Data generation method and apparatus, storage medium, and electronic device

Abstract

The application provides a data generation method and device, a storage medium and an electronic device. The method comprises: obtaining initial time sequence characteristics of each of a plurality of basic variables, wherein one initial time sequence characteristic comprises a feature of a corresponding variable on a target time sequence; determining at least one derived variable, and calculating initial time sequence characteristics of each of the at least one derived variable based on the initial time sequence characteristics of each of the basic variables; generating target time sequence characteristics of each of the basic variables and target time sequence characteristics of each of the derived variables based on the initial time sequence characteristics of each of the basic variables and the initial time sequence characteristics of each of the derived variables, to generate to-be-analyzed data, wherein the to-be-analyzed data supports model input data of a target model, and the model input data is used for prediction by the target model. The embodiments of the application can generate to-be-analyzed data with relatively comprehensive feature coverage, thereby improving the accuracy of the prediction result through the to-be-analyzed data.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a data generation method, apparatus, storage medium, and electronic device. Background Technology

[0002] Currently, time series data is widely used in various application scenarios, such as user retention and sales forecasting. Existing technologies typically use target models to directly predict raw tabular data, but the feature coverage of the raw tabular data is incomplete, leading to low accuracy in the prediction results. Therefore, how to generate data to be analyzed with more comprehensive feature coverage to improve the accuracy of prediction results has become a research hotspot. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a data generation method, apparatus, storage medium, and electronic device to solve the problem that the prior art uses raw tabular data with incomplete feature coverage for prediction, resulting in low accuracy of prediction results; that is, embodiments of the present invention can generate data to be analyzed with more comprehensive feature coverage, thereby improving the accuracy of prediction results through the data to be analyzed.

[0004] According to one aspect of the present invention, a data generation method is provided, the method comprising:

[0005] Obtain the initial time series features of each basic variable among multiple basic variables. An initial time series feature includes the features of the corresponding variable on the target time series.

[0006] Identify at least one derived variable, and calculate the initial time-series characteristics of each derived variable among the at least one derived variable based on the initial time-series characteristics of each of the basic variables;

[0007] Based on the initial time-series features of each basic variable and each derived variable, target time-series features of each basic variable and target time-series features of each derived variable are generated respectively to generate data to be analyzed. The data to be analyzed supports the determination of model input data for the target model, so as to predict the model input data through the target model.

[0008] According to another aspect of the present invention, a data generation apparatus is provided, the apparatus comprising:

[0009] The acquisition unit is used to acquire the initial time series features of each basic variable among multiple basic variables. An initial time series feature includes the features of the corresponding variable on the target time series.

[0010] A processing unit is configured to determine at least one derived variable and, based on the initial time-series characteristics of each of the at least one derived variable, calculate the initial time-series characteristics of each derived variable among the at least one derived variable.

[0011] The processing unit is further configured to generate target time-series features of each basic variable and target time-series features of each derived variable based on the initial time-series features of each basic variable and the initial time-series features of each derived variable, respectively, to generate data to be analyzed. The data to be analyzed supports the determination of model input data for a target model, so as to predict the model input data through the target model.

[0012] According to another aspect of the present invention, an electronic device is provided, the electronic device including a processor and a memory storing a program, wherein the program includes instructions that, when executed by the processor, cause the processor to perform the methods mentioned above.

[0013] According to another aspect of the present invention, a non-transitory computer-readable storage medium is provided storing computer instructions for causing a computer to perform the methods mentioned above.

[0014] In this embodiment of the invention, after obtaining the initial time-series features of each of the multiple basic variables, at least one derived variable can be determined. Based on the initial time-series features of each basic variable, the initial time-series features of each derived variable are calculated. Each initial time-series feature includes the characteristics of the corresponding variable on the target time series. Furthermore, based on the initial time-series features of each basic variable and each derived variable, target time-series features of each basic variable and each derived variable can be generated respectively to generate data to be analyzed. The data to be analyzed supports the model input data for determining the target model, so as to predict the model input data through the target model. Therefore, this embodiment of the invention can generate data to be analyzed with relatively comprehensive feature coverage, thereby improving the accuracy of prediction results. In addition, the data to be analyzed has strong scalability and is easy to maintain. Attached Figure Description

[0015] Further details, features, and advantages of the invention are disclosed in the following description of exemplary embodiments in conjunction with the accompanying drawings, in which:

[0016] Figure 1 A flowchart illustrating a data generation method according to an exemplary embodiment of the present invention is shown;

[0017] Figure 2 A schematic diagram of a basic variable according to an exemplary embodiment of the present invention is shown;

[0018] Figure 3 A flowchart illustrating another data generation method according to an exemplary embodiment of the present invention is shown;

[0019] Figure 4 A schematic diagram of a derived variable according to an exemplary embodiment of the present invention is shown;

[0020] Figure 5 A schematic diagram of a timing feature according to an exemplary embodiment of the present invention is shown;

[0021] Figure 6 A schematic block diagram of a data generation apparatus according to an exemplary embodiment of the present invention is shown;

[0022] Figure 7 A structural block diagram of an exemplary electronic device that can be used to implement embodiments of the present invention is shown. Detailed Implementation

[0023] Embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While some embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of the invention. It should be understood that the accompanying drawings and embodiments are for illustrative purposes only and are not intended to limit the scope of protection of the invention.

