Data parameter determination method and apparatus, electronic device, and program product

Abstract

The present disclosure discloses a data parameter determination method, comprising: obtaining an initial data set, the data of the initial data set being suitable for inputting into a main model to obtain a first analysis result; when the initial data set and / or the first analysis result meets a preset condition, obtaining a disaster recovery data set, and inputting data extracted from the disaster recovery data set into a disaster recovery model, wherein the disaster recovery model is used to obtain a second analysis result, and the disaster recovery data set is at least partially different from the initial data set; performing scaling processing on the second analysis result according to a scaling strategy to obtain a third analysis result corresponding to the main model and used to replace the first analysis result, wherein the scaling strategy represents the stretching or shrinking ratio of the first analysis result range of the main model and the second analysis result range of the disaster recovery model. The present disclosure proposes a data parameter determination scheme for realizing effective disaster recovery in a simple way, which has the advantages of high analysis result stability, high disaster recovery logic flexibility, and small constraint degree on the disaster recovery model and data.

Description

Technical Field

[0001] This disclosure relates to the field of financial technology, and in particular to methods and apparatus for determining data parameters, as well as related electronic devices and software products. Background Technology

[0002] In the fintech sector, data analytics models (such as scoring models) are an important service product, and their stability and quality are of paramount importance.

[0003] To ensure service stability and quality, current data analysis models provide corresponding disaster recovery mechanisms to address issues such as service failures and data source failures. However, these disaster recovery mechanisms have many shortcomings.

[0004] In some solutions, existing data analysis model disaster recovery mechanisms are often designed for model service failures. When the data source fails, the solution is typically to automatically return a backup analysis result (such as a backup score) and use that result as the final analysis outcome – a rather simplistic and crude approach. Current technologies also propose deploying dual models to address data source failures, with the auxiliary model (sometimes called the backup model) using a different data source than the primary model. However, the differences between the auxiliary and primary models can lead to instability issues in the analysis results, resulting in significant discrepancies between the analysis results generated by the auxiliary and primary models.

[0005] In view of this, there is an urgent need for improved data analysis models for financial technology.

[0006] The background description is provided for the purpose of understanding the relevant technologies in this field and is not intended as an admission of prior art. Summary of the Invention

[0007] Accordingly, this disclosure provides a data parameter determination method and apparatus, as well as related electronic devices and program products, which at least partially solve the above-mentioned problems.

[0008] In a first aspect, a method for determining data parameters is provided, comprising:

[0009] Obtain an initial dataset, the data of which is suitable for input into the main model to obtain the first analysis results;

[0010] When the initial dataset and / or the first analysis result meet the preset conditions...

[0011] Obtain a disaster recovery dataset, and input the data extracted from the disaster recovery dataset into a disaster recovery model, wherein the disaster recovery model is used to obtain a second analysis result, and the disaster recovery dataset is at least partially different from the initial dataset;

[0012] The second analysis result is scaled according to the scaling strategy to obtain a third analysis result corresponding to the main model and used to replace the first analysis result. The scaling strategy represents the stretching or shrinking ratio between the range of the first analysis result of the main model and the range of the second analysis result of the disaster recovery model.

[0013] In a second aspect, a data parameter determining apparatus is provided, comprising:

[0014] The first acquisition unit is configured to acquire an initial dataset, the data of which is suitable for inputting into the main model to obtain a first analysis result.

[0015] The second acquisition unit is configured to acquire a disaster recovery dataset when the initial dataset and / or the first analysis result meet the preset conditions, and input the data extracted from the disaster recovery dataset into the disaster recovery model, wherein the disaster recovery model is used to obtain the second analysis result, and the disaster recovery dataset is at least partially different from the initial dataset;

[0016] The scaling processing unit is configured to perform scaling processing on the second analysis result according to a scaling strategy to obtain a third analysis result corresponding to the main model and used to replace the first analysis result, wherein the scaling strategy characterizes the stretching or shrinking ratio of the range of the first analysis result of the main model and the range of the second analysis result of the disaster recovery model.

[0017] In a third aspect, an electronic device is provided, comprising: a processor and a memory storing a computer program, the processor being configured to perform the method described in the first aspect when the computer program is executed.

[0018] In a fourth aspect, a program product is provided, comprising a computer program, wherein the computer program, when executed by a processor, implements the method as described in the first aspect.

[0019] The data parameter determination scheme provided in this embodiment can obtain an initial dataset, the data of which is suitable for input into the main model to obtain a first analysis result; when the initial dataset and / or the first analysis result meet preset conditions, a disaster recovery dataset is obtained, and the data extracted from the disaster recovery dataset is input into the disaster recovery model, wherein the disaster recovery model is used to obtain a second analysis result, and the disaster recovery dataset is at least partially different from the initial dataset; the second analysis result is scaled according to a scaling strategy to obtain a third analysis result corresponding to the main model and used to replace the first analysis result, wherein the scaling strategy characterizes the stretching or shrinking ratio of the range of the first analysis result of the main model to the range of the second analysis result of the disaster recovery model. The technical solution provided by this disclosure can address specific issues with the main model or its dataset, such as missing data or abnormal first analysis results due to missing data or other reasons. In such cases, a disaster recovery model can be used, and its corresponding dataset can be called to obtain a second analysis result. A scaling (stretching and / or shrinking) strategy can then be employed to scale the second analysis model to obtain a third analysis result, replacing the first analysis result that should have been output by the main model. Therefore, the disaster recovery data parameter determination solution provided by this disclosure through an effective scaling (stretching and / or shrinking) strategy can fully meet the stability requirements of the analysis results and effectively overcome the distribution differences between the disaster recovery analysis results and the main analysis results. Furthermore, compared to some known solutions that improve analysis results based on the quantiles of historical analysis results generated by auxiliary models, the solution of this disclosure still maintains high stability of the analysis results. Moreover, the scaling strategy-based solution of this disclosure can dynamically adjust the disaster recovery logic by adjusting the scaling degree, thereby ensuring the stability of the analysis results for different clients over a longer period. Furthermore, compared to the quantile-based solutions known to the public, the scaling strategy-based solution of this disclosure offers flexible configuration options, such as based on model correspondence, model data density, and customer diversity. Additionally, compared to the quantile-based solutions known to the public, the scaling strategy-based solution of this disclosure imposes fewer constraints on the disaster recovery model, allowing for the selection of a suitable model from among numerous disaster recovery models through appropriate scaling strategy configuration. This does not sacrifice the input integrity of the disaster recovery model, expanding the range of available disaster recovery models and facilitating support for various types of disaster recovery. Moreover, compared to the quantile-based solutions known to the public, the scaling strategy-based solution of this disclosure can fully utilize third-party backup sources to replace faulty third-party sources in the disaster recovery model, resolving the issue of missing model input data.

[0020] Other optional features and technical effects of the embodiments of this disclosure are described in part below, and in part will be apparent from reading this document. Attached Figure Description

[0021] The embodiments of this disclosure will now be described in detail with reference to the accompanying drawings. The elements shown are not limited to the scale shown in the drawings, and the same or similar reference numerals in the drawings denote the same or similar elements, wherein:

[0022] Figure 1 A system architecture diagram for implementing the data parameter determination method of embodiments of the present disclosure is shown;

[0023] Figure 2 A flowchart of a data parameter determination method according to an embodiment of this disclosure is shown;

[0024] Figure 3 A flowchart of a data parameter determination method according to an embodiment of this disclosure is shown;

[0025] Figure 4 A flowchart of a data parameter determination method according to an embodiment of this disclosure is shown;

[0026] Figure 5 A flowchart of a data parameter determination method according to an embodiment of this disclosure is shown;

[0027] Figure 6 A flowchart of a data parameter determination method according to an embodiment of this disclosure is shown;

[0028] Figure 7 A schematic diagram of a user interface for configuring scaling strategies according to an embodiment of the present disclosure is shown;

[0029] Figure 8 A block diagram of a data parameter determination apparatus according to an embodiment of the present disclosure is shown; and

[0030] Figure 9 A structural diagram of an electronic device for implementing the methods of embodiments of the present disclosure is shown. Detailed Implementation

[0031] To make the objectives, technical solutions, and advantages of this disclosure clearer, the disclosure will be further described in detail below with reference to specific embodiments and accompanying drawings. Here, the illustrative embodiments and descriptions of this disclosure are used to explain this disclosure, but are not intended to limit this disclosure.

