluated layer to obtain the sum of the positive sample hit rates of the to-be-stratified population in each to-be-evaluated layer; Determine the number of the subsample sets corresponding to each layer to be evaluated of the population to be stratified according to the sum of the positive sample hit rates of the population to be stratified in each layer to be evaluated; Calculate the quotient of the number of the sub-sample sets corresponding to each layer to be evaluated of the population to be stratified and the preset number of copies to obtain the proportion of the number of people in each layer to be evaluated of the population to be stratified; According to the order of the predicted probability corresponding to each person to be stratified from high to low, the persons to be stratified are sorted to obtain a target sequence; The target sequence is split according to the proportion of the population to be stratified in each layer to be evaluated to obtain the stratification result.

7. An object stratification device based on artificial intelligence, characterized in that: include: An acquisition unit, used to acquire historical recruitment data of an object, construct a sample set using the historical recruitment data, and split the sample set to obtain a training set and a validation set; A training unit, used to obtain a preset model with classification function, and train the preset model using the training set to obtain an intermediate model; An input unit, used to input the validation set into the intermediate model to obtain the predicted probability of each sample in the validation set; A sorting unit, used to sort each sample in the validation set in descending order of the predicted probability to obtain a sample sequence; A splitting unit, used to split the sample sequence from the beginning of the column to the end of the column according to a preset number of parts to obtain at least one sub-sample set; A calculation unit, used to calculate the overall positive sample hit rate of the verification set and the positive sample hit rate of each sub-sample set; A determination unit, used to obtain the positive sample hit rate of each level of the intermediate model, and determine at least one layer to be evaluated according to the positive sample hit rate of each level; A verification unit, used to verify the rationality of the positive sample hit rate of each layer to be evaluated according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set; The determining unit is further configured to determine the intermediate model as the target model when the rationality of the positive sample hit rate of each layer to be evaluated passes the verification; A prediction unit, used to obtain a population to be stratified and predict a positive sample hit rate of the population to be stratified; The input unit is further used to obtain the to-be-processed data corresponding to each person to be stratified in the to-be-stratified population, and input the to-be-processed data into the target model to obtain the predicted probability corresponding to each person to be stratified; A stratification unit, used to stratify the persons to be stratified according to the positive sample hit rate of the population to be stratified, the prediction probability corresponding to each person to be stratified and the positive sample hit rate of each layer to be evaluated, to obtain a stratification result; The verification unit verifies the rationality of the positive sample hit rate of each layer to be evaluated according to the overall positive sample hit rate and the positive sample hit rate of each sub-sample set, including: Determine the subsample set contained in each layer to be evaluated; Calculate the sum of the positive sample hit rates of the subsample sets contained in each layer to be evaluated, and use it as the positive sample hit rate of each layer to be evaluated; Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the overall positive sample hit rate to obtain the positive sample lift of each layer to be evaluated; Obtain the maximum value of the positive sample hit rate of the subsample set contained in each layer to be evaluated, and use it as the first threshold of each layer to be evaluated; Determine a database corresponding to the historical recruitment data of the object, and obtain from the database a minimum value of the positive sample hit rate corresponding to the data contained in each preset period within a preset time range; Calculate the quotient of the positive sample hit rate of each layer to be evaluated and the minimum value, and use it as the maximum positive sample hit rate of each layer to be evaluated; When it is detected that the positive sample lift and the maximum positive sample hit rate of the first to-be-evaluated layer are both less than or equal to the first threshold of the first to-be-evaluated layer, it is determined that the rationality of the positive sample hit rate of the first to-be-evaluated layer has passed the verification; or When it is detected that the positive sample lift and the maximum positive sample hit rate of the second layer to be evaluated are not both less than or equal to the first threshold of the second layer to be evaluated, it is determined that the rationality of the positive sample hit rate of the second layer to be evaluated has failed the verification.

8. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the artificial intelligence-based object stratification method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, which, when executed by a processor, enables the processor to perform the artificial intelligence-based object stratification method according to any one of claims 1 to 6.