[0024] It should be understood that the various steps described in the method embodiments of the present invention may be performed in different orders and / or in parallel. Furthermore, the method embodiments may include additional steps and / or omit the steps shown. The scope of the present invention is not limited in this respect.

[0025] The term "comprising" and its variations as used herein are open-ended, meaning "including but not limited to". The term "based on" means "at least partially based on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". Definitions of other terms will be given in the following description. It should be noted that the concepts of "first", "second", etc., mentioned in this invention are used only to distinguish different devices, modules, or units, and are not intended to limit the order of functions performed by these devices, modules, or units or their interdependencies.

[0026] It should be noted that the terms "a" and "a plurality of" used in this invention are illustrative rather than restrictive. Those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".

[0027] The names of the messages or information exchanged between the multiple devices in the embodiments of the present invention are for illustrative purposes only and are not intended to limit the scope of these messages or information.

[0028] It should be noted that the execution subject of the data generation method provided in this embodiment of the invention can be one or more electronic devices, and this invention does not limit this; wherein, the electronic device can be a terminal (i.e., a client) or a server. Therefore, when the execution subject includes multiple electronic devices, and among the multiple electronic devices includes at least one terminal and at least one server, the data generation method provided in this embodiment of the invention can be jointly executed by the terminal and the server. Accordingly, the terminal mentioned herein may include, but is not limited to: smartphones, tablets, laptops, desktop computers, smartwatches, smart voice interaction devices, smart home appliances, vehicle terminals, aircraft, etc. The server mentioned herein can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms, etc.

[0029] Based on the above description, this embodiment of the invention proposes a data generation method, which can be executed by the aforementioned electronic device (terminal or server); or, the data generation method can be executed jointly by the terminal and the server. For ease of explanation, the following description will use the execution of the data generation method by an electronic device as an example; such as Figure 1 As shown, the data generation method may include the following steps S101-S103:

[0030] S101, obtain the initial time series features of each basic variable among multiple basic variables. An initial time series feature includes the features of the corresponding variable on the target time series.

[0031] It should be noted that the data generation method mentioned in this invention can be applied to object retention scenarios, as well as sales forecasting scenarios, and this invention does not limit it in this regard. For ease of explanation, the embodiments of this invention will subsequently be described using object retention scenarios as the application scenario, and the above-mentioned basic variables as variables under the object retention scenario.

[0032] In one implementation, when the aforementioned multiple basic variables are variables in an object retention scenario, these multiple basic variables may include user-side basic variables, product-side basic variables (i.e., internal product strategy-side basic variables), and environment-side basic variables (such as competitor-side basic variables). Specifically, the aforementioned multiple basic variables may include, but are not limited to: at least one basic variable under [user] internal borrowing behavior (i.e., at least one basic variable under user-side internal borrowing behavior), at least one basic variable under [user] other internal behaviors, at least one basic variable under [product] internal strategy actions (i.e., at least one basic variable under product-side internal strategy actions), at least one basic variable under [user] competitor borrowing behavior (i.e., external borrowing behavior), at least one basic variable under [environment] competitor strategy actions (i.e., external strategy actions), and at least one basic variable under [environment] other external behaviors (i.e., at least one basic variable under environment-side other external behaviors), etc. Figure 2 As shown; however, this invention does not limit this. In other words, multiple basic variables can be divided into multiple target categories, which may include, but are not limited to: internal borrowing performance, other internal performance, internal strategic actions, competitor borrowing performance, competitor strategic actions, and other external behaviors, etc. Among them, external influences such as competitors and the environment can be referred to as external, and internal factors such as the product and the user's use of the product can be referred to as internal.

[0033] Optionally, at least one basic variable under loan performance (such as internal loan performance and external loan performance) may include, but is not limited to: loan amount, actual loan amount, whether overdue, credit utilization rate, and whether prepayment was made, etc., and this invention does not limit this. Optionally, at least one basic variable under other internal performance may include, but is not limited to: accessed pages, number of logins, use of APP (Application), and data tracking, etc., and this invention does not limit this. Optionally, at least one basic variable under strategic actions (such as internal strategic actions and external strategic actions), may include, but is not limited to: number of telemarketing calls, number of times price reductions were offered (i.e., number of interest rate reductions), and coupon issuance, etc., and this invention does not limit this. Optionally, at least one basic variable under other external behaviors may include, but is not limited to: number of multiple loans by commercial banks and number of multiple loans by rural banks, etc., etc., and this invention does not limit this. Here, multiple loans refer to credit activities within financial institutions.

[0034] For example, such as Figure 2As shown, the basic variables between users and products can include the number of telemarketing calls, the number of price reductions, and product usage behavior (such as basic variables under the performance of internal loan documents). The basic variables between users and the environment can include basic variables under the performance of competitor loan documents. The basic variables between products and the environment can include basic variables such as interest rates and credit limit comparisons. Optionally, multiple basic variables can also include the user's age, gender, income, real estate, and owned businesses.