[0032] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0033] In the fintech field, risk or security data from relevant users or organizations (such as resource requesters) is typically used to build data analysis models. This data includes, but is not limited to, operational risk parameters, security parameters, compliance parameters, and credit stability parameters. These models are usually provided to users as Software as a Service (SaaS) through an analysis model platform (system). Users can request data analysis results from the model; in one example, the user is a client of the analysis model platform, such as a bank, insurance institution, or credit institution. In response to the data analysis request, the model retrieves the relevant data from the database to obtain the analysis results. However, due to data gaps or model anomalies, the analysis results obtained by the model may be abnormal, such as no results being obtained or abnormal results (e.g., deteriorated results). To address these anomalies, the analysis model platform is equipped with disaster recovery strategies. Currently, there are disaster recovery solutions for model service failures and data source failures; however, as mentioned in the background, both solutions have their own limitations. Furthermore, the disaster recovery mechanisms of current analytical models are often tightly bound to the model's own logic, resulting in insufficient flexibility. When the disaster recovery logic needs to be modified, the code must be modified and the model redeployed. In addition, existing disaster recovery solutions usually only support a single disaster recovery, which is difficult to adapt to different business scenarios and customers' different requirements for the stability of analytical results. Therefore, there is still room for improvement in disaster recovery solutions for analytical models.

[0034] Therefore, refer to Figure 1 The diagram illustrates a system architecture of an analytical model platform (system) according to an illustrative embodiment of this disclosure. This analytical model platform (system) features disaster recovery capabilities based on a scaling strategy, as further described below.

[0035] like Figure 1 As shown, the analysis model platform (system) 100 may include a model group 110 for one or more analysis models, one or more databases 120, a model management module 130, an automated testing module 140, an interface gateway 150, and a cache unit 160. Each database stores one or more datasets.

[0036] In this embodiment, model group 110 can be deployed in a distributed or clustered manner. In a distributed deployment, the various analysis models of model group 110 can be distributed across different geographical locations or different data centers to improve system scalability and fault tolerance. In a clustered deployment, the various analysis models of model group 110 can work collaboratively as a whole within the same data center, providing high availability and load balancing; this can be referred to as a model group cluster. Model group 110 can contain multiple analysis models, such as Model 1, Model 2, Model 3...Model N shown in the figure. At least one of these models can serve as the primary model when a client requests data analysis results, and at least another of these models can serve as a disaster recovery model for the primary model. It is conceivable that various different model invocation methods and disaster recovery modes are possible. For example, in some embodiments, the selection of the master model and disaster recovery model in model group 110 is fixed, that is, several models serve as the master model and several models serve as disaster recovery models for the master model; in other embodiments, the selection of the master model and disaster recovery model in model group 110 is variable, for example, but not limited to: mutual disaster recovery models, such as model 2 being the disaster recovery model of model 1 and model 1 being the disaster recovery model of model 2; the disaster recovery models are nested, such as model 2 being the disaster recovery model of model 1 and model 3 being the disaster recovery model of model 2; based on changes in customers or other conditions, the model can be selectively used as the master model or disaster recovery model, for example, model 1 is the master model for a data analysis request from a first customer, and model 2 is the master model and model 1 is the disaster recovery model of model 2 for a data analysis request from a second customer. The above and other various selection schemes for master models and disaster recovery models, as well as their variations or combinations, all fall within the scope of this disclosure, and specific selection schemes combined with certain specific embodiments of this disclosure can obtain further preferred embodiments.

[0037] In this disclosure, the datasets on which these models depend can be stored in database 120. In embodiments of this disclosure, multiple databases 120 may be provided, for example... Figure 1 DB1, DB2, and DB3, etc., are shown. In some embodiments, database 120 may store one or more datasets. In embodiments of this disclosure, database 120 may contain datasets in any suitable form and may invoke the corresponding datasets as needed, such as based on user-requested feature variables (which may be referred to as input parameters) and / or disaster recovery requirements. For example, in Figure 1In the illustrated embodiments, database DB1 can store several datasets A0, A1, A2, A3...An; database DB2 can store several datasets B0, B1, B2, B3...Bn; and database DB3 can store several datasets C0, C1, C2, C3...Cn. In different embodiments, database 120 can be distributed. For example, some databases may be databases storing data provided by storage platform providers, some databases may be databases storing all customer data, some databases may be third-party source databases, or any combination thereof; these situations fall within the scope of this disclosure. Therefore, different storage methods for datasets in the database, as well as different configuration forms of the database, can be reasonably set as needed in the embodiments of this disclosure, all of which fall within the scope of the embodiments of this disclosure.

[0038] like Figure 1 As shown, the interface gateway 150 is configured to receive a user's request to obtain analysis results and return the analysis results according to the user's request. In the event of disaster recovery, the disaster recovery method based on the scaling strategy of this embodiment can be used to obtain the disaster recovery analysis results, as further described below.

[0039] like Figure 1 As shown, cache unit 160 can store related model configurations, such as model product disaster recovery configurations (e.g., the dataset corresponding to the main model and the corresponding disaster recovery model) and scaling strategies (e.g., scaling function structures and their scaling parameters) and other middleware.

[0040] exist Figure 1 In the illustrated embodiment, a model management module 130 may optionally be provided, which may include or integrate a scaling configuration module. In other examples, it is conceivable that the scaling configuration module may be configured independently or integrated into other modules. With the help of the model management module 130 (scaling configuration module), operators, such as analysts, can configure scaling strategies (such as stretching or shrinking rules and their scaling parameters) as needed. In this case, the scaling strategy can be dynamically adjusted as needed. Additionally, the model management module 130 can also be used to configure other disaster recovery settings for the model product, such as determining the primary model used for a given customer's evaluation request, and determining or adjusting the disaster recovery model corresponding to the given primary model.

[0041] exist Figure 1 In the illustrated embodiment, an automated testing module 140 may also be provided, which can be used to test the effectiveness of scaling strategies or other disaster recovery strategies configured or changed by the analyst, such as whether the deployment was successful, the logic correctness, etc.

[0042] In the following text, we will also combine Figure 1The analytical model system shown is used to illustrate the methods, apparatus, electronic devices, program products, and storage media of different embodiments of this disclosure.

[0043] Therefore, this disclosure provides a method for determining data parameters. For example... Figure 2 As shown, the method for determining the data parameters may include steps S210 to S240:

[0044] S210: Obtain the initial dataset.

[0045] In this embodiment of the disclosure, the initial dataset corresponds to the main model, and the data in the initial dataset is suitable for input into the main model to obtain a first analysis result. In this embodiment of the disclosure, the analysis model (such as the main model) can provide analysis results based on the input data according to a specific operation process and / or logical algorithm, which is here the first analysis result.

[0046] In this embodiment, the datasets (including the initial dataset and the disaster recovery dataset) can have different dataset fields. For example, dataset Ai can have fields ai_field1...ai_fieldn; dataset Bi can have fields bi_field1...bi_fieldn; dataset Ci can have fields ci_field1...ci_fieldn (i refers to the i-th field from 0 to n, 0≤i≤n). Correspondingly, the data in the datasets (including the initial dataset and the disaster recovery dataset) can be sample data with different field values, that is, the sample data in the dataset can have corresponding values ​​for each field. In this embodiment, the field values ​​of the sample data in the dataset can be input into the analysis model as needed, and the corresponding fields constitute the feature variables (also called input parameters) used by the analysis model. In this embodiment, the data in the initial dataset can include sample data with different field (feature variable / input parameter) values, which can be input into the main model, and the main model can output the first analysis result accordingly based on the input data. The dataset fields (feature variables / input parameters) involve information about a certain dimension of the sample data (sample type), such as, but not limited to, age, education level, income status, default status, etc.

