Model generation method, device and equipment based on customized joint model
By generating and matching standardized models, the problem of high and low efficiency of customized joint modeling is solved, and the rapid adaptation of production and multi-customer needs is achieved, reducing costs and improving efficiency.
Patent Information
- Application Number
- CN202410914541.9
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-09
- Publication Date
- 2024-11-12
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Customized joint modeling has high cost and low efficiency problems in the financial technology field, mainly due to the need to spend a lot of time and data resources for model development and tuning.
By generating a customized joint model based on the first customer data, and selecting the target model as a standardized model, it is used to match the second customer data and directly generate the corresponding model, thereby reducing model development costs and improving efficiency.
This method greatly reduces the cost and time of model development, improves efficiency, can quickly adapt to production and meet customized modeling needs of multiple customers, and reduces the cost of subsequent testing and launch.
Smart Images

Figure CN118940868A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of financial technology, and in particular to a model generation method, device and equipment based on a customized joint model. Background Art
[0002] In the field of financial technology, especially for Internet finance, banking, consumer finance, and risk control industries of small loan institutions, customized joint modeling products can accurately meet customers' specific customized modeling needs and are gradually gaining attention in the industry.
[0003] Customized joint modeling is based on the customized modeling needs of customers for their customer base, i.e., their loan customization population. According to the customized modeling needs, the data sources required by customers are determined, the pain points are solved, and specific data sources are selected for backtracking according to the needs. The backtracking time is 5-7 working days. After the backtracking data source results, the customer's on-site modeling development is carried out, and customized models are made according to the Y definition label provided by the customer. The general modeling process includes data source analysis, preprocessing, single feature analysis, model input variable screening, model development and optimization, and final model finalization. Because different data sources are needed for model development, it may be necessary to repeat the modeling process for multiple data sources in the early stage, and finally determine the model effects of different data sources; analyze the preliminary effects of different data sources to determine whether they can meet the customer's modeling needs. If they can meet the needs, the model will be reported and delivered for acceptance after the finalization. If the effects of different data sources cannot meet the customer's modeling needs, it is necessary to combine the models of different data sources, try different algorithms, and optimize the models until the customer's modeling requirements are met. After the model is finalized, it will be reported and delivered to the customer for acceptance.
[0004] Therefore, it usually takes a lot of time and data resources to develop customized models and tune the models. The customized joint modeling service model has gradually exposed the problems of high cost and low efficiency. Summary of the invention
[0005] In view of this, the purpose of the present invention is to provide a model generation method, device and equipment based on a customized joint model to overcome the problem of high cost and low efficiency of current customized joint modeling.
[0006] To achieve the above objectives, the present invention adopts the following technical solutions:
[0007] In a first aspect, an embodiment of the present application provides a model generation method based on a customized joint model, comprising:
[0008] generating a customized federated model based on the first customer data;
[0009] Based on the model effects and model information of the customized joint models, selecting a target customized joint model from the customized joint models as a standardized model;
[0010] Based on the second customer data, matching a target standardized model from the standardized models;
[0011] A model corresponding to the second customer data is generated based on the target standardized model.
[0012] Further, in some embodiments of the present application, the first customer data includes multiple groups of customer information and business information, and each group of the customer information corresponds to one or more groups of the business information;
[0013] The generating a customized joint model based on the first customer data includes:
[0014] For each set of customer information, determine a data source combination to be used based on the set of customer information and corresponding business information;
[0015] Perform data source backtracking based on the determined data source combination;
[0016] Based on the data source tracing results, analyze and screen the data source variables to determine the target variables;
[0017] Based on the target variable, a sub-model corresponding to each data source or combination of data sources of each group of business information under the group of customer information is generated.
[0018] Furthermore, in some embodiments of the present application, generating a customized joint model based on the first customer data further includes:
[0019] The sub-models corresponding to the multiple data sources are fused to obtain a fusion model.
[0020] Further, in some embodiments of the present application, the selecting a target customized joint model from the customized joint models as a standardized model based on the model effect and model information of the customized joint model includes:
[0021] Based on the business information in the first customer data, for the channel or customer group, a customized joint model with excellent model effect is selected from the multiple customized joint models as the first standardized model.