[0035] In another implementation, the multiple basic variables may include at least one variable under each of the multiple target categories. That is, the multiple basic variables can be divided into multiple target categories, and each target category may include at least one variable. Optionally, the multiple target categories may include, but are not limited to: external credit reporting - multiple borrowing, external credit reporting - pricing and credit limit, external credit reporting - balance, external credit reporting - new loan, internal - loan, internal - data tracking, internal - strategy - operation, and internal - strategy - risk control, etc. The present invention does not limit this. Optionally, based on the user's internal and external behavior, the electronic device may divide the multiple basic variables into two major categories: external and internal, and further into multiple target categories. It should be noted that this invention does not limit the specific content of the multiple basic variables, that is, it does not limit the number of basic variables among the multiple basic variables. Optionally, the number of basic variables included in the multiple basic variables can be 266, and the external credit reporting - multiple borrowing can include 19 basic variables, the external credit reporting - pricing and credit limit can include 17 basic variables, the external credit reporting - balance can include 9 basic variables, the external credit reporting - new loan can include 15 basic variables, the internal - loan can include 30 basic variables, the internal - data tracking can include 122 basic variables, the internal - strategy - operation can include 33 basic variables, and the internal - strategy - risk control can include 16 basic variables, and so on. Here, pricing can also be referred to as interest rate or price, etc.

[0036] Optionally, the basic variables under External Credit Reporting - Multiple Loans (also known as External Multiple Loans) may include, but are not limited to: the number of multiple loans to microfinance institutions, the number of multiple loans to commercial banks, and the number of multiple loans to rural banks; the basic variables under External Credit Reporting - Pricing and Limits (also known as External Pricing and Limits) may include, but are not limited to: the maximum limit of competitors, the minimum price of competitors, and the average limit of consumer loan competitors; the External Credit Reporting - Balance (also known as External Balance) may include, but are not limited to: bank balance, consumer finance balance (i.e., consumer finance balance), and the balance of interest-only repayment; the External Credit Reporting - New Loan Invoices (also known as External Loan Invoices) may include, but are not limited to: the number of personal business loan loan invoices, the number of personal consumer loan loan invoices, and so on. And the number of commercial bank loan receipts, etc.; internal loan receipts (also known as internal loan receipts) may include, but are not limited to: the number of times credit is used in the current month, the current maximum remaining period, and the overdue amount in the current month; internal tracking points (also known as internal tracking points) may include, but are not limited to: the number of clicks on the product homepage, the number of times the Satisfactory Loan credit page is displayed, and the number of visits to the trial calculation page; internal strategy-operation (also known as internal strategy operation) may include, but are not limited to: the number of times coupons are issued in the current month, the coupon discount rate in the current month, and the number of times SMS messages are sent in the current month; internal strategy-risk control (also known as internal strategy risk control) may include, but are not limited to: the number of times credit limits are increased in the current month, the price change range in the current month, and the number of times temporary credit limit methods are used in the current month; this invention does not limit these.

[0037] In summary, the embodiments of the present invention can classify multiple basic variables, thereby making the classification clear and easy to manage, so that data users (such as managers or model developers) can use it easily and with strong interpretability; it can be seen that the embodiments of the present invention can ensure that feature classification is more systematic through the designed feature generation framework.

[0038] In this embodiment of the invention, the target time series may include 12 sequence points, 24 sequence points, etc.; the invention does not limit this. It should be understood that the target time series may refer to a time series within a target time range. The target time range may include multiple time sub-ranges. The sequence points in the target time series correspond one-to-one with the time sub-ranges in the target time range, and the characteristic of any basic variable at any sequence point can be: the value of any basic variable within the time sub-range corresponding to any sequence point. Accordingly, multiple time sub-ranges may be 12 consecutive months (in which case one time sub-range is one month), 24 consecutive months (in which case one time sub-range is one month), or multiple quarters (in which case one time sub-range is one quarter), etc.; the invention does not limit this. For example, assuming that multiple time sub-ranges include 12 consecutive months (such as the 12 months of last year), taking the basic variable "internal actual loan amount" as an example, the initial time series characteristic of the internal-actual loan amount may include the values ​​of 12 consecutive months, that is, it may include the value of the internal actual loan amount of each month in the 12 consecutive months.

[0039] In this embodiment of the invention, the methods for obtaining the initial temporal characteristics of each basic variable may include, but are not limited to, the following:

[0040] The first method of acquisition: The electronic device can obtain the feature download link of the feature set, which includes the initial time series features of each basic variable. In this case, the electronic device can download the feature set based on the feature download link to obtain the initial time series features of each basic variable.

[0041] The second acquisition method: Electronic devices can collect data based on variables to determine the initial time-series characteristics of each basic variable, and so on. Specifically, for any basic variable among multiple basic variables, a target time range is obtained. The target time range includes multiple time sub-ranges, and each time sub-range corresponds one-to-one with a sequence point in the target time series. Then, the variable collection data can be determined. The variable collection data includes variable data within each time sub-range of the multiple time sub-ranges. The variable data within a time sub-range includes the characteristics of each basic variable within the corresponding time sub-range. Furthermore, the characteristics of any basic variable within each time sub-range can be determined from the variable collection data, and the initial time-series characteristics of any basic variable are generated according to the target time series using the characteristics of any basic variable within each time sub-range.

[0042] The target time range can refer to the past year, the past month, etc., and this invention does not limit this. Optionally, a time sub-range can refer to a month, a quarter, etc., and this invention does not limit this. Optionally, the electronic device can also determine a time division rule and divide the target time range according to the time division rule to obtain multiple time sub-ranges, so that the target time range includes multiple time sub-ranges. Optionally, the time division rule can refer to division by month, division by quarter, etc., and this invention does not limit this. For example, assuming the target time range refers to the past year and the time division rule refers to division by month, then the electronic device can divide the target time range into 12 time sub-ranges, and each time sub-range is one month. At this time, the target time series can include 12 sequence points. That is, the initial time series features of a basic variable can include the features at each of the 12 sequence points, that is, it can include the values ​​(i.e., features) of the corresponding basic variable in each of the 12 time sub-ranges.

