Index generation method and device

Abstract

The invention is suitable for the technical field of Internet, and provides an index generation method and device. In the embodiment, the data of the source end is flattened, the front end of the page can generate and define the indexes through screening configuration, a set of systematic index generation rules is formed and stored in the database, and then the data of the indexes is generated according to the generation rules of the indexes through the middleware index interpretation generation algorithm. According to the method, various subsequent data products such as reports, visual reports, cockpit wide screens and supervision submitted indexes can select and use corresponding index data according to own index definitions and data verification, and subsequent management can form data changes caused by system and source end business development, so that the data wide tables of the source end can be changed by controlling all data, and the data quality is improved. Therefore, the data correctness of all indexes can be ensured. The problems that in an existing index system, index management is not unified, index calibers are not consistent, and the index process is not standard are solved.

Description

Technical Field

[0001] The present application relates to the field of Internet technology, and in particular to an indicator generation method and device. Background Art

[0002] In today's information age, enterprises have an increasing demand for data management and analysis, especially in the big data environment. How to extract valuable information from massive data has become the key to enterprise decision-making. Traditional data processing and analysis methods often face problems such as low efficiency and lack of flexibility, and it is difficult to meet the needs of modern enterprises for rapid response and deep insight into data.

[0003] As a core component of enterprise-level data storage and analysis, the Data Warehouse (DW) provides a unified and standardized data storage environment by integrating data from different sources. In the process of building a data warehouse, it usually includes multiple layers such as the Operational Data Store (ODS) and the Data Warehouse (DW). The ODS layer is mainly used for the initial integration and cleaning of data to provide a basis for subsequent data analysis; while the DW layer further processes and integrates the data in depth to form an analytical data set that meets business needs.

[0004] However, in actual applications, how to efficiently build wide table data that meets business needs from the ODS layer to the DW layer, and how to generate indicators with business value based on these wide table data, are important challenges faced in the process of data warehouse construction. Traditional practices often rely on manual definition and writing of complex SQL query statements, which is not only time-consuming and labor-intensive, but also difficult to ensure data consistency and accuracy.

[0005] In addition, with the continuous development of machine learning technology, especially the widespread application of neural networks in the field of data analysis and prediction, new ideas have been provided for the generation of indicators in data warehouses. By utilizing the powerful learning ability of neural networks, valuable indicators can be automatically learned and extracted from data, thereby improving the efficiency and accuracy of indicator generation.

[0006] Therefore, there is an urgent need for a method that can automatically build wide table data from the ODS layer to the DW layer and generate target indicators based on these wide table data using machine learning technology to meet the needs of modern enterprises for rapid response and deep insight into data. This indicator generation method can not only improve the efficiency of data processing and analysis, but also enhance the flexibility and interpretability of data, providing more powerful support for enterprise decision-making. Summary of the invention

[0007] In view of this, the present application provides an indicator generation method and device to solve the indicator management problems existing in the existing indicator system, such as inconsistent indicator management, inconsistent indicator caliber, and non-standard indicator process.

[0008] The first aspect of the present application provides an indicator generation method, the method comprising:

[0009] After obtaining the target source data, statistical analysis is performed on the source data at the ODS layer to determine the logical relationship between the tables at the ODS layer;

[0010] Acquire business logic instance information, and determine initial wide table data of the business logic instance information according to the logical relationship between the tables at the DW layer, wherein the business logic instance information includes business instance data and logical relationship;

[0011] Based on the initial wide table data, the target wide table data is flattened and designed to be output in combination with the logical relationship in the business logic instance information;

[0012] Generate a front-end page through the target wide table data, and determine a parameter list of the target wide table data according to the user's operation on the front-end page, wherein the parameter list includes a plurality of initial indicators, the operation logic and budget parameters of each initial indicator;

[0013] Determine the wide table information corresponding to each initial indicator from the target wide table data, determine the operation logic and budget parameters corresponding to each wide table information, input the wide table information, operation logic and budget parameters into a pre-trained indicator generation model to obtain the target indicator.