[0022] Further, in some embodiments of the present application, the selecting a target customized joint model from the customized joint models as a standardized model based on the model effect and model information of the customized joint model includes:
[0023] Based on the business information in the first customer data, a customized joint model that is targeted at the target data source and has excellent model effects for different channels or customer groups is selected from the multiple customized joint models as the second standardized model.
[0024] Further, in some embodiments of the present application, the second customer data includes multiple groups of customer information and business information, and each group of the customer information corresponds to one or more groups of the business information;
[0025] The matching a target standardized model from the standardized model based on the second customer data includes:
[0026] A first standardized model that is the same as the channel or customer group in the business information in the second customer data is matched from the first standardized model as the target standardized model.
[0027] Further, in some embodiments of the present application, generating a model corresponding to the second customer data based on the target standardized model includes:
[0028] The target standardized model is verified and evaluated, and if the verification and evaluation result meets the requirements, the target standardized model is used as the model corresponding to the second customer data.
[0029] Furthermore, in some embodiments of the present application, the standardized model includes a fusion model, and a sub-model corresponding to the data source or the combination of data sources.
[0030] In a second aspect, an embodiment of the present application further provides a model generation device based on a customized joint model, comprising:
[0031] A first generating module, configured to generate a customized joint model based on first customer data;
[0032] A screening module, configured to select a target customized joint model from the customized joint models as a standardized model based on the model effects and model information of the customized joint models;
[0033] A matching module, configured to match a target standardized model from the standardized model based on the second customer data;
[0034] The second generating module is used to generate a model corresponding to the second customer data based on the target standardized model.
[0035] In a third aspect, an embodiment of the present application further provides a model generation device based on a customized joint model, characterized in that it includes a processor and a memory, wherein the processor is connected to the memory:
[0036] Wherein, the processor is used to call and execute the program stored in the memory;
[0037] The memory is used to store the program, and the program is at least used to execute the above-mentioned model generation method based on the customized joint model.
[0038] The present invention relates to the field of financial technology, and specifically to a model generation method, device and equipment based on a customized joint model, the method specifically comprising: generating a customized joint model based on first customer data; selecting a target customized joint model from the customized joint model as a standardized model based on the model effect and model information of the customized joint model; matching a target standardized model from the standardized model based on second customer data; and generating a model corresponding to the second customer data based on the target standardized model. In this way, based on the previously generated customized joint model, a standardized model is determined, and when there is a new modeling requirement, the corresponding standardized model is directly matched for model development, which can greatly reduce the model development cost and improve the model development efficiency. BRIEF DESCRIPTION OF THE DRAWINGS
[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.
[0040] Figure 1 It is a flowchart of a model generation method based on a customized joint model provided in an embodiment of the present invention;
[0041] Figure 2 is a structural schematic diagram of a model generation device based on a customized joint model provided in an embodiment of the present invention;
[0042] Figure 3 It is a structural diagram of a model generation device based on a customized joint model provided in an embodiment of the present invention. DETAILED DESCRIPTION
[0043] To make the purpose, technical solution and advantages of the present invention clearer, the technical solution of the present invention will be described in detail below. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other implementation methods obtained by ordinary technicians in this field without creative work belong to the scope of protection of the present invention.
[0044] Figure 1This is a flow chart of a model generation method based on a customized joint model provided by an embodiment of the present invention. Figure 1 , this embodiment may include the following steps:
[0045] S101. Generate a customized joint model based on first customer data.
[0046] Specifically, the first customer data here is data provided to customers for building a customized joint model, including customer information, business information, sample data information, etc.
[0047] S102: Based on the model effects and model information of the customized joint models, a target customized joint model is selected from the customized joint models as a standardized model.
[0048] It should be noted that the first customer data mentioned in the above steps is not the data of one customer, but may be data provided by multiple groups of customers, such as multiple groups of customer information, each of which corresponds to at least one business information, and each business information may correspond to a data source or a combination of data sources. In the process of building a customized joint model, each data source or combination of data sources can eventually form a model, namely a sub-model. On this basis, according to actual needs (such as whether the model result data meets customer requirements), multiple sub-models can be fused to obtain a fusion model (which is actually a fusion of multiple data sources). The customized joint model mentioned in this application may include the sub-models and / or fusion models mentioned in the above process.
[0049] In practical applications, the model effects and model information of the customized joint model are collected, organized and summarized, from which the standardized model is determined and the relevant information is saved.