[0043] S102, determine at least one derived variable, and calculate the initial time series characteristics of each derived variable in the at least one derived variable based on the initial time series characteristics of each basic variable.

[0044] In embodiments of the present invention, at least one derived variable may include, but is not limited to, the total number of credit transactions (internal and external), the ratio of internal to external prices, the monthly credit limit, and the difference in credit limits between internal and external users, etc.; the present invention does not limit this. Optionally, at least one derived variable may include derived variables under addition rules, derived variables under division rules, derived variables under subtraction rules, derived variables under multiplication rules, and derived variables under extraction rules, etc. Specifically, the initial time-series characteristics of the derived variables under addition rules are obtained through addition operations, the initial time-series characteristics of the derived variables under subtraction rules are obtained through subtraction operations, the initial time-series characteristics of the derived variables under multiplication rules are obtained through multiplication operations, the initial time-series characteristics of the derived variables under division rules are obtained through division operations, and the initial time-series characteristics of the derived variables under extraction rules are obtained through extraction operations. Optionally, the extraction operation can refer to extracting features that meet the extraction conditions and setting features that do not meet the extraction conditions to 0; or, the extraction operation can refer to setting features that meet the extraction conditions to preset values ​​and setting features that do not meet the extraction conditions to 0; or, the extraction operation can refer to maximum value extraction operation or minimum value extraction operation, etc.; this invention does not limit this. Optionally, the preset value can be set according to experience or according to actual needs; this invention does not limit this. It should be understood that when the extraction operation refers to maximum value extraction operation, the maximum value can be extracted as the value of the derived variable; when the extraction operation refers to minimum value extraction operation, the minimum value can be extracted as the value of the derived variable.

[0045] S103, based on the initial time-series characteristics of each basic variable and each derived variable, generate the target time-series characteristics of each basic variable and each derived variable respectively, so as to generate the data to be analyzed. The data to be analyzed supports the determination of the model input data of the target model, so as to predict the model input data through the target model.

[0046] The data to be analyzed may include the target time-series features of each basic variable and the target time-series features of each derived variable. Optionally, the data to be analyzed can also be further extended, such as by extending the initial time-series features of multiple basic variables to obtain the initial time-series features of other derived variables besides at least one derived variable, thereby obtaining the target time-series features of other derived variables, which can then be added to the data to be analyzed, and so on.

[0047] It should be noted that this invention designs a general feature extraction framework (i.e., feature generation framework). For any variable among multiple basic variables and at least one derived variable, the initial time-series features of any variable can be extracted using the feature extraction framework, thereby obtaining the target time-series features of any variable. Optionally, the feature dimension of a target time-series feature can be 500 dimensions, 600 dimensions, etc., and this invention does not limit this; when more than 500 dimensions of features are extracted for each variable, the total number of features generated in the data to be analyzed can exceed 100,000 dimensions.

[0048] It should be understood that, in the context of object retention, the aforementioned target model is an object retention model (also known as a retention analysis model); optionally, the target model can be a time series model. In this embodiment of the invention, the electronic device can generate target time series features of each basic variable under any object, and generate target time series features of each derived variable under any object, thereby obtaining the data to be analyzed under any object; that is, the electronic device can generate the data to be analyzed under each of multiple objects.

[0049] In this embodiment of the invention, after obtaining the initial time-series features of each of the multiple basic variables, at least one derived variable can be determined. Based on the initial time-series features of each basic variable, the initial time-series features of each derived variable are calculated. Each initial time-series feature includes the characteristics of the corresponding variable on the target time series. Furthermore, based on the initial time-series features of each basic variable and each derived variable, target time-series features of each basic variable and each derived variable can be generated respectively to generate data to be analyzed. The data to be analyzed supports the model input data for determining the target model, so as to predict the model input data through the target model. Therefore, this embodiment of the invention can generate data to be analyzed with relatively comprehensive feature coverage, thereby improving the accuracy of prediction results. In addition, the data to be analyzed has strong scalability and is easy to maintain.

[0050] Based on the above description, this embodiment of the invention also proposes a more specific data generation method. Accordingly, this data generation method can be executed by the aforementioned electronic device (terminal or server); or, the data generation method can be executed jointly by the terminal and the server. For ease of explanation, the following description will use the execution of this data generation method by an electronic device as an example; please refer to [link to previous text]. Figure 3 The data generation method may include the following steps S301-S305:

[0051] S301, obtain the initial time series features of each basic variable among multiple basic variables. An initial time series feature includes the features of the corresponding variable on the target time series.

[0052] S302, determine at least one derived variable, and calculate the initial time series characteristics of each derived variable in the at least one derived variable based on the initial time series characteristics of each basic variable.

[0053] Specifically, when calculating the initial time-series characteristics of each derived variable among at least one derived variable based on the initial time-series characteristics of each basic variable, for any derived variable among at least one derived variable, the electronic device can determine the initial time-series characteristics of each calculated variable among at least one calculated variable corresponding to any derived variable from the initial time-series characteristics of each basic variable. Each calculated variable is a variable among multiple basic variables used to calculate the initial time-series characteristics of any derived variable. Then, the initial time-series characteristics of each calculated variable can be used to calculate the initial time-series characteristics of any derived variable.