[0047] In this embodiment, an initial dataset can be obtained in response to a request, such as a customer's data analysis request. In this embodiment, all information, data, and signals involved are authorized by the user or fully authorized by all parties, and the collection, use, and processing of related data comply with the relevant laws, regulations, and standards of the relevant countries and regions. In this embodiment, the user or customer can be replaced by, for example, but not limited to, banks, insurance companies, and credit institutions, who can request analysis results as users, but generally cannot access or understand the configured disaster recovery strategy. This differs from operators, such as analysts, who can change (add, modify, and delete) the disaster recovery strategy.

[0048] In one specific embodiment, when a user request is received through the interface gateway 150, the user's (customer's) request, such as a user data analysis request, can be parsed to obtain the data information required by the user's request (such as a rating analysis request). In a specific example, the feature variables (input parameters) required for data analysis can be determined, and the dataset to be obtained can be determined based on the determined feature variables (input parameters). For example, for the first part of the input parameters (e.g., a0_field1 and a0_field2), it is determined that dataset A0 in database DB1 needs to be obtained; for the second part of the input parameters (e.g., b0_field2), it is determined that dataset B0 in database DB2 needs to be obtained; and for the third part of the input parameters (e.g., c0_field1, c0_field2, and c0_field3), it is determined that dataset C0 in database DB3 needs to be obtained. Alternatively, the dataset information requested by the user can be directly obtained to determine and obtain the corresponding initial dataset.

[0049] In another specific embodiment, the main model for analysis can be determined based on the user's data analysis request, and then, for example, the dataset required by the main model can be determined based on the determined main model. For example, the main model corresponding to the data analysis can be determined as Model 1 based on the user's request, and the associated initial dataset (such as A0 stored in DB1, B0 stored in DB2, and C0 stored in DB3) can be determined and obtained based on the determined main model (such as Model 1), for example, based on the feature variables (input parameters) required by the main model (such as Model 1).

[0050] In the embodiments of this disclosure, the initial dataset obtained can be one or more, and can come from a single database or multiple databases (data sources). This disclosure does not limit this.

[0051] In this embodiment of the disclosure, step S220 (not shown in the figure) may also be included: determining whether the initial dataset and / or the first analysis result meet the preset conditions.

[0052] In some embodiments, it can be determined whether the first analysis result meets preset conditions.

[0053] In this embodiment, before determining whether the first analysis result meets the preset conditions, the data parameter determination method also includes: S211 (not shown in the figure): inputting the data extracted from the initial dataset into the main model to obtain the first analysis result.

[0054] Therefore, determining whether the initial dataset and / or the first analysis result meet the preset conditions may include: S221 (not shown in the figure): verifying whether the first analysis result is abnormal, wherein the preset conditions include the first analysis result being abnormal.

[0055] In this embodiment, after receiving the relevant data from the initial dataset, the main model will process (infer or perform calculations) based on the relevant data from the initial dataset, such as the sample data corresponding to the input parameters involved in the main model, thereby obtaining the first analysis result. If there are problems with the data in the initial dataset and / or problems with the main model itself, it may lead to abnormal output results of the main model, including but not limited to no output value or errors, or output values ​​that do not meet the requirements. Therefore, the operation of obtaining the disaster recovery dataset described below can be performed.

[0056] In another embodiment, it can be determined whether the initial dataset meets preset conditions. Here, it is conceivable that step S211 above is an optional step. In other words, in this embodiment, although the data of the obtained initial dataset is suitable for input into the main model to obtain the first analysis result, it does not mean that the data of the initial dataset must be input into the main model. In this embodiment, the data of the initial dataset can be validated, without needing to input the data of the initial dataset into the main model to attempt to obtain the first analysis result. Specifically, determining whether the initial dataset and / or the first analysis result meets preset conditions may include: S222 (not shown in the figure): verifying whether the data of the initial dataset is missing, wherein the preset conditions include missing data in the initial dataset. Similarly, if the data of the initial dataset is found to be missing, the operation of obtaining the disaster recovery dataset described below can be performed. In embodiments of this disclosure, missing data in the initial dataset can refer to substantial missing data. For example, the missing data refers to the missing sample data corresponding to the fields (feature variables / input parameters) required to determine the analysis result according to the user request. In another example, the missing data refers to the missing proportion of sample data corresponding to the fields (feature variables / input parameters) being greater than a predetermined threshold.

[0057] In specific embodiments of this disclosure, a status flag can be set to indicate whether disaster recovery logic is executed. For example, when the preset conditions described above are met through the judgment / verification process, the status flag is set to active. Therefore, whether to execute disaster recovery logic can be determined simply by judging the status flag.

[0058] S230: When the initial dataset and / or the first analysis result meet the preset conditions, obtain the disaster recovery dataset and input the data extracted from the disaster recovery dataset into the disaster recovery model.

[0059] In this embodiment, the disaster recovery dataset corresponds to the disaster recovery model, and the data in the initial dataset is suitable for input into the disaster recovery model to obtain the second analysis result.

[0060] In this embodiment of the disclosure, the disaster recovery dataset may include sample data with different field (feature variable / input parameter) values, which can be input into the disaster recovery model, and the disaster recovery model can output a second analysis result accordingly based on the input data. Similarly, the disaster recovery dataset may be stored in database 120, such as database DB3.

[0061] In this embodiment of the disclosure, the disaster recovery dataset is at least partially different from the initial dataset. For example, the initial dataset can be A0, B0, and C0 as described above, and the disaster recovery dataset can be dataset C1 stored in database DB3. In this embodiment of the disclosure, the obtained disaster recovery dataset can be one or more, and can come from a single database or multiple databases (data sources), and this disclosure does not impose any limitations on this.

[0062] In this embodiment of the disclosure, the disaster recovery model associated with the main model can be determined based on the main model, and then the disaster recovery dataset can be determined and obtained accordingly based on the feature variables (input parameters) required by the disaster recovery model. For example, if the main model is determined to be model 1, the disaster recovery model configured for model 1 can be determined to be model 3, and the required disaster recovery dataset C1 can be determined based on the feature variables (such as c1_field1, c1_field2) required by model 3.

[0063] Furthermore, data from the disaster recovery dataset (such as C1) (such as sample data corresponding to the feature variables required by the disaster recovery model, such as c1_field1 and c1_field2) can be input into the disaster recovery model (such as model 3). The disaster recovery model will provide analysis results based on the input data, which is the second analysis result in this case.

[0064] It will become clear that the description of the dataset and the initial dataset above can be applied to the description of the disaster recovery dataset, and will not be repeated here. Similarly, the description of the main model above can be applied to the description of the disaster recovery model, and will not be repeated here.

[0065] S240: Scale the second analysis result according to the scaling strategy to obtain the corresponding master model and the third analysis result used to replace the first analysis result.

[0066] In this embodiment, the scaling strategy characterizes the stretching or shrinking ratio between the first analysis result range of the master model and the second analysis result range of the disaster recovery model. In this embodiment, stretching or shrinking, and the corresponding stretching or shrinking ratio, can be determined based on the correspondence between the analysis result ranges of the master model and the disaster recovery model. In a preferred embodiment, the scaling strategy is a stretching strategy, meaning the analysis result range of the disaster recovery model is always smaller than the analysis result range of the master model.