[0050] S103. Based on the second customer data, a target standardized model is matched from the standardized models.
[0051] S104. Generate a model corresponding to the second customer data based on the target standardized model.
[0052] Specifically, when there is a new modeling requirement, the customer provides new data, namely the second customer data (it should be noted that the customer at this time may be the same as or different from the customer in the above steps), and the new data is directly analyzed and judged in the above-mentioned standardized model to match the second customer data with the corresponding standardized model, namely the target standardized model, and generate a final model corresponding to the second customer data based on the target standardized model.
[0053] The model generation method based on the customized joint model provided by the present invention first generates a customized joint model based on the first customer data; then, based on the model effect and model information of the customized joint model, a target customized joint model is selected from the customized joint model as a standardized model; then, based on the second customer data, a target standardized model is matched from the standardized model; and finally, a model corresponding to the second customer data is generated based on the target standardized model. In this way, based on the previously generated customized joint model, a standardized model is determined, and when there is a new modeling requirement, the corresponding standardized model is directly matched for model development, which can greatly reduce the model development cost and improve the model development efficiency.
[0054] Further, in some embodiments of the present application, the first customer data includes multiple groups of customer information and business information, and each group of customer information corresponds to one or more groups of business information; on this basis, based on the first customer data, a customized joint model is generated, which may include: for each group of customer information, based on the group of customer information and the corresponding business information, determining the data source combination to be used; tracing back the data source based on the determined data source combination; based on the data source tracing results, analyzing and screening the data source variables to determine the target variables; based on the target variables, generating a sub-model corresponding to each data source or data source combination of each group of business information under the group of customer information.
[0055] For example, this can be achieved through the following specific steps:
[0056] Step 1: Receive multiple sets of customer information, business information and sample data information provided by the customer.
[0057] Step 2: Based on the customer information, business information, and sample data information provided by the customer, determine the multiple data source combinations used to build a customized joint model.
[0058] Step 3: After determining the sample data information and the data source combination used, perform backtracking on multiple groups of data sources.
[0059] Step 4: Based on the data source tracing results, implement data source variable analysis and screening.
[0060] For example, we can screen out variables with IV values (information value) greater than 0.2 and variable stability index PSI values less than 0.1. In practical applications, through the above variable analysis and screening, we can screen out variables with more information value, better variable effects, more stable variables, and more stable model effects for subsequent construction.
[0061] Step 5: Develop the model based on the variables screened above.
[0062] Specifically, a model may be fitted based on the variables obtained through the above screening, and algorithms such as decision trees and logistic regression may be used to fit the model, thereby obtaining a sub-model for each data source or combination of data sources.
[0063] Step 6: Based on multiple groups of data sources, fit different groups of data source models (i.e., the sub-models mentioned in the above steps) to finally form a fusion model.
[0064] In practical applications, it is also necessary to evaluate the effect of the model and monitor it online, which can be achieved through the following steps:
[0065] Step 7: Evaluation of model effectiveness.
[0066] In the embodiment of the present application, the effect of the model can be evaluated by using AUC, the model discrimination index ks value and the model stability evaluation PSI. Specifically, the AUC value range is [0,1], and the larger the value, the better the model effect; the ks value range is [0,1], and the larger the value, the better the model distinguishes good and bad samples; the model stability evaluation PSI is less than 0.1, and the model is relatively stable.
[0067] Step 8: Model online monitoring.
[0068] In practical applications, when monitoring the model effect, the effect of the model can be monitored by monitoring the changes in indicators such as AUC and KS of the model over time. In the embodiment of the present application, it can be stipulated that the KS after the model is online is above 0.2, and the AUC is above 0.6 to indicate that the model effect of the online operation is relatively good; and the model stability can also be monitored, such as detecting the PSI, an indicator of modeling stability. If the PSI is less than 0.1, it means that the model is relatively stable.
[0069] It should be noted that the process of building the customized joint model can be a daily operation process, which is to generate many customized joint models after a long period of accumulation, and on this basis, determine the standardized model based on the model effect and model information. In the embodiment of the present application, the generated customized joint models can be collected and summarized, the model effect can be determined, and the model that can be used as a standardized model in the future can be screened to obtain a standardized model.