[0014] Optionally, the performing statistical analysis on the source data at the ODS layer includes:

[0015] According to the actual business of the system and whether subsequent data analysis requires detailed records of its historical development changes, the source data is divided into static data and dynamic data, wherein the static data includes user information, asset information, and approval information, and the dynamic data includes financing information, clearing information, and circulation information.

[0016] Optionally, the flattening design outputting the target wide table data in combination with the logical relationship in the business logic instance information includes:

[0017] A separation design for concurrent processing of the logical relationship is performed, whereby a target logical relationship in the logical relationship that has no dependencies on previous and subsequent data and does not affect each other is split into multiple sub-threads, and the first wide table data is outputted concurrently according to the target logical relationship through each sub-thread;

[0018] The second wide table data of the business logic instance information is output through the unsplit logical relationship, and the target wide table data is determined and output through the first wide table data and the second wide table data.

[0019] Optionally, the operation logic includes:

[0020] Time dimension processing logic, measurement statistical logic, and group statistical logic.

[0021] Optionally, the training method of the indicator generation model includes:

[0022] Acquire initial training data including historical wide table data, operation logic and budget parameters, pre-process the initial training data, and generate historical feature data;

[0023] Determine a training set, a test set, and a validation set of the indicator generation model according to the historical feature data;

[0024] The indicator generation model is trained using the training set data, and after each training, the performance of the indicator generation model is evaluated through the validation set and the parameters of the indicator generation model are adjusted to determine the target indicator generation model;

[0025] The target indicator generation model is evaluated through a test set, and after the evaluation passes, the target indicator generation model is determined as a trained indicator generation model.

[0026] A second aspect of the present application provides an indicator generation device, the device comprising:

[0027] A unit for determining the logical relationship between tables, for performing statistical analysis on the source data at the ODS layer after acquiring the target source data, to determine the logical relationship between tables at the ODS layer;

[0028] An initial wide table data determination unit, used to obtain business logic instance information, and determine the initial wide table data of the business logic instance information according to the logical relationship between the tables at the DW layer, wherein the business logic instance information includes business instance data and logical relationship;

[0029] A target wide table data determination unit, configured to output the target wide table data by flattening design based on the initial wide table data and in combination with the logical relationship in the business logic instance information;

[0030] A parameter list determination unit, used to generate a front-end page through the target wide table data, and determine a parameter list of the target wide table data according to the user's operation on the front-end page, wherein the parameter list includes a plurality of initial indicators, an operation logic and a budget parameter of each initial indicator;

[0031] The target indicator acquisition unit is used to determine the wide table information corresponding to each initial indicator from the target wide table data, determine the operation logic and budget parameters corresponding to each wide table information, and input the wide table information, operation logic and budget parameters into a pre-trained indicator generation model to obtain the target indicator.

[0032] Optionally, the unit for determining the logical relationship between tables performs statistical analysis on the source data at the ODS layer, including:

[0033] According to the actual business of the system and whether subsequent data analysis requires detailed records of its historical development changes, the source data is divided into static data and dynamic data, wherein the static data includes user information, asset information, and approval information, and the dynamic data includes financing information, clearing information, and circulation information.

[0034] Optionally, the target wide table data determination unit outputs the target wide table data by flattening the design in combination with the logical relationship in the business logic instance information, including:

[0035] A separation design for concurrent processing of the logical relationship is performed, whereby a target logical relationship in the logical relationship that has no dependencies on previous and subsequent data and does not affect each other is split into multiple sub-threads, and the first wide table data is outputted concurrently according to the target logical relationship through each sub-thread;

[0036] The second wide table data of the business logic instance information is output through the unsplit logical relationship, and the target wide table data is determined and output through the first wide table data and the second wide table data.

[0037] Optionally, the operation logic in the parameter list determination unit includes:

[0038] Time dimension processing logic, measurement statistical logic, and group statistical logic.