[0067] In this embodiment of the disclosure, the scaling strategy may have a scaling factor. In a further embodiment, the scaling strategy is adjustable, for example, the scaling function structure and / or the scaling factor is adjustable. For example, the scaling factor can be adjusted using the model management module 130 (scaling configuration module) described above. By using disaster recovery processing based on scaling relationships, this embodiment of the disclosure can handle problems such as missing data (sources) and other reasons causing main model anomalies in a simple and flexible manner. Moreover, it has fewer constraints on the models and data used for disaster recovery and is highly flexible. As will be described in detail below, the scaling strategy can also be reasonably configured according to the characteristics of the main model and the disaster recovery model and / or the characteristics of its data and / or customer characteristics, providing disaster recovery data analysis results with small differences and high stability.

[0068] In this embodiment, the scaling strategy is specifically implemented as a scaling function, which includes a univariate polynomial function, the coefficient of which is the scaling coefficient of the scaling function relationship. In a preferred embodiment, the scaling function relationship includes a univariate quadratic function.

[0069] For example, a quadratic scaling function can be expressed as:

[0070] y = a * x 2 +b*x+c

[0071] Where a, b, and c are scaling coefficients, y is the result of the third analysis, and x is the result of the second analysis.

[0072] Table 1 below provides an exemplary model data dependency and stretching strategy relationship table.

[0073] Table 1. Relationship between model data dependency and stretching strategy

[0074]

[0075] In the example shown in the table above, the primary model is Model 1, and its corresponding disaster recovery model is Model 3. The primary model depends on the initial datasets A0, B0, and C0. More specifically, the primary model depends on the feature variables (fields) a0_field1, a0_field2 (all of A0), b0_field2 (all of B0), c0_field1, c0_field2, and c0_field3 (all of C0). The disaster recovery model depends on the disaster recovery dataset C1. More specifically, the disaster recovery model depends on the feature variables (fields) c1_field1, c1_field2, and c1_field3 (all of C0). 1. All); Under normal circumstances, in response to a customer's request, the main model (Model 1) will be used to obtain the first analysis result (Score1) using the dataset on which it depends (sample data corresponding to the feature variables on which it depends); however, if it is necessary to obtain the disaster recovery analysis result due to the data source or model, the disaster recovery model (Model 3) will be used to obtain the second analysis result Score2, and then the disaster recovery analysis result will be determined according to the given scaling function relationship, which is a univariate quadratic scaling function relationship in this case, the third analysis result Score3 used to replace the first analysis result (Score1 in parentheses in Table 1).

[0076] In this example, the correspondence between the range of the disaster recovery analysis result (Score3) of Model 3 and the range of the analysis result (Score1) of Model 1 will determine whether the scaling strategy is to stretch or shrink. Preferably, the range of the disaster recovery analysis result (Score3) of Model 3 is smaller than the range of the analysis result (Score1).

[0077] The table also displays the user (client) API, which allows for different scaling strategies to be configured for different users (clients). In some embodiments, different scaling strategies refer to the same function structure (e.g., all are univariate quadratic functions) but with different scaling coefficients; in other embodiments, different scaling strategies refer to different function structures (e.g., all are univariate quadratic functions); and in some embodiments, different scaling strategies refer to a combination of both.

[0078] Here, step S240 may include: determining the scaling strategy corresponding to the user based on the user request, and scaling the second analysis result according to the scaling strategy corresponding to the user.

[0079] Furthermore, when the scaling strategy includes a univariate multivariate function with adjustable scaling coefficients, this univariate multivariate scaling function can be backward compatible with lower-order scaling functions by adjusting the scaling coefficients. For example, when the scaling strategy includes a univariate quadratic scaling function with adjustable coefficients, the scaling coefficient 'a' can be set to be non-zero. In this case, the scaling function will represent the quadratic relationship between the analysis result range of the master model and the analysis result range of the disaster recovery model. Alternatively, the scaling coefficient 'a' can be set to zero, and the scaling coefficient 'b' can be non-zero. In this case, the scaling function will represent the linear relationship between the analysis result range of the master model and the analysis result range of the disaster recovery model. In certain specific cases (as described below), the scaling coefficients 'a' and 'b' can be set to zero, and the scaling coefficient 'c' can be non-zero. In this case, the corresponding third analysis result within the analysis result range of the disaster recovery model will be constant. Further leveraging disaster recovery processing based on univariate multivariate scaling functions will have further advantages. As an explanation, and not a limitation, the analysis data of the master model and the analysis data of the disaster recovery model are usually not linearly corresponding, and their distribution densities may also differ. Scaling processing based on univariate multivariate scaling functions can better address this issue. Disaster recovery processing based on univariate multivariate scaling functions combined with adjustable scaling relationships offers greater technical benefits than simple superposition. For example, disaster recovery processing based on univariate multivariate scaling functions combined with adjustable scaling relationships allows operators, such as analysts, to handle dynamic changes in the correspondence between the master model and the disaster recovery model caused by various dynamic factors (such as time variations or different customer needs for the distribution of analysis results or disaster recovery analysis results), especially in cases of nonlinear correspondence and differences in distribution density.

[0080] In a further embodiment of this disclosure, multiple scaling functions can be configured for multiple second analysis result ranges of a single disaster recovery model. Each scaling function corresponds to one second analysis result range, and the multiple second analysis result ranges do not overlap with each other.

[0081] Accordingly, such as Figure 3 As shown, scaling the second analysis result according to the scaling strategy may include:

[0082] S241: Determine the range of the second analysis result to determine the corresponding scaling function, and

[0083] S242: Use the corresponding scaling function to scale the second analysis result.

[0084] In this embodiment, after scaling the second analysis result based on steps S241 and S241, a third analysis result for substitution can be obtained directly, or it can undergo further processing (e.g., in conjunction with other embodiments described herein, such as...). Figure 4 or Figure 5 The embodiments described herein all fall within the scope of this disclosure.

[0085] The technical solution of configuring multiple analysis result ranges and multiple scaling functions for a single disaster recovery model offers further advantages. For clarification, the description of a single disaster recovery model being configured with multiple analysis result ranges does not mean that the main model is equipped with only one disaster recovery model, but rather that multiple analysis result ranges are configured for one specific disaster recovery model. In a clarification, and not a limitation, the analysis results of the main model and the disaster recovery model may have corresponding differences across different ranges, such as differences in distribution density and / or slope and / or power factor and / or customer requirements. This problem can be effectively overcome by configuring different scaling functions for different ranges.

[0086] Reference Figure 3 and Figure 7 In a specific example, three scaling functions can be configured for the three ranges of the disaster recovery model (such as Model 3), where the scaling function for the range [n0, n1] is y = a0 * x. 2 +b0*x+c0; The scaling function for the range (n1,n2) is y=a1*x 2 +b1*x+c1; The scaling function for the range (n2,n3) is y=a2*x 2 +b²*x+c². Although Figure 7 Three ranges are shown, but more or fewer ranges are conceivable. Furthermore, it is also conceivable that the scaling strategy in this embodiment, such as the structure of the scaling function and / or the scaling coefficients, can be adjustable, and that the start point, end point, and open / closedness of the ranges are also adjustable. Although in Figure 7 In the examples shown, multiple scaling functions are presented as quadratic functions and can be configured as needed to be compatible with or incompatible with lower powers, as described above. For example, in one specific example, the scaling coefficient 'a' for multiple ranges can all be 0; in another specific example, the scaling coefficient 'a' for multiple ranges can be at least partially non-zero; in yet another specific example, in certain specific ranges, such as the range [n0, n1] or the range [n0, n1], or an additionally set lower limit range [0, n0) or upper limit range (n3, +∞), the scaling coefficients 'a' and 'b' can both be set to 0. In this case, for the second analysis results falling within these ranges, a corresponding constant alternative analysis result (the third analysis result) will be given. All of the above schemes fall within the scope of this disclosure.