[0070] Specifically, based on the business information in the first customer data (such as information on channels or customer groups), a customized joint model with excellent model effects is selected from multiple customized joint models for the channel or customer group as the first standardized model; and based on the business information in the first customer data, a customized joint model with excellent model effects for the target data source and for different channels or customer groups is selected from multiple customized joint models as the second standardized model (which can be applied to multiple channels or customer groups).
[0071] In addition, it should be noted that in the above embodiment, the second customer data is similar to the first customer data, including multiple groups of customer information and business information, and each group of customer information corresponds to one or more groups of business information (the only difference is that they are provided by different customers. The first customer data is used to generate respective customized joint models for corresponding customers before determining the standardized model, and the first customer data is used to generate a model for the corresponding customer in combination with the standardized model after determining the standardized model).
[0072] On this basis, in some embodiments of the present application, based on the second customer data, a target standardized model is matched from the standardized model, which may specifically include: matching the first standardized model that is the same as the channel or customer group in the business information in the second customer data from the first standardized model as the target standardized model, or directly using the second standardized model as the target standardized model, and then verifying and evaluating the target standardized model. If the verification and evaluation results meet the requirements, the target standardized model is used as the model corresponding to the second customer data.
[0073] In some embodiments of the present application, in order to reduce the computing power consumption of determining the standardized model in the early stage, after generating a plurality of customized joint models, only the information of the model is collected, summarized and analyzed. When there is a new model generation requirement (that is, when the second customer data is received), the target standardized model corresponding to the second customer data is directly determined in combination with the second customer data, and then the model corresponding to the second customer data is generated based on the target standardized model. The specific process may include the following steps:
[0074] Step 1: Collect and summarize model information and the effects of the corresponding model.
[0075] Specifically, in the process of building and generating a customized joint model, based on the above multiple groups of customers, multiple groups of business information of each customer, and the sub-models of the fusion of multiple groups of data sources corresponding to each group of business information or the fusion model formed by the fusion of multiple data source sub-models, the model information is collected and summarized, as well as information on indicators such as AUC\KS and other indicators of model effect. Among them, the model with excellent model effect is the model with no obvious increase / decrease in indicator monitoring, such as the model stability indicator PSI and the model discrimination indicator KS after the model is launched.
[0076] Step 2: Match the existing customized joint model based on the new modeling requirements.
[0077] Specifically, when a customer has a new modeling need or a new customer has a modeling need, when seeking customized modeling needs based on multiple sets of customer information and business information (i.e., second customer data) provided by the customer, the customized joint modeling products collected through statistics are analyzed and judged with the customer information or business information provided by the customer for matching.
[0078] For example, when a customer provides multiple groups of business information that is identical to the collected business information, the corresponding data source model and the corresponding model information are matched, including the discrimination value ks, data source combination information, and the like.
[0079] Step 4: Determine the multiple combined data sources of the model, and trace back the new data source (determined based on the second customer data).
[0080] Step 5: After the backtracking results of multiple sets of data sources are returned, the fused model is used to score the data (i.e., using the model, such as including an evaluation of the customer's customers, etc.) and the results of multiple sets of data are returned to the customer.
[0081] Step 6: The customer verifies multiple sets of data results and evaluates the model.
[0082] Specifically, the evaluation can be based on AUC, the model discrimination index ks value, and the model stability evaluation PSI. Among them, the AUC value range is [0,1], and the larger the value, the better the model effect; the ks value range is [0,1], and the larger the value, the better the model distinguishes good and bad samples; the model stability evaluation PSI is less than 0.1, and the model is relatively stable.
[0083] Step 7: If the test results of multiple sets of data meet the customer's modeling requirements, the joint modeling product of the customized requirements is converted into a standard product, and the model corresponding to the new requirements is obtained.
[0084] For example, based on the multiple groups of customer information and business information for building customized joint models, it is summarized and found that channel A / customer group has carried out customized joint modeling in different modeling customers, and has been called online, the model effect is relatively good, the call time is at least more than half a year, and there are models customized with different or the same data sources in channel A / customer group. At this time, when a new customized modeling demand is received, if the customer has channel A / customer group, it is possible to directly test the data source model (the data source model is the sub-model or fusion model mentioned in the above embodiment) that has a better effect and is used more frequently among different customers (generated models), directly backtest, analyze and judge the test results, and if it meets the requirements of new customers, it can be developed and called online. It can be understood that in this embodiment, the sub-model fusion model customized based on channel A / customer group is used as a standardized model.