[0039] Optionally, the training method of the indicator generation model in the target indicator acquisition unit includes:

[0040] Acquire initial training data including historical wide table data, operation logic and budget parameters, pre-process the initial training data, and generate historical feature data;

[0041] Determine a training set, a test set, and a validation set of the indicator generation model according to the historical feature data;

[0042] The indicator generation model is trained using the training set data, and after each training, the performance of the indicator generation model is evaluated through the validation set and the parameters of the indicator generation model are adjusted to determine the target indicator generation model;

[0043] The target indicator generation model is evaluated through a test set, and after the evaluation passes, the target indicator generation model is determined as a trained indicator generation model.

[0044] In the embodiment provided in the present application, by flattening the source data, the front end of the page can generate and define the indicators through screening configuration, form a set of systematic indicator generation rules and store them in the database, and then generate the indicator data according to the indicator generation rules through the middleware indicator interpretation generation algorithm. Subsequent data products such as reports, visualization reports, cockpit wide screens, regulatory reporting indicators, etc. can choose to use the corresponding indicator data according to their own indicator definitions and data verification, and subsequent management can be systematic. Data changes caused by the development of the source business can be controlled by one control, changing the source data wide table, that is, the data correctness of all indicators can be guaranteed. This solves the problems of inconsistent indicator management, inconsistent indicator caliber, and non-standard indicator processes in the existing indicator system. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 A flow chart of the method provided in the embodiment of the present application;

[0046] Figure 2 Another flow chart provided for an embodiment of the present application;

[0047] Figure 3 Another flow chart provided for an embodiment of the present application;

[0048] Figure 4 A structural diagram of a device provided in an embodiment of the present application;

[0049] Figure 5 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] Exemplary embodiments will be described in detail herein, examples of which are shown in the accompanying drawings. When the following description refers to the drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application. Instead, they are merely examples of devices and methods consistent with some aspects of the present application as detailed in the appended claims.

[0051] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.

[0052] It should be understood that although the terms first, second, third, etc. may be used in the present application to describe various information, these information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of the present application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0053] The present application provides an indicator generation method to solve the problems existing in the existing indicator system, such as inconsistent indicator management, inconsistent indicator caliber, and non-standard indicator process.

[0054] The technical solution of the present application is described in detail with specific embodiments below. The following specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described in detail in some embodiments.

[0055] like Figure 1 As shown, it is a flowchart of an indicator generation method provided by the present application, and the process may include the following steps:

[0056] Step S101, after acquiring the target source data, statistical analysis is performed on the source data at the ODS layer to determine the logical relationship between tables at the ODS layer.

[0057] In this embodiment, ODS stands for Operational Data Store, which is an important component of the data warehouse architecture. It is located between the data source system and the data warehouse, and is mainly used to store the original data extracted from various business systems. The main features of the ODS layer include subject-oriented data organization, retention of the level of detail of the original data, real-time or near real-time data update frequency, and retention of historical data to support data backtracking.

[0058] This embodiment adopts the principle of data warehouse technology to classify the source data, perform statistical analysis on the source data at the ODS layer, and sort out the logical relationship between the tables at the ODS layer. The logical relationship between the tables at the ODS layer can be determined by the data flow diagram drawn by historical data, which shows the process of data flowing from various business systems to the ODS layer and then to higher levels such as DWD and DWS.

[0059] In another embodiment, performing statistical analysis on source data at the ODS layer includes:

[0060] According to the actual business of the system and whether subsequent data analysis requires detailed records of its historical development changes, the source data is divided into static data and dynamic data, wherein the static data includes user information, asset information, and approval information, and the dynamic data includes financing information, clearing information, and circulation information.

[0061] Step S102, obtaining business logic instance information, and determining initial wide table data of the business logic instance information according to the logical relationship between the tables at the dw layer, wherein the business logic instance information includes business instance data and logical relationship.

[0062] The DW layer, or data warehouse layer, is a core layer in the data warehouse architecture, responsible for storing and managing cleansed, integrated, and standardized data. The data at this layer is usually stored in a structured manner for easy query and analysis. Its main goal is to provide a unified data view, allowing enterprises to obtain data from multiple source systems and conduct comprehensive data analysis.