[0087] As an optional step, the data parameter determination method of this disclosure embodiment may further include returning a first analysis result or a third analysis result to the client. In some embodiments, such as when using... Figure 1When using the analysis model system shown, the client can initiate a request to the interface gateway through the client. Correspondingly, the final analysis result can be returned to the client as either a first analysis result or a third analysis result, depending on whether disaster recovery has been performed. In this embodiment of the disclosure, the generation of the disaster recovery analysis result can be seamless for the client; that is, the client is unaware that the returned result is a disaster recovery analysis result (third analysis result), and always treats the returned analysis result as a normal analysis result. However, other embodiments of this disclosure also cover providing the client with a status indicating whether the result is a disaster recovery analysis result, in which case the client can selectively perform further processing based on this status.

[0088] The above-mentioned data parameter determination scheme has the advantages of high stability of analysis results, high flexibility of disaster recovery logic, and low constraint on disaster recovery models and data.

[0089] Further embodiments of this disclosure also propose a solution where a master model can be equipped with multiple disaster recovery models, thereby meeting customers' requirements for the stability of analysis results.

[0090] In further embodiments of this disclosure, the scaling process may also include filtering and / or averaging processes, which will be described in detail in the embodiments below, such as steps S441-S442 and S541-S542.

[0091] Another embodiment of this disclosure provides a method for determining data parameters. For example... Figure 4 As shown, the method for determining this data parameter may include steps S410, S431, S432, S441, and S442:

[0092] S410: Obtain the initial dataset.

[0093] The data in the initial dataset is suitable for input into the main model to obtain the first analysis results.

[0094] Step S410 can be referred to the above embodiments, especially step S210 and its features, and will not be repeated here.

[0095] In this embodiment, step S420 (not shown in the figure) may also be included: determining whether the initial dataset and / or the first analysis result meet the preset conditions.

[0096] Step S420 can be referred to the above embodiments, especially step S220 and its features, and will not be repeated here.

[0097] In this embodiment, when preset conditions are met, the data parameter determination method may further include S430 (not shown in the figure): obtaining a disaster recovery dataset, inputting the data extracted from the disaster recovery dataset into the disaster recovery model to obtain a second analysis result, and may specifically include:

[0098] S431: When the initial dataset and / or the first analysis result meet the preset conditions, obtain multiple sets of different disaster recovery datasets corresponding to multiple different disaster recovery models, and each set of disaster recovery datasets includes at least one disaster recovery dataset.

[0099] In this embodiment, multiple disaster recovery models can be configured accordingly.

[0100] In this embodiment, each disaster recovery model corresponds to a disaster recovery dataset group, which includes at least one disaster recovery dataset. Furthermore, any disaster recovery dataset can correspond to one disaster recovery model or multiple disaster recovery models. In a specific example, disaster recovery model A corresponds to one disaster recovery dataset group {a, b, c}, while disaster recovery model B corresponds to another disaster recovery dataset group {c, d}. However, it is understood that, as mentioned above, the disaster recovery datasets and their number within the disaster recovery dataset groups corresponding to the disaster recovery models are determined based on the feature variables required by the specific model, and will not be elaborated further here.

[0101] In a further embodiment, the main model can be directly equipped with multiple disaster recovery models. In this embodiment, a directly equipped disaster recovery model means that the disaster recovery analysis results of the directly equipped disaster recovery model, determined by a scaling strategy, can be used to replace the analysis results of the main model. In a specific example, the main model 1 is directly equipped with two disaster recovery models: Model 2 and Model 3. In some embodiments, the main model can be indirectly equipped with multiple sets of disaster recovery models, such as the nested disaster recovery models described above. In this embodiment, an indirectly equipped disaster recovery model means that the disaster recovery analysis results of the indirectly equipped disaster recovery model, determined by a scaling strategy, cannot be directly used to replace the analysis results of the main model, but is used to replace the analysis results of another disaster recovery model; and the analysis results of that other disaster recovery model, determined by another scaling strategy, can be used to replace the analysis results of the main model, or to replace the analysis results of another nested disaster recovery model, which can sequentially generate alternative analysis results according to its respective scaling strategy until an alternative analysis result is obtained to replace the analysis results of the main model. In one specific example, the main model 1 is directly equipped with a disaster recovery model: model 2, and model 2 itself is also directly equipped with a disaster recovery model: model 3. In this case, the main model 1 is indirectly equipped with disaster recovery model 3. These disaster recovery models can then be obtained as needed, for example, based on a given nesting depth (such as two levels) or other given conditions (as described below, until the termination condition for obtaining alternative disaster recovery analysis results is met). In some other embodiments, it can be a combination of both.

[0102] Accordingly, multiple disaster recovery datasets corresponding to the disaster recovery model used can be determined. Other embodiments disclosed herein can be referred to, especially step S230 and its features, which will not be elaborated here.

[0103] S432: Input multiple sets of data extracted from multiple disaster recovery datasets into multiple different disaster recovery models to obtain multiple alternative second analysis results.

[0104] In some embodiments, the multiple sets of data can be input into their respective corresponding disaster recovery models to obtain multiple alternative second analysis results corresponding to multiple different disaster recovery models. For example, the dataset includes disaster recovery dataset a, disaster recovery dataset b, and disaster recovery dataset c. Data a1 is extracted from disaster recovery dataset a, data b1 and b2 are extracted from disaster recovery dataset b, and data c3 is extracted from disaster recovery dataset c. Disaster recovery model A corresponds to one disaster recovery dataset group {a, b, c}, while disaster recovery model B corresponds to another disaster recovery dataset group {b, c}, and disaster recovery model C corresponds to yet another disaster recovery dataset group {a, c}. At this time, a set of data [a1, b1, b2, c3] from the dataset group corresponding to disaster recovery model A is input into disaster recovery model A. Similarly, data [b1, b2, c3] is input into disaster recovery model B, and data [a1, c3] is input into disaster recovery model C. The analysis results A1, B1, and C1 output by disaster recovery model A, disaster recovery model B, and disaster recovery model C are obtained respectively, thus obtaining three alternative second analysis results: A1, B1, and C1.

[0105] In this embodiment, the data parameter determination method may further include S440 (not shown in the figure): scaling the second analysis result according to the scaling strategy to obtain a third analysis result corresponding to the master model and used to replace the first analysis result, and may specifically include:

[0106] S441: Screen multiple alternative second analysis results to determine the target second analysis result that meets the conditions and its corresponding target disaster recovery model.

[0107] In the embodiments of this disclosure, various different filtering conditions and / or filtering methods and / or filtering rules can be used to determine the target second analysis result that meets the criteria. For example, the second analysis result closest to the average or located at the median can be used as the target second analysis result. Alternatively, another filtering module, such as a historical data filtering module, can be set to filter and obtain the target second analysis result. In another embodiment, for example, the fastest returning qualified second analysis result can be used as the target second analysis result. In some specific embodiments, a given threshold or range for the second analysis result can be set, thereby considering second analysis results that are greater than or equal to, less than, less than or equal to, or fall within the range or do not fall within the range as qualified (or unqualified) analysis results. The aforementioned threshold or range for the second analysis result can be determined based on experience or historical data.

[0108] In another embodiment, when nested disaster recovery models are included, the second analysis results can be obtained sequentially according to the nesting order. The filtering rules include, but are not limited to, the average of the analysis results closest to historical data. This filtering can be performed in conjunction with steps S431 and S432. In a specific example, for the same request, the second analysis result 100 of the sub-disaster recovery model 2 directly configured by the main model 1 is obtained, followed by the second analysis result 85 of the disaster recovery model 3 directly configured by the sub-disaster recovery model 2 (i.e., the disaster recovery model 3 indirectly configured by the main model 1). Since the average of the analysis results in the historical data of the main model is 88, 85 is selected as the target second analysis result.

[0109] In some cases where a rapid response is required, the second analysis result with the fastest response can be used as the target second analysis result and the acquisition of the second analysis result can be stopped.