[0054] Furthermore, when calculating the initial time-series characteristics of any derived variable using the initial time-series characteristics of each computational variable, the electronic device can determine the calculation rule corresponding to any derived variable. The calculation rule can include any of the following: addition rule, division rule, subtraction rule, multiplication rule, and extraction rule; and using the initial time-series characteristics of each computational variable, calculate the initial time-series characteristics of any derived variable according to the calculation rule. It should be understood that any two initial time-series characteristics have the same dimension; that is, the dimension of the initial time-series characteristic of a derived variable is the same as the dimension of the initial time-series characteristic of any basic variable.

[0055] Optionally, the derived variables under the addition operation rule may include, but are not limited to: total internal and external balance (internal balance + external balance, i.e., at least one calculation variable includes internal balance and external balance), total number of credit uses (internal credit use + external credit use), etc. Here, credit use can refer to borrowing or using a credit line.

[0056] For example, such as Figure 4 As shown, assuming any derived variable is the total number of internal and external communications, the calculation rule corresponding to any derived variable is the addition operation rule. At least one calculation variable corresponding to any derived variable includes the number of internal communications and the number of external communications. Then, the electronic device can use the initial time-series characteristics of the number of internal communications (such as the value of the number of internal communications in each month of 24 consecutive months) and the initial time-series characteristics of the number of external communications (such as the value of the number of external communications in each month of 24 consecutive months) to calculate the initial time-series characteristics of the total number of internal and external communications according to the addition operation rule. That is to say, the initial time-series characteristics of the number of internal communications and the initial time-series characteristics of the number of external communications can be added to obtain the initial time-series characteristics of the total number of internal and external communications.

[0057] Optionally, the derived variables under the subtraction operation rules may include, but are not limited to: internal and external price difference (internal price - external price, i.e., at least one calculation variable includes internal price and external price), internal and external credit limit difference (internal credit limit - external credit limit), internal and external credit limit utilization rate difference (internal credit limit utilization rate - external credit limit utilization rate), internal and external balance difference (internal balance - external balance), internal and external credit limit difference when no credit is used (~Bool(internal credit usage times) * (internal credit limit - external credit limit)), internal and external credit limit difference when credit is used (Bool(internal credit usage times) * (internal credit limit - external credit limit)), internal and external price difference when no credit is used (~Bool(internal credit usage times) * (internal price - external price)), internal and external price difference when credit is used (Bool(internal credit usage times) * (internal price - external price)), etc. Here, Bool is the definition symbol for a logical variable, and Bool can take the value of true or false, with 0 being false and 1 being true; correspondingly, ~Bool is the result of inverting Bool. For example, when the value of the internal message usage count is greater than 0 (i.e., message usage), the value of Bool(internal message usage count) can be 1, and the value of ~Bool(internal message usage count) can be 0; when the value of the internal message usage count is greater than 0 (i.e., message not used), the value of Bool(internal message usage count) can be 0, and the value of ~Bool(internal message usage count) can be 1. For example, assuming the initial time series characteristics of the internal message usage count are (0, 1, 3, 0, 6), then Bool(internal message usage count) can be (0, 1, 1, 0, 1), and ~Bool(internal message usage count) can be (1, 0, 0, 1, 0). Optionally, Bool(internal message usage count) can also be referred to as whether the message was used internally; that is, performing a Bool operation on the basic variable "internal message usage count" can be transformed into the basic variable "whether the message was used internally".

[0058] For example, suppose any derived variable is the difference between internal and external credit limits. In this case, the calculation rule corresponding to any derived variable is the subtraction operation rule, and at least one calculation variable corresponding to any derived variable can include internal credit limit and external credit limit (i.e., competitor credit limit). Then, the electronic device can use the initial time-series characteristics of the internal credit limit (such as the internal credit limit value of each month in 24 consecutive months) and the initial time-series characteristics of the external credit limit (such as the external credit limit value of each month in 24 consecutive months) to calculate the initial time-series characteristics of the difference between internal and external credit limits according to the subtraction operation rule. That is, the initial time-series characteristics of the internal credit limit and the initial time-series characteristics of the external credit limit can be subtracted to obtain the initial time-series characteristics of the difference between internal and external credit limits.