[0063] In this embodiment, based on the mathematical logic between tables specified by ODS and the logical relationship in the business logic instance information, the business instance data is processed into a business fact data table, such as the user table + external enterprise check data + approval data, and the detailed data of the user's entire life cycle is processed; the asset detail table + transaction record table + clearing record table + fund settlement table are processed into detailed data of the asset's entire life cycle. The detailed data is the initial wide table data mentioned above.

[0064] Step S103: Based on the initial wide table data and in combination with the logical relationship in the business logic instance information, the target wide table data is flatly designed and output.

[0065] In this embodiment, based on the initial wide table data of the dw layer, the logical relationship of each data in the initial wide table data is determined through the business logic instance information, and the target wide table data is flattened and designed to be output. For example: the detailed data of the user's entire life cycle, with the user as the main dimension [the finest granularity], the approval records, related information, asset overview, various attributes and other information are processed into fields by flattening, which can be used for the integrated output of indicators such as user management, approval statistics, and customer portraits in the future; the detailed data of the asset's entire life cycle, with the sub-assets of the transaction flow as the main dimension, the key transaction information (financing amount, financing time, split time, split amount, creditor holder, outflow party, transfer party, etc.), issuance key information, clearing key information, etc. are processed into fields by flattening, which can be used for the integrated management of indicators such as balance management, loan management, and quota management in the future.

[0066] In another embodiment, the flattening design outputting the target wide table data in combination with the logical relationship in the business logic instance information includes:

[0067] A separation design for concurrent processing of the logical relationship is performed, whereby a target logical relationship in the logical relationship that has no dependencies on previous and subsequent data and does not affect each other is split into multiple sub-threads, and the first wide table data is outputted concurrently according to the target logical relationship through each sub-thread;

[0068] The second wide table data of the business logic instance information is output through the unsplit logical relationship, and the target wide table data is determined and output through the first wide table data and the second wide table data.

[0069] In this embodiment, by performing a separation design for concurrent processing of logical relationships, logical relationships that have no data dependency and do not affect each other are split into multi-threaded concurrent data processing functions to improve data processing efficiency and ensure data repairability. In addition, the split data lines can be parallelized to ensure data integrity.

[0070] Step S104, generating a front-end page through the target wide table data, and determining a parameter list of the target wide table data according to the user's operation on the front-end page, wherein the parameter list includes several initial indicators, the operation logic and budget parameters of each initial indicator.

[0071] In this embodiment, the front-end page can be implemented by the finereport report visualization development tool to reduce the development work of code writing through the tool. However, it can also be implemented using Javaweb, JavaScript, C# and other development technologies.

[0072] The front-end page will reference the target wide table data that has been processed in the above steps, and visualize the table structure information in the target wide table data. For example, the rule configuration for indicator generation is converted from the code-written method to the drop-down multiple-select method on the front-end of the page, and a variety of statistical analysis methods are pre-designed for "metrics". At the same time, the SQL code input of the metric is retained, and multiple analysis scenarios of the wildcard metric are used. In addition, corresponding operation logic is configured for different analysis scenarios, such as time dimension processing logic, metric statistical logic, group statistical logic, and budget parameters.

[0073] In this embodiment, the front-end page needs to collect data processing rules, that is, "front-end indicator rule screening configuration", and convert the user's demand for the target wide table data into a parameter list required by the next middleware through page visualization operations. Figure 2 shown.

[0074] pass Figure 2 -S21 manages the indicators as a whole and uses Figure 2 -S22 and Figure 2 -S23 completes the addition and modification / deletion of indicators. The specific conversion logic design is Figure 2 -S24 and Figure 2 -S25, based on the principle of SQL query syntax, is designed into the front end of the page.

[0075] Step S105, determining the wide table information corresponding to each initial indicator from the target wide table data, determining the operation logic and budget parameters corresponding to each wide table information, and inputting the wide table information, operation logic and budget parameters into a pre-trained indicator generation model to obtain the target indicator.