[0110] Figure 4 The illustrated embodiments can be combined with other embodiments in a non-contradictory manner, especially Figure 2 and Figure 3 The method features, steps, and sub-steps of the illustrated embodiments, and such combinations fall within the scope of this disclosure. In particular, Figure 4 The illustrated embodiments can explicitly combine univariate multivariate scaling functions and / or adjustable scaling coefficients and / or multiple scaling functions corresponding to multiple ranges. Here, in one or more of the multiple disaster recovery models provided, a single disaster recovery model can be equipped with multiple analysis result ranges. As previously stated, the multiple analysis result ranges configured for a single disaster recovery model do not mean that the main model is equipped with only one disaster recovery model, but rather that multiple analysis result ranges are configured for one of the disaster recovery models themselves.

[0111] S442: Based on the scaling strategy corresponding to the target disaster recovery model, convert the target second analysis result into the target third analysis result.

[0112] In this embodiment, for the multiple disaster recovery datasets obtained, each disaster recovery model can be used to obtain the corresponding second analysis results. Other embodiments disclosed herein can be referred to, especially step S230 and its features, which will not be repeated here.

[0113] In this embodiment, a target second analysis result can be obtained by screening multiple unscaled candidate second analysis results, and then scaling processing can be performed to obtain a third analysis result to replace the first analysis result (main analysis result). This screening process can eliminate outliers and unstable second analysis results, thereby selecting the most reliable and stable result for subsequent processing, further improving the stability of disaster recovery analysis results.

[0114] In another embodiment where a primary model can be equipped with multiple disaster recovery models, multiple disaster recovery analysis results can be obtained by stretching multiple second analysis results, and the stretched multiple disaster recovery analysis results are averaged to obtain a third analysis result to replace the first analysis result (primary analysis result). This method disperses the risks that may arise from the analysis results of a single disaster recovery model, making the overall analysis result independent of the output of a single disaster recovery model, thereby reducing the impact of fluctuations in the analysis results of a single disaster recovery model on the final result and improving the stability of the disaster recovery analysis results.

[0115] Therefore, this other embodiment provides a method for determining data parameters. For example... Figure 5 As shown, the method for determining this data parameter may include steps S510, S531, S532, S541, and S542:

[0116] S510: Obtain the initial dataset.

[0117] The data in the initial dataset is suitable for input into the main model to obtain the first analysis results.

[0118] Step S510 can be referred to the above embodiments, especially step S210 and its features, and will not be repeated here.

[0119] In this embodiment, step S520 (not shown in the figure) may also be included: determining whether the initial dataset and / or the first analysis result meet the preset conditions.

[0120] Step S520 can be referred to the above embodiments, especially step S220 and its features, and will not be repeated here.

[0121] In this embodiment, when the preset conditions are met, the data parameter determination method may further include S530 (not shown in the figure): obtaining a disaster recovery dataset, inputting the data extracted from the disaster recovery dataset into the disaster recovery model to obtain a second analysis result, and may specifically include:

[0122] S531: When the initial dataset and / or the first analysis result meet the preset conditions, obtain multiple sets of different disaster recovery datasets corresponding to multiple different disaster recovery models. Each set of disaster recovery datasets includes at least one disaster recovery dataset.

[0123] S532: Input multiple sets of secondary data extracted from multiple disaster recovery datasets into multiple different disaster recovery models to obtain multiple secondary analysis results.

[0124] Step S520 can be referenced and combined with the above embodiments, especially steps S431 and S432 and their features, in a non-contradictory manner, and will not be repeated here. However, it will be understood that in this embodiment, the multiple second analysis results obtained will not be screened (optimized), and the multiple second analysis results will be scaled rather than "alternatives", as described in steps S541 and S542 below.

[0125] In this embodiment, the data parameter determination method may further include S540 (not shown in the figure): scaling the second analysis result according to the scaling strategy to obtain a third analysis result corresponding to the master model and used to replace the first analysis result, and may specifically include:

[0126] S541: Based on the scaling strategies corresponding to multiple different disaster recovery models, convert multiple second analysis results into multiple fourth analysis results, and

[0127] S542: The results of multiple fourth analyses are averaged to obtain the results of the third analysis.

[0128] In this embodiment of the disclosure, multiple second analysis results can be scaled to obtain multiple scaled intermediate analysis results, i.e., fourth analysis results. Then, the multiple fourth analysis results are averaged, for example, by calculating the average or the median, to obtain the final disaster recovery score (i.e., the third analysis result). In one example, when calculating the average, the averaging process may include averaging or weighted average, etc., and this disclosure does not limit this.

[0129] Figure 5 The illustrated embodiments can be combined with other embodiments in a non-contradictory manner, especially Figure 2 , Figure 3 and Figure 4 The method features, steps, and sub-steps of the illustrated embodiments, and such combinations fall within the scope of this disclosure. In particular, Figure 5 The illustrated embodiments can explicitly combine univariate multivariate scaling functions and / or adjustable scaling coefficients and / or multiple scaling functions corresponding to multiple ranges. Here, in one or more of the multiple disaster recovery models provided, a single disaster recovery model can be equipped with multiple analysis result ranges. As previously stated, the multiple analysis result ranges configured for a single disaster recovery model do not mean that the main model is equipped with only one disaster recovery model, but rather that multiple analysis result ranges are configured for one of the disaster recovery models themselves.

[0130] In this embodiment of the disclosure, scaling policies and / or other disaster recovery configurations can also be updated, for example, through the model management module 130 (scaling configuration module). In a further embodiment, the updated scaling policies and / or other disaster recovery configurations can be automatically tested and verified, and the verification results can be returned.

[0131] Specifically, such as Figure 6 As shown, the method for determining data parameters may also include:

[0132] S650: Real-time monitoring of scaling policy update operations.

[0133] S660: When an update operation is detected, retrieve the historical data of the primary model and disaster recovery model corresponding to the updated scaling strategy.

[0134] In this embodiment, as described above, the scaling strategy characterizes the stretching or shrinking ratio of the first analysis result range of the master model and the second analysis result range of the disaster recovery model. Based on this, the historical running data of the master model and the disaster recovery model involved in the updated scaling strategy can be retrieved accordingly.

[0135] S670: Based on an updated scaling strategy, use historical data from the disaster recovery model to perform scaling test processing to obtain third test analysis results.

[0136] In this embodiment, after obtaining the historical data of the disaster recovery model corresponding to the updated scaling strategy, the analysis results in the historical data of the disaster recovery model will be stretched or shrunk according to the updated scaling strategy to generate new third test analysis results.

[0137] S680: Based on the historical data of the master model and the results of the third test analysis, verify the validity of the scaling strategy update operation and return the verification result.

[0138] In this embodiment, after obtaining a new third test analysis result, the system compares and verifies this result with the historical analysis results in the historical data of the main model to evaluate the effectiveness of the new scaling strategy. In one example, the error between the third test analysis result and the historical analysis results in the historical data of the main model can be verified. If the error is less than a preset error threshold, the verification result is returned as the scaling strategy being effective; otherwise, the scaling strategy is returned as invalid. In a specific example, for instance, the scaling strategy involves one main model and three corresponding disaster recovery models, and the preset error threshold is 5%. At this point, tests are conducted based on, for example, historical requests A, B, and C. The historical analysis results in the main model's historical data are {70, 80, 90}. The third test analysis result obtained after applying the updated scaling strategy to the three disaster recovery models is {63, 72, 81}. Error calculation shows that {(65-63) / 65, (75-72) / 75, (85-81) / 85} ≈ {3.08%, 4.00%, 4.71%}, all of which are less than the error threshold of 5%. Therefore, the verification result returned indicates that the scaling strategy is effective.

[0139] Figure 6 The data parameter determination method and steps of the illustrated embodiment can preferably be implemented in the following ways: Figure 1 The analysis model system 100 shown is implemented.