[0059] Optionally, derived variables under the multiplication rule may include, but are not limited to: competitor price increase - whether the product has been used within six months (competitor price increase * whether the product has been used internally within six months, where at least one calculation variable may include both competitor price increase and whether the product has been used internally within six months), competitor credit limit decrease - whether the product has been used within six months (competitor credit limit decrease * whether the product has been used internally within six months), competitor price decrease - whether the product has been prepaid within six months (competitor price decrease * whether the product has been prepaid within six months), competitor credit limit increase - whether the product has been prepaid within six months (competitor credit limit increase * whether the product has been prepaid within six months), competitor price increase - whether the product has been used within three months (competitor price increase * whether the product has been used internally within three months), competitor... Credit limit reduction - whether the product has been used within the past three months (competitors' credit limit reduction * internal credit usage within the past three months); Competitor price reduction - whether the loan has been prepaid within the past three months (competitors' price reduction * internal loan usage within the past three months); Competitor credit limit increase - whether the loan has been prepaid within the past three months (competitors' credit limit increase * internal loan usage within the past three months); Competitor price increase - whether the product has been used within the past month (competitors' price increase * internal loan usage within the past month); Competitor credit limit reduction - whether the product has been used within the past month (competitors' credit limit reduction * internal loan usage within the past month); Competitor price reduction - whether the loan has been prepaid within the past month (competitors' price reduction * internal loan usage within the past month); Competitor credit limit increase - whether the loan has been prepaid within the past month (competitors' credit limit increase) *Whether to prepay within one month), whether to use credit within 3 months after credit enhancement (Bool(submitted credit enhancement information) * Bool(number of credit uses in the current month)), whether to lower the price within 3 months after credit enhancement (Bool(submitted credit enhancement information) * Bool(number of price reductions in the current month)), whether to increase the credit limit within 3 months after credit enhancement (Bool(submitted credit enhancement information) * Bool(number of credit limit increases in the current month)), whether to add external loan items in the current month when accessing the trial page (Bool(number of times the trial page is accessed) * number of external loan items in the current month), whether to settle the loan early within 6 months after internal price reduction (Bool(number of times the price reduction is applied in the current month * whether to settle the loan early within 6 months), whether to settle the loan early within 3 months after internal price reduction (Bool(number of times the price reduction is applied in the current month)). *Did you settle the loan early within 3 months?), Did you settle the loan early in the same month after the internal price reduction (Bool(number of price reductions in the current month) * whether you settled the loan early in the current month?), Did you repay the loan early within 6 months after the internal price reduction (Bool(number of price reductions in the current month) * whether you repaid the loan early within 6 months?), Did you repay the loan early within 3 months after the internal price reduction (Bool(number of price reductions in the current month) * whether you repaid the loan early within 3 months?), Did you repay the loan early in the same month after the internal price reduction (Bool(number of price reductions in the current month) * whether you repaid the loan early in the current month?), Did you use credit within 6 months after the internal price reduction (Bool(number of price reductions in the current month) * whether you used credit within 6 months?), Did you use credit within 3 months after the internal price reduction (Bool(number of price reductions in the current month) * whether you used credit within 3 months?)Whether credit was used in the month following an internal price reduction (Bool(number of price reductions in the current month) * whether credit was used in the current month), whether credit was used within 6 months following an internal credit limit increase (Bool(number of credit limit increases in the current month) * whether credit was used within 6 months), whether credit was used within 3 months following an internal credit limit increase (Bool(number of credit limit increases in the current month) * whether credit was used within 3 months), whether credit was used in the month following an internal credit limit increase (Bool(number of credit limit increases in the current month) * whether credit was used in the current month), etc.

[0060] Here, Bool(Number of Price Reductions in the Current Month) can refer to whether a price reduction occurred in the current month, Bool(Number of Trial Page Visits) can refer to whether the trial page was visited, Bool(Submission of Credit Enhancement Information) can refer to whether credit enhancement information was submitted, Bool(Number of Credit Limit Increases in the Current Month) can refer to whether a credit limit was increased in the current month, and so on. It should be noted that the symbol "_" can be used to represent "and / or"; for example, taking "Competitor Price Increase_Whether the Product Has Been Credited in the Last Six Months" as an example, when the value of "Competitor Price Increase_Whether the Product Has Been Credited in the Last Six Months" is 1, it can mean that the competitor's price increased and the product was credited within six months; when the value of "Competitor Price Increase_Whether the Product Has Been Credited in the Last Six Months" is 0, it can mean that the competitor's price did not increase and / or the product was not credited within six months.

[0061] For example, suppose any derived variable is "competitor price increase - whether the product uses credit within six months". In this case, the calculation rule corresponding to any derived variable is the subtraction operation rule, and at least one calculation variable corresponding to any derived variable can include whether the competitor has raised prices and whether the product uses credit within six months. Then, electronic devices can use the initial time-series characteristics of whether the competitor has raised prices and whether the product uses credit within six months, and calculate the initial time-series characteristics of "competitor price increase - whether the product uses credit within six months" according to the multiplication operation rule. That is, the initial time-series characteristics of whether the competitor has raised prices and whether the product uses credit within six months can be multiplied to obtain the initial time-series characteristics of "competitor price increase - whether the product uses credit within six months". For example, suppose the initial time series characteristic of whether a competitor raises its price is (1, 0, 1, 1, 0), where 1 indicates that the competitor raised its price in the current month, and 0 indicates that the competitor did not raise its price in the current month; and suppose the initial time series characteristic of whether credit was used internally within six months is (0, 1, 1, 0, 1), where 1 indicates that credit was used internally within six months, and 0 indicates that credit was not used internally within six months. Then the initial time series characteristic of whether a competitor raises its price and whether credit was used within six months is (0, 0, 1, 0, 0).

[0062] Optionally, the derived variables under the division operation rules may include, but are not limited to: internal and external price ratio (internal price / external price), internal and external quota ratio (internal quota / external quota), internal and external quota utilization rate ratio (internal quota utilization rate / external quota utilization rate), internal and external balance ratio (internal balance / external balance), etc.