[0076] This embodiment adopts the middleware indicator interpretation generation method, such as Figure 3 As shown, in Figure 3 - In S31, the page maintenance indicator rule triggers the complete indicator rule life cycle, and after the indicator is stable, it can be Figure 3 -The “analysis indicator rules” in S31 are updated regularly to form a data closed loop.

[0077] exist Figure 3 -S31 passes the parsed indicator rules to Figure 3 - Before S32, "analysis indicator rules" need to read the parameters of indicator generation from the indicator rule table of the database: flattened data wide table name, time dimension processing logic, measurement statistical logic, grouping statistical logic and supplementary SQL analysis and filling. The algorithm principle referred to by this method is the "back propagation algorithm", which is the BP algorithm of multi-layer neural network.

[0078] In actual work, we face a variety of systems, a wide variety of back-end mathematical logic, and a variety of indicator rules, but referring to the principle of the BP algorithm, we can eventually return the indicators to the calculation of the input layer, hidden layer, and output layer. Therefore, our general indicator generation method was implemented, through the input layer of the "business logic instance information" construction method, the "source data flattening" construction of the hidden layer, and the "middleware indicator interpretation generation method" construction of the output layer, so that it can be universally adapted to the generation of any indicator in any system, so that complex and diverse indicators can be systematically managed in an integrated manner.

[0079] In another embodiment, the training method of the indicator generation model includes:

[0080] Acquire initial training data including historical wide table data, operation logic and budget parameters, pre-process the initial training data, and generate historical feature data;

[0081] Determine the training set, test set and validation set of the indicator generation model according to the historical feature data; use the training set data to train the indicator generation model, and after each training, evaluate the performance of the indicator generation model through the validation set and adjust the parameters of the indicator generation model, such as the learning rate, regularization coefficient, number of network layers and number of neurons, to determine the target indicator generation model;

[0082] The target indicator generation model is evaluated through a test set, and after the evaluation passes, the target indicator generation model is determined as a trained indicator generation model.

[0083] In this embodiment, different abnormal data can be preprocessed by different methods. For example, missing values ​​can be processed by deletion, filling or prediction; abnormal values ​​can be identified by statistical methods or based on domain knowledge; and duplicate values ​​can be deleted. After the processing is completed, feature extraction is performed to generate historical feature data.

[0084] After feature extraction, the historical feature data is split into training set, validation set and test set. Common split ratios are 70% training set, 15% validation set, 15% test set, or 80% training set, 10% validation set, 10% test set.

[0085] So far, completed Figure 1 The process shown.

[0086] In an embodiment of the present application, by flattening the source data, the front end of the page can generate and define indicators through screening configuration, form a set of systematic indicator generation rules and store them in the database, and then generate indicator data according to the indicator generation rules through the middleware indicator interpretation generation algorithm. Subsequent data products such as reports, visualization reports, cockpit wide screens, regulatory reporting indicators, etc. can choose to use the corresponding indicator data according to their own indicator definitions and data verification, and subsequent management can be systematic. Data changes caused by the development of the source business can be controlled by one control, changing the source data wide table, that is, the data correctness of all indicators can be guaranteed. This solves the problems of inconsistent indicator management, inconsistent indicator caliber, and non-standard indicator processes in the existing indicator system.

[0087] The present application also provides an indicator generation device, such as Figure 4 As shown, the device comprises:

[0088] The table logical relationship determination unit 401 is used to obtain the target source data, perform statistical analysis on the source data at the ODS layer, and determine the table logical relationship at the ODS layer;

[0089] The initial wide table data determining unit 402 is used to obtain business logic instance information, and determine the initial wide table data of the business logic instance information according to the logical relationship between the tables at the DW layer, wherein the business logic instance information includes business instance data and logical relationship;

[0090] A target wide table data determination unit 403 is configured to output the target wide table data by flattening the design based on the initial wide table data and in combination with the logical relationship in the business logic instance information;

[0091] A parameter list determination unit 404 is used to generate a front-end page through the target wide table data, and determine a parameter list of the target wide table data according to the user's operation on the front-end page, wherein the parameter list includes a plurality of initial indicators, an operation logic and a budget parameter of each initial indicator;

[0092] The target indicator acquisition unit 405 is used to determine the wide table information corresponding to each initial indicator from the target wide table data, determine the operation logic and budget parameters corresponding to each wide table information, and input the wide table information, operation logic and budget parameters into a pre-trained indicator generation model to obtain the target indicator.