[0140] This combination Figure 6 and Figure 1 Several more specific embodiments are described. For example... Figure 1 As shown, operators, such as analysts, can configure (update) scaling strategies, such as scaling factors (e.g., customer-based scaling factors), and / or other disaster recovery strategies in the model management module 130 (scaling configuration module). Operators, such as analysts, can implement this through a user graphical interface, and optionally through forms or specific data formats. (See reference) Figure 7 As shown, configurable scaling factors can be configured (updated) for example, by filling out a form to configure (update) the scaling factors corresponding to the disaster recovery model, such as the coefficients of a quadratic function, and the corresponding range of second analysis results can also be configured, and so on. Figure 7 As mentioned above, multiple ranges of second analysis results can be configured for a single model. As an alternative or supplement, such as... Figure 7 As shown, the scaling factor corresponding to the disaster recovery model can also be configured (updated) using a certain data structure. An example data structure can be {"n0,n1,0":"a0,b0,c0"}, which, for example, corresponds to the scaling function of the range [n0,n1]. The third "0" can represent the open and closed state of the range, which is marked as a left and right closed range in this case. Other open and closed state ranges can be marked with other state values, such as left open and right closed.

[0141] Once configured, operators, such as analysts, will upload the scaling policies and / or other disaster recovery configurations configured in the model management module 130 (scaling configuration module) (e.g., by clicking...). Figure 7 (As shown by the parsing button), updated information can be automatically pushed to cache unit 160; the service of interface gateway 150 can monitor the updates of cache unit in real time, automatically obtain cache update data, and store it in program memory. In response to the upload operation, model management module 130 (scaling configuration module) also notifies automated testing module 140 of successful deployment. Automated testing module 140 can execute automated test scripts to verify validity, such as whether scaling strategies, scaling parameters, and / or other disaster recovery configurations have been successfully deployed and are logically correct.

[0142] Therefore, in this embodiment, operators, such as analysts, can configure (update) scaling policies and / or other disaster recovery configurations online via an interface. The configured (updated) policies are automatically synchronized to relevant components, such as cache units and testing modules. Furthermore, automated testing and effective verification are achieved through real-time monitoring by the interface gateway 150, enabling online updates of effective disaster recovery configurations (including scaling policies). This improves the stability and reliability of scaling policy updates and enhances testing efficiency, facilitating rapid response to changes in business needs.

[0143] In the embodiments disclosed herein, such as Figure 6 The illustrated embodiments can be combined with other embodiments in a non-contradictory manner, especially Figure 2 , Figure 3 , Figure 4 and Figure 5 The method features, steps, and sub-steps of the illustrated embodiments, and such combinations fall within the scope of this disclosure. Furthermore, Figure 6 Steps S650 to S680 of the illustrated embodiment can be combined with Figure 2 , Figure 3 , Figure 4 and Figure 5 The method steps and sub-steps in the illustrated embodiment are executed in parallel, therefore, there are no time constraints between them, for example... Figure 6 Steps S650 to S680 of the illustrated embodiment can be performed in... Figure 2 , Figure 3 , Figure 4 and Figure 5 The method steps and sub-steps of the illustrated embodiments are executed before, after, or between them, and vice versa.

[0144] Here, the term "update operation" will be interpreted broadly to encompass adding, deleting, and adjusting (not adding or deleting) the corresponding operated object. In this context, "update operation" in a scaling strategy will cover updating the structure of the scaling function (such as a univariate polynomial function) itself (e.g., adjusting from a quadratic function to a cubic function), as well as updating the coefficients of the scaling function (such as a univariate polynomial function).

[0145] Correspondingly, this disclosure also provides a data parameter determination device. In a preferred embodiment, this data parameter determination device can be implemented in software.

[0146] In such Figure 8 In the illustrated embodiment, the data parameter determination device 800 may include a first acquisition unit 810, a second acquisition unit 830, and a scaling processing unit 840. The first acquisition unit 810 is configured to acquire an initial dataset, the data of which is suitable for input into the main model to obtain a first analysis result. The second acquisition unit 830 is configured to acquire a disaster recovery dataset when the initial dataset and / or the first analysis result meet preset conditions, and input the data extracted from the disaster recovery dataset into the disaster recovery model to obtain a second analysis result, wherein the disaster recovery dataset is at least partially different from the initial dataset. The scaling processing unit 840 is configured to scale the second analysis result according to a scaling strategy to obtain a third analysis result corresponding to the main model and used to replace the first analysis result, wherein the scaling strategy characterizes the scaling ratio between the range of the first analysis result of the main model and the range of the second analysis result of the disaster recovery model. Figure 8 In the embodiment shown, the data parameter determination device 800 may optionally include a judgment unit 820 (not shown in the figure), which can be configured to determine whether the initial dataset and / or the first analysis result meet preset conditions.

[0147] In the embodiments of this disclosure, one or more component features or sub-component features corresponding to the step features of the foregoing method embodiments may be optionally provided in the data parameter determination device as needed. For example, the first acquisition unit 810 of the data parameter determination device 800 may optionally have an input function, which may be further configured to input data extracted from the initial dataset into the main model to obtain a first analysis result. For ease of reading, many details of the device embodiments and several possible device embodiments will not be described in detail herein, but it will be understood that the structure or composition of the data parameter determination device in different embodiments of this disclosure can correspond to and implement any of the foregoing method embodiments or reasonable combinations thereof.

[0148] Furthermore, while the various units of the aforementioned data parameter determination device 800 are generally divided according to function, it should be understood that, particularly in embodiments where the data parameter determination device 800 is implemented in software, the various units of the data parameter determination device are not strictly implemented by one-to-one corresponding program modules. It is conceivable that the function of a single unit can be implemented by multiple modules, or that multiple units and their functions can be implemented by an integrated module, or that multiple units can be implemented by multiple but not one-to-one corresponding functional modules. For example, in one example, the function of the aforementioned second acquisition unit can be implemented by a second acquisition (sub) module and a second input (sub) module, respectively; in another example, the functions of the aforementioned first and second acquisition units can be implemented by an integrated acquisition module; in yet another example, the acquisition functions of the first and second acquisition units are implemented by a common acquisition (sub) module, and the output functions of the first and second acquisition units are implemented by a common input (sub) module. These variations all fall within the scope of the embodiments of this disclosure.

[0149] In embodiments of this disclosure, an electronic device may also be provided, including: a processor and a memory storing a computer program, the processor being configured to perform the method of any of the embodiments of this disclosure when running the computer program.

[0150] Figure 9 The illustration shows a method or electronic device 900 that can implement embodiments of the present disclosure. In some embodiments, it may include more or fewer electronic devices than illustrated. In some embodiments, it may be implemented using a single or multiple electronic devices. In some embodiments, it may be implemented using cloud-based or distributed electronic devices.

[0151] like Figure 9 As shown, the electronic device 900 includes a processor 901, which can perform various appropriate operations and processes based on programs and / or data stored in read-only memory (ROM) 902 or programs and / or data loaded from storage portion 908 into random access memory (RAM) 903. The processor 901 can be a multi-core processor or may contain multiple processors. In some embodiments, the processor 901 may include a general-purpose main processor and one or more special coprocessors, such as a graphics processing unit (GPU), a neural network processor (NPU), a digital signal processor (DSP), etc. The RAM 903 also stores various programs and data required for the operation of the electronic device 900. The processor 901, ROM 902, and RAM 903 are interconnected via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0152] The processor and memory described above are used together to execute a program stored in the memory. When the program is executed by a computer, it can implement the steps or functions of the file transfer method described in the above embodiments.

[0153] The following components are connected to I / O interface 905: an input section 906 including a keyboard, mouse, touchscreen, etc.; an output section 907 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 908 including a hard disk, etc.; and a communication section 909 including a network interface card such as a LAN card, modem, etc. The communication section 909 performs communication processing via a network such as the Internet. A drive 910 is also connected to I / O interface 905 as needed. A removable medium 911, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 910 as needed so that computer programs read from it can be installed into storage section 908 as needed. Figure 9 The diagram only shows a portion of the components and does not imply that the computer system 900 only includes... Figure 9 The components shown.