[0063] For example, suppose any derived variable is the ratio of internal to external prices (i.e., interest rates). In this case, the calculation rule corresponding to any derived variable is the division operation rule, and at least one calculation variable corresponding to any derived variable can include internal price and external price (i.e., competitor price). Then, electronic devices can use the initial time-series characteristics of internal price (such as the internal price value of each month in 24 consecutive months) and the initial time-series characteristics of external price (such as the external price value of each month in 24 consecutive months) to calculate the initial time-series characteristics of the ratio of internal to external prices according to the division operation rule. That is, the initial time-series characteristics of internal price and the initial time-series characteristics of external price can be divided to obtain the initial time-series characteristics of the ratio of internal to external prices.

[0064] Optionally, the derived variables under the extraction operation rules may include, but are not limited to: credit limit per month (f(internal credit usage times, internal credit limit)), maximum internal and external price (Max(internal price, external price)), minimum internal and external price (Min(internal price, external price)), maximum internal and external credit limit (Max(internal credit limit, external credit limit)), minimum internal and external credit limit (Min(internal credit limit, external credit limit)), maximum internal and external credit limit utilization rate (maximum internal and external credit limit utilization rate), minimum internal and external credit limit utilization rate (Min(internal credit limit utilization rate, external credit limit utilization rate)), maximum internal and external balance (Max(internal balance, external balance)), minimum internal and external balance (Min(internal balance, external balance)), etc. Where f can represent extraction based on the extraction condition "internal message usage count is greater than the preset message usage count threshold or internal message usage", Max can represent extraction based on the extraction condition "maximum value" (i.e., used to extract the maximum value, in which case the extraction operation refers to the maximum value extraction operation), and Min can represent extraction based on the extraction condition "minimum value" (i.e., used to extract the minimum value, in which case the extraction operation refers to the minimum value extraction operation).

[0065] For example, assuming any derived variable is the credit limit for a given month, the calculation rule corresponding to this derived variable is an extraction operation rule, and at least one calculation variable corresponding to this derived variable can include internal credit usage counts and internal credit limits. Furthermore, assuming the extraction condition for the credit limit for a given month is that the value of the internal credit usage counts is greater than a preset credit usage count threshold, then when the electronic device uses the initial temporal characteristics of the internal credit usage counts and the initial temporal characteristics of the internal credit limits to calculate the credit limit for a given month according to the extraction operation rule, it can extract the internal credit limit value that meets the extraction condition (i.e., the value of the internal credit usage counts is greater than the preset credit usage count threshold), thus obtaining the initial temporal characteristics of the credit limit for a given month. Optionally, the preset credit usage count threshold can be 0 (in which case the internal credit limit value within the time sub-range of credit usage is extracted), or it can be 1, etc., and this invention does not limit this. For example, assuming the initial time-series characteristics of internal credit usage frequency are (0, 1, 3, 0, 5), the initial time-series characteristics of internal credit limit are (10000, 20000, 20000, 10000, 10000), and the preset credit usage frequency threshold is 0, then the initial time-series characteristics of the credit limit for a given month can be (0, 20000, 20000, 0, 10000). Optionally, when any derived variable is the credit limit for a given month, at least one calculated variable corresponding to any derived variable can also include whether internal credit is used and the internal credit limit. The initial time-series characteristics of whether internal credit is used can include a credit usage identifier (e.g., 1) and an unused credit identifier (e.g., 0). In this case, the extraction condition can refer to the feature of whether internal credit is used being the credit usage identifier. Then, the electronic device can extract the internal credit limit value with the feature of whether internal credit is used being the credit usage identifier from the initial time-series characteristics of the internal credit limit to obtain the initial time-series characteristics of the internal credit limit.

[0066] For example, suppose any derived variable is the internal or external maximum price. In this case, the calculation rule corresponding to any derived variable is the extraction operation rule. Here, the extraction operation refers to the maximum value extraction operation. And at least one calculation variable corresponding to any derived variable can include the internal price and the external price. Then, the electronic device can perform the maximum value extraction operation on the initial time series characteristics of the internal price and the initial time series characteristics of the external price to obtain the initial time series characteristics of the internal and external maximum prices. For example, suppose the initial time series characteristics of the internal price are (0.3, 0.5, 0.1, 0.2, 0.2) and the initial time series characteristics of the external price are (0.2, 0.3, 0.5, 0.2, 0.3). Then the initial time series characteristics of the internal and external maximum prices are (0.3, 0.5, 0.5, 0.2, 0.3).

[0067] In summary, at least one derived variable can have at least 50 derived variables.

[0068] S303, for any one of multiple basic variables and at least one derived variable, determine at least one target domain, wherein the at least one target domain includes at least one of the following: time domain, spectral domain, and statistical domain.

[0069] Optionally, time-domain-based time series features may include, but are not limited to: autocorrelation, mean differences, mean absolute differences, median differences, median absolute differences, sum of absolute differences, entropy, peak to peak distance, number of maximum peaks, number of minimum peaks, etc.; this invention does not limit these features.

[0070] Optionally, the time-series features based on the spectral domain may include, but are not limited to: Fast Fourier Transform, FFT Mean Coefficient, Wavelet Transform, Wavelet Absolute Mean, Wavelet Standard Deviation, Wavelet Variance, Spectral Distance, Spectral Fundamental Frequency, Spectral Maximum Frequency, Spectral Median Frequency, Spectral Maximum Peaks, etc.; this invention does not limit these. In embodiments of this invention, the spectral domain may also be referred to as the frequency domain.