[0093] In another embodiment, the inter-table logical relationship determination unit performs statistical analysis on the source data at the ODS layer, including:

[0094] According to the actual business of the system and whether subsequent data analysis requires detailed records of its historical development changes, the source data is divided into static data and dynamic data, wherein the static data includes user information, asset information, and approval information, and the dynamic data includes financing information, clearing information, and circulation information.

[0095] In another embodiment, the target wide table data determination unit outputs the target wide table data by flattening the design in combination with the logical relationship in the business logic instance information, including:

[0096] A separation design for concurrent processing of the logical relationship is performed, whereby a target logical relationship in the logical relationship that has no dependencies on previous and subsequent data and does not affect each other is split into multiple sub-threads, and the first wide table data is outputted concurrently according to the target logical relationship through each sub-thread;

[0097] Output the second wide table data of the business logic instance information through the unsplit logical relationship, and then determine the target wide table data through the first wide table data and the second wide table data and output it

[0098] In another embodiment, the operation logic in the parameter list determination unit includes:

[0099] Time dimension processing logic, measurement statistical logic, and group statistical logic.

[0100] In another embodiment, the training method of the indicator generation model in the target indicator acquisition unit includes:

[0101] Acquire initial training data including historical wide table data, operation logic and budget parameters, pre-process the initial training data, and generate historical feature data;

[0102] Determine a training set, a test set, and a validation set of the indicator generation model according to the historical feature data;

[0103] The indicator generation model is trained using the training set data, and after each training, the performance of the indicator generation model is evaluated through the validation set and the parameters of the indicator generation model are adjusted to determine the target indicator generation model;

[0104] The target indicator generation model is evaluated through a test set, and after the evaluation passes, the target indicator generation model is determined as a trained indicator generation model.

[0105] The above embodiment of the present invention provides an indicator generation method, and based on this method provides an indicator generation device. Through the above method and device, the problems of inconsistent indicator management, inconsistent indicator caliber, and non-standard indicator process existing in the existing indicator system are solved.

[0106] This embodiment also discloses a computer device, such as Figure 3 As shown, the computer device includes a processor and a memory, wherein the memory stores at least one instruction, and the at least one instruction is loaded and executed by the processor to implement any of the above-mentioned indicator generation methods.

[0107] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A method for generating an indicator, characterized in that: The method comprises: After obtaining the target source data, the data warehouse technology principle is used to perform statistical analysis on the target source data at the ODS layer to determine the logical relationship between the tables at the ODS layer; Acquire business logic instance information, and determine initial wide table data of the business logic instance information according to the logical relationship between the tables at the DW layer, wherein the business logic instance information includes business instance data and logical relationship; Based on the initial wide table data, the target wide table data is flattened and designed to be output in combination with the logical relationship in the business logic instance information; Generate a front-end page through the target wide table data, and determine a parameter list of the target wide table data according to the user's operation on the front-end page, wherein the parameter list includes a plurality of initial indicators, the operation logic and budget parameters of each initial indicator; Determine the wide table information corresponding to each initial indicator from the target wide table data, determine the operation logic and budget parameters corresponding to each wide table information through the BP algorithm of the multi-layer neural network, and input the wide table information, operation logic and budget parameters into a pre-trained indicator generation model to obtain the target indicator.

2. The method according to claim 1, characterized in that The statistical analysis of the source data at the ODS layer includes: According to the actual business of the system and whether subsequent data analysis requires detailed records of its historical development changes, the source data is divided into static data and dynamic data, wherein the static data includes user information, asset information, and approval information, and the dynamic data includes financing information, clearing information, and circulation information.