[0154] In some embodiments, the electronic device refers to a mobile terminal or computer, including mobile phones, vehicle terminals, smart TVs, etc. Taking a mobile phone as an example, the electronic device also includes a touch screen, external speaker, gyroscope, camera, 4G / 5G antenna, and other device modules.

[0155] The systems, devices, modules, or units described in the above embodiments can be implemented by a computer or its associated components. The computer may be, for example, a mobile terminal, smartphone, personal computer, laptop computer, in-vehicle human-machine interface device, personal digital assistant, media player, navigation device, game console, tablet computer, wearable device, smart TV, Internet of Things system, smart home, industrial computer, server, or a combination thereof.

[0156] Although not shown, in embodiments of this disclosure, a program product is provided, the program product including a computer program configured to be run to perform the methods of any of the embodiments of this disclosure.

[0157] Although not shown, in embodiments of this disclosure, a storage medium is provided storing a computer program configured to be executed, when run, to perform the methods of any of the embodiments of this disclosure.

[0158] The storage media in embodiments of this disclosure include articles that are permanent and non-permanent, removable and non-removable, capable of storing information by any method or technology. Examples of storage media include, but are not limited to, phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0159] The methods, programs, systems, apparatuses, etc., of the embodiments of this disclosure can be executed or implemented in a single or multiple networked computers, or practiced in a distributed computing environment. In the embodiments of this specification, in these distributed computing environments, tasks can be performed by remote processing devices connected via a communication network.

[0160] Those skilled in the art will understand that the embodiments described in this specification can be provided as methods, systems, or computer program products. Therefore, those skilled in the art will realize that the functional modules / units or controllers and related method steps described in the above embodiments can be implemented in software, hardware, or a combination of both.

[0161] Unless explicitly stated otherwise, the actions or steps of the methods or procedures described in the embodiments of this disclosure do not necessarily have to be performed in a specific order and can still achieve the desired results. In some implementations, multitasking and parallel processing are also possible or may be advantageous.

[0162] This document describes several embodiments of the present disclosure; however, for the sake of brevity, the descriptions of the embodiments are not exhaustive, and identical or similar features or portions between the embodiments may be omitted. In this document, "one embodiment," "some embodiments," "example," "specific example," or "some examples" refers to at least one embodiment or example applicable to the present disclosure, but not all embodiments. The above terms do not necessarily refer to the same embodiment or example. Without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of the different embodiments or examples.

[0163] The exemplary systems and methods of this disclosure have been specifically shown and described with reference to the foregoing embodiments, and are merely examples of the best mode for implementing the systems and methods. Those skilled in the art will understand that various changes can be made to the embodiments of the systems and methods described herein without departing from the spirit and scope of this disclosure as defined in the appended claims when implementing the systems and / or methods.

Claims

1. A method for determining data parameters, characterized in that, include: Obtain an initial dataset, the data of which is suitable for input into the main model to obtain the first analysis results; When the initial dataset and / or the first analysis result meet the preset conditions, a disaster recovery dataset is obtained, and the data extracted from the disaster recovery dataset is input into the disaster recovery model. The disaster recovery model is used to obtain a second analysis result. The disaster recovery dataset is at least partially different from the initial dataset. The preset conditions include missing data in the initial dataset and / or abnormal first analysis result. The second analysis result is scaled according to the scaling strategy corresponding to the disaster recovery model to obtain a third analysis result corresponding to the main model and used to replace the first analysis result. The scaling strategy represents the stretching or shrinking ratio between the range of the first analysis result of the main model and the range of the second analysis result of the disaster recovery model.

2. The method according to claim 1, characterized in that, The scaling strategy includes multiple scaling functions corresponding to multiple second analysis result ranges for a single disaster recovery model. Each scaling function corresponds to a second analysis result range, and the second analysis result ranges do not overlap with each other. The scaling process for the second analysis result according to the scaling strategy includes: Determine the range of the second analysis result and the corresponding scaling function, and The second analysis result is scaled using the corresponding scaling function.

3. The method according to claim 1, characterized in that, The scaling process also includes filtering and / or averaging.

4. The method according to claim 3, characterized in that, The process of obtaining the disaster recovery dataset, and inputting the data extracted from the disaster recovery dataset into the disaster recovery model to obtain the second analysis result, includes: Obtain multiple sets of different disaster recovery datasets, each corresponding to a different disaster recovery model. Each set of disaster recovery datasets includes at least one disaster recovery dataset. Multiple sets of data extracted from the multiple disaster recovery datasets are input into the multiple different disaster recovery models to obtain multiple alternative second analysis results; The step of scaling the second analysis result according to the scaling strategy to obtain a third analysis result corresponding to the main model and used to replace the first analysis result includes: The multiple candidate second analysis results are screened to determine the target second analysis result that meets the conditions and the target disaster recovery model corresponding to the target second analysis result. Based on the scaling strategy corresponding to the target disaster recovery model, the second analysis result of the target is transformed into the third analysis result.

5. The method according to claim 1, characterized in that, The process of obtaining the disaster recovery dataset, and inputting the data extracted from the disaster recovery dataset into the disaster recovery model to obtain the second analysis result, includes: Obtain multiple sets of different disaster recovery datasets, each corresponding to a different disaster recovery model. Each set of disaster recovery datasets includes at least one disaster recovery dataset. Multiple sets of second data extracted from the multiple disaster recovery datasets are input into the multiple different disaster recovery models to obtain multiple second analysis results; The step of scaling the second analysis result according to the scaling strategy to obtain a third analysis result corresponding to the main model and used to replace the first analysis result includes: Based on the scaling strategies corresponding to the various disaster recovery models, the multiple second analysis results are respectively converted into multiple fourth analysis results, and The third analysis result is obtained by averaging the multiple fourth analysis results.

6. The method according to any one of claims 1 to 4, characterized in that, Before obtaining the disaster recovery dataset, the data parameter determination method further includes: inputting the data extracted from the initial dataset into the main model to obtain a first analysis result; The initial dataset and / or the first analysis result meet preset conditions, including: The verification shows that the first analysis result is abnormal, wherein the preset conditions include the first analysis result being abnormal; and / or The verification revealed that the initial dataset was missing data, wherein the preset condition included the missing data in the initial dataset.

7. The method according to any one of claims 1 to 4, characterized in that, The method for determining data parameters further includes: Monitor the update operations of the scaling strategy in real time; When the update operation is detected, historical data of the main model and disaster recovery model corresponding to the updated scaling strategy are obtained. Based on the updated scaling strategy, historical data from the disaster recovery model is used to perform scaling tests, resulting in a third test analysis. Based on the historical data of the master model and the results of the third test analysis, the validity of the update operation of the scaling strategy is verified, and the verification result is returned.

8. A data parameter determination device, characterized in that, The data parameter determination device includes: The first acquisition unit is configured to acquire an initial dataset, the data of which is suitable for inputting into the main model to obtain a first analysis result. The second acquisition unit is configured to acquire a disaster recovery dataset when the initial dataset and / or the first analysis result meet preset conditions, and input the data extracted from the disaster recovery dataset into the disaster recovery model, wherein the disaster recovery model is used to obtain a second analysis result, and the disaster recovery dataset is at least partially different from the initial dataset, wherein the preset conditions include data missing in the initial dataset and / or abnormality in the first analysis result; The scaling processing unit is configured to scale the second analysis result according to the scaling strategy corresponding to the disaster recovery model to obtain a third analysis result corresponding to the main model and used to replace the first analysis result, wherein the scaling strategy characterizes the stretching or shrinking ratio between the range of the first analysis result of the main model and the range of the second analysis result of the disaster recovery model.

9. An electronic device, characterized in that, include: A processor and a memory storing a computer program, the processor being configured to perform the method of any one of claims 1-7 when running the computer program.

10. A program product comprising a computer program, wherein, When the computer program is executed by a processor, it implements the method as described in any one of claims 1-7.

Images