[0085] By analogy, when there is a DEF channel / customer group, customized modeling has been carried out and called online among different modeling customers. When the effect is relatively good and the model effect has not decayed after a period of time, that is, the model effect is excellent, then when new customers have customized modeling needs in the DEF channel customer group, the sub-models and fusion models customized for the DEF channel / customer group can be directly used as the standardized model evaluation output.
[0086] In addition, standardized models that can be more widely used can also be determined. For example, the B data source sub-model or fusion model has customized models for different channels / customer groups (customer groups of customers who customize models) among different customers (customer groups of customers who customize models), and the data source effect ks is relatively good, and the various data indicators are relatively stable, and the online call situation is good without attenuation. Therefore, it can be determined that B data has relatively good effects on different channels / customer groups and different pricing (pricing for different customers by modeling customers, such as loan-related pricing), and the data source performance is relatively stable.
[0087] At this time, based on this situation, the sub-model and fusion model corresponding to the B data source can be output as a more generally applicable standard product (i.e., the second standardized model mentioned in the above embodiment). When a new customer has a customized modeling requirement, the sub-model or fusion model corresponding to the B data source can be directly tested (the fusion model corresponding to the B data source is a model obtained by fusing the sub-model corresponding to the B data source and the sub-models of other data sources based on the customer's effect requirements when the customized joint model was generated before) to see the test results and customer requirements. If the customer requirements are met, the development call can be put online and the sub-model or fusion model corresponding to the B data source can be output as a standard product.
[0088] It is understandable that standardized products, namely standardized models, can achieve rapid adaptation to production and large-scale replication, and can meet customers' customized modeling needs more quickly, thereby reducing costs and improving production efficiency. In addition, standardized products also have wide applicability and substitutability, and can meet the needs of more customers.
[0089] The model generation method based on customized joint model provided by the present invention has accumulated models of multiple customized joint modeling companies and the collection of related information data in the early stage, including the channel, customer group, pricing, sub-model KS effect, fusion model KS effect, whether it has been launched, the online time, the call duration, the online call volume, whether the online effect has decayed, etc. of the modeling samples, and analyzed, judged and matched them to match the models with the same channel or customer group and excellent model effect (i.e., matching the corresponding standardized model) for new modeling needs, which greatly reduces the development cost of subsequent models and improves the efficiency of subsequent model development.
[0090] The model generation method based on the customized joint model provided in the present application, on the basis of the customized joint model, after obtaining the customized joint model, collects information such as the customer group, pricing, data source, sub-model effect, fusion model effect, etc. used in the customized joint model, determines the standardized model, and matches it with the new modeling demand (that is, summarizes the sub-models and fusion models with the same customer group, pricing, and good effects as the new demand project, and can first analyze and determine the standardized model, and then match it; or analyze and match it after obtaining the information of the new demand), directly conducts an online test on the standardized model, and if the test effect is good, it can be directly put online for application, thereby reducing the testing cost of each data and improving production efficiency, without customized scoring based on customized customer groups, completing the test online in a short time, reducing the modeling cost, and can also quickly meet the customer's modeling needs.
[0091] Based on the same inventive concept, the present invention also provides a model generation device based on a customized joint model, which is used to implement the above method embodiment. Figure 2 Schematic diagram of the structure of a model generation device based on a customized joint model provided by an embodiment of the present invention. Figure 2 As shown, including:
[0092] A first generating module 21, configured to generate a customized joint model based on first customer data;
[0093] A screening module 22, configured to select a target customized joint model from the customized joint models as a standardized model based on the model effects and model information of the customized joint models;
[0094] A matching module 23, configured to match a target standardized model from the standardized model based on the second customer data;
[0095] The second generating module 24 is used to generate a model corresponding to the second customer data based on the target standardized model.
[0096] Regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.
[0097] Based on the same inventive concept, the present invention also provides a model generation device based on a customized joint model, which is used to implement the above method embodiment. Figure 3 Schematic diagram of the structure of a model generation device based on a customized joint model provided by an embodiment of the present invention. Figure 3As shown, the model generation device based on the customized joint model of this embodiment includes a processor 31 and a memory 32, and the processor 31 is connected to the memory 32. Among them, the processor 31 is used to call and execute the program stored in the memory 32; the memory 32 is used to store the program, and the program is at least used to execute the model generation method based on the customized joint model in the above embodiment.