3. The method according to claim 1, characterized in that The output of the target wide table data by flattening the design in combination with the logical relationship in the business logic instance information includes: A separation design for concurrent processing of the logical relationship is performed, whereby a target logical relationship in the logical relationship that has no dependencies on previous and subsequent data and does not affect each other is split into multiple sub-threads, and the first wide table data is outputted concurrently according to the target logical relationship through each sub-thread; The second wide table data of the business logic instance information is output through the unsplit logical relationship, and the target wide table data is determined and output through the first wide table data and the second wide table data.

4. The method according to claim 1, characterized in that The operation logic includes: Time dimension processing logic, measurement statistical logic, and group statistical logic.

5. The method according to claim 1, characterized in that The training method of the indicator generation model includes: Acquire initial training data including historical wide table data, operation logic and budget parameters, pre-process the initial training data, and generate historical feature data; Determine a training set, a test set, and a validation set of the indicator generation model according to the historical feature data; The indicator generation model is trained using the training set data, and after each training, the performance of the indicator generation model is evaluated through the validation set and the parameters of the indicator generation model are adjusted to determine the target indicator generation model; The target indicator generation model is evaluated through a test set, and after the evaluation passes, the target indicator generation model is determined as a trained indicator generation model.

6. An indicator generating device, characterized in that: The device comprises: A unit for determining the logical relationship between tables, which is used to obtain the target source data and then perform statistical analysis on the target source data at the ODS layer using the data warehouse technology principle to determine the logical relationship between tables at the ODS layer; An initial wide table data determination unit, used to obtain business logic instance information, and determine the initial wide table data of the business logic instance information according to the logical relationship between the tables at the DW layer, wherein the business logic instance information includes business instance data and logical relationship; A target wide table data determination unit, configured to output the target wide table data by flattening design based on the initial wide table data and in combination with the logical relationship in the business logic instance information; A parameter list determination unit, used to generate a front-end page through the target wide table data, and determine a parameter list of the target wide table data according to the user's operation on the front-end page, wherein the parameter list includes a plurality of initial indicators, an operation logic and a budget parameter of each initial indicator; The target indicator acquisition unit is used to determine the wide table information corresponding to each initial indicator from the target wide table data, and the BP algorithm of the multi-layer neural network determines the operation logic and budget parameters corresponding to each wide table information, and inputs the wide table information, operation logic and budget parameters into a pre-trained indicator generation model to obtain the target indicator.

7. The device according to claim 6, characterized in that The unit for determining the logical relationship between tables performs statistical analysis on the source data at the ODS layer, including: According to the actual business of the system and whether subsequent data analysis requires detailed records of its historical development changes, the source data is divided into static data and dynamic data, wherein the static data includes user information, asset information, and approval information, and the dynamic data includes financing information, clearing information, and circulation information.

8. The device according to claim 6, characterized in that The target wide table data determination unit outputs the target wide table data by flattening the design in combination with the logical relationship in the business logic instance information, including: A separation design for concurrent processing of the logical relationship is performed, whereby a target logical relationship in the logical relationship that has no dependencies on previous and subsequent data and does not affect each other is split into multiple sub-threads, and the first wide table data is outputted concurrently according to the target logical relationship through each sub-thread; The second wide table data of the business logic instance information is output through the unsplit logical relationship, and the target wide table data is determined and output through the first wide table data and the second wide table data.

9. The device according to claim 6, characterized in that The operation logic in the parameter list determination unit includes: Time dimension processing logic, measurement statistical logic, and group statistical logic.

10. The device according to claim 6, characterized in that The training method of the indicator generation model in the target indicator acquisition unit includes: Acquire initial training data including historical wide table data, operation logic and budget parameters, pre-process the initial training data, and generate historical feature data; Determine a training set, a test set, and a validation set of the indicator generation model according to the historical feature data; The indicator generation model is trained using the training set data, and after each training, the performance of the indicator generation model is evaluated through the validation set and the parameters of the indicator generation model are adjusted to determine the target indicator generation model; The target indicator generation model is evaluated through a test set, and after the evaluation passes, the target indicator generation model is determined as a trained indicator generation model.

Images