[0098] The specific implementation scheme of the model generation device based on the customized joint model provided in the embodiment of the present application can refer to the implementation scheme of the model generation method based on the customized joint model in any of the above embodiments, which will not be repeated here.
[0099] It can be understood that the same or similar parts of the above embodiments can be referenced to each other, and the contents not described in detail in some embodiments can refer to the same or similar contents in other embodiments.
[0100] It should be noted that, in the description of the present invention, the terms "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "plurality" refers to at least two.
[0101] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, segment or portion of code that includes one or more executable instructions for implementing the steps of a specific logical function or process, and the scope of the preferred embodiments of the present invention includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by those skilled in the art to which the embodiments of the present invention belong.
[0102] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, a plurality of steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.
[0103] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.
[0104] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.
[0105] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc.
[0106] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.
[0107] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.
Claims
1. A model generation method based on a customized joint model, characterized in that: include: generating a customized federated model based on the first customer data; Based on the model effects and model information of the customized joint models, selecting a target customized joint model from the customized joint models as a standardized model; Based on the second customer data, matching a target standardized model from the standardized models; A model corresponding to the second customer data is generated based on the target standardized model.
2. The model generation method based on the customized joint model according to claim 1, characterized in that: The first customer data includes multiple sets of customer information and business information, and each set of the customer information corresponds to one or more sets of the business information; The generating a customized joint model based on the first customer data includes: For each set of customer information, determine a data source combination to be used based on the set of customer information and corresponding business information; Perform data source backtracking based on the determined data source combination; Based on the data source tracing results, analyze and screen the data source variables to determine the target variables; Based on the target variable, a sub-model corresponding to each data source or combination of data sources of each group of business information under the group of customer information is generated.
3. The model generation method based on the customized joint model according to claim 2 is characterized in that: The generating a customized joint model based on the first customer data further includes: The sub-models corresponding to the multiple data sources are fused to obtain a fusion model.
4. The model generation method based on the customized joint model according to claim 2 is characterized in that: The selecting a target customized joint model from the customized joint models as a standardized model based on the model effect and model information of the customized joint models includes: Based on the business information in the first customer data, for the channel or customer group, a customized joint model with excellent model effect is selected from the multiple customized joint models as the first standardized model.
5. The model generation method based on the customized joint model according to claim 2, characterized in that: The selecting a target customized joint model from the customized joint models as a standardized model based on the model effect and model information of the customized joint models includes: Based on the business information in the first customer data, a customized joint model that is targeted at the target data source and has excellent model effects for different channels or customer groups is selected from the multiple customized joint models as the second standardized model.
6. The model generation method based on the customized joint model according to claim 4 is characterized in that: The second customer data includes multiple sets of customer information and business information, and each set of the customer information corresponds to one or more sets of the business information; The matching a target standardized model from the standardized model based on the second customer data includes: A first standardized model that is the same as the channel or customer group in the business information in the second customer data is matched from the first standardized model as the target standardized model.
7. The model generation method based on the customized joint model according to claim 1, characterized in that: The generating a model corresponding to the second customer data based on the target standardized model includes: The target standardized model is verified and evaluated, and if the verification and evaluation result meets the requirements, the target standardized model is used as the model corresponding to the second customer data.
8. The model generation method based on the customized joint model according to claim 2, characterized in that: The standardized model includes a fusion model, and a sub-model corresponding to the data source or the combination of data sources.
9. A model generation device based on a customized joint model, characterized in that: include: A first generating module, configured to generate a customized joint model based on first customer data; A screening module, configured to select a target customized joint model from the customized joint models as a standardized model based on the model effects and model information of the customized joint models; A matching module, configured to match a target standardized model from the standardized model based on the second customer data; The second generating module is used to generate a model corresponding to the second customer data based on the target standardized model.
10. A model generation device based on a customized joint model, characterized in that: The invention comprises a processor and a memory, wherein the processor is connected to the memory: Wherein, the processor is used to call and execute the program stored in the memory; The memory is used to store the program, and the program is at least used to execute the model generation method based on the customized joint model described in any one of claims 1-8.