[0071] Optionally, the time-series characteristics based on the statistical domain may include, but are not limited to: maximum, minimum, mean, median, skewness, kurtosis, interquartile range, mean absolute deviation, median absolute deviation, root mean square, standard deviation, variance, empirical distribution function percentile count, empirical cumulative distribution function slope, etc.; this invention does not limit these.

[0072] S304, based on the initial time-series characteristics of any variable, calculate at least one feature of any variable in each target domain in at least one target domain.

[0073] Among them, features under a target domain can also be called time-series features based on the corresponding target domain.

[0074] In one implementation, at least one target domain may include a spectral domain, and the spectral domain may include at least one transformation method. Based on this, when calculating at least one feature of any variable in each target domain based on the initial temporal characteristics of any variable, the electronic device may traverse each transformation method among the at least one transformation method and take the currently traversed transformation method as the current transformation method. Then, according to the current transformation method, the initial temporal characteristics of any variable may be subjected to spectral domain transformation processing to obtain the spectral domain transformation result of any variable under the current transformation method. At least one spectral domain variable corresponding to the current transformation method is determined, and based on the spectral domain transformation result, the variable information of each spectral domain variable in the at least one spectral domain variable is calculated, and the variable information of each spectral domain variable is added to at least one feature of any variable in the spectral domain. After traversing each transformation method among the at least one transformation method, at least one feature of any variable in the spectral domain is obtained.

[0075] Optionally, the above-mentioned at least one transformation method may include, but is not limited to: Fast Fourier Transform, Wavelet Transform, and Discrete-Time Fourier Transform, etc.; the present invention does not limit this. Optionally, at least one spectral domain variable corresponding to a transformation method may include, but is not limited to: standard deviation variable, variance variable, absolute mean variable, maximum frequency variable, fundamental frequency variable, intermediate frequency variable, spectral distance variable, and maximum peak value variable, etc.; the present invention does not limit this.

[0076] For example, suppose the current transformation method is wavelet transform, and the at least one spectral domain variable corresponding to the current transformation method includes: wavelet absolute value variable, wavelet standard deviation variable, wavelet variance variable, and wavelet maximum frequency variable. In this case, the electronic device can perform wavelet transform on the initial time-series features of any variable to achieve spectral domain transformation processing on the initial time-series features of any variable, thereby obtaining the wavelet transform result of any variable under wavelet transform. Further, based on the wavelet transform result, the electronic device can calculate the variable information of the wavelet absolute value variable (i.e., wavelet absolute value), the variable information of the wavelet standard deviation variable (i.e., wavelet standard deviation), the variable information of the wavelet variance variable (i.e., wavelet variance), and the variable information of the wavelet maximum frequency variable (i.e., wavelet maximum frequency), and add them to at least one feature of any variable in the spectral domain. In this case, at least one feature of any variable in the spectral domain may include wavelet absolute value, wavelet standard deviation, wavelet variance, and wavelet maximum frequency, etc.

[0077] In another implementation, when at least one target domain includes a time domain, for any variable among multiple basic variables and at least one derived variable, the electronic device can determine at least one time-domain variable in the time domain, and calculate the variable information of each time-domain variable in the at least one time-domain variable based on the initial time-series characteristics of any variable, to obtain at least one feature of any variable in the time domain, so that at least one feature of any variable in the time domain includes the variable information of each time-domain variable; wherein, at least one time-domain variable may include, but is not limited to: autocorrelation variable, difference mean variable, difference absolute value mean variable, and minimum peak count variable, etc.; the present invention does not limit this.

[0078] For example, assuming that at least one time-domain variable includes a difference mean variable, a difference absolute mean variable, and a minimum peak count variable, then the electronic device can calculate the variable information of the difference mean variable (i.e., difference mean), the variable information of the difference absolute mean variable (i.e., difference absolute mean), and the variable information of the minimum peak count variable (i.e., minimum peak count) based on the initial time-series characteristics of any variable, thereby obtaining at least one feature of any variable in the time domain. At this time, at least one feature of any variable in the time domain may include the difference mean, the difference absolute mean, and the minimum peak count.

[0079] In another implementation, when at least one target domain includes a statistical domain, for any variable among multiple basic variables and at least one derived variable, the electronic device can determine at least one statistical domain variable under the statistical domain, and calculate the variable information of each statistical domain variable among the at least one statistical domain variable based on the initial time-series characteristics of any variable, to obtain at least one feature of any variable under the statistical domain, so that at least one feature of any variable under the statistical domain includes the variable information of each statistical domain variable; wherein, at least one statistical domain variable may include, but is not limited to: maximum value variable, minimum value variable, mean variable, median variable, and interquartile range variable, etc.; the present invention does not limit this.

[0080] For example, assuming that at least one statistical domain variable includes a mean variable, a median variable, and an interquartile range variable, the electronic device can calculate the variable information of the mean variable (i.e., mean), the variable information of the median variable (i.e., median), and the variable information of the interquartile range variable (i.e., interquartile range) based on the initial time-series characteristics of any variable, thereby obtaining at least one feature of any variable in the statistical domain, and at this time, at least one feature of any variable in the statistical domain may include the mean, the median, and the interquartile range.

[0081] S305, based on the initial time-series features of any variable and at least one feature of any variable in each target domain, generate the target time-series features of any variable to generate the data to be analyzed. The data to be analyzed supports the model input data for determining the target model, so as to predict the model input data through the target model.

[0082] It should be understood that the target time-series feature of any variable may include the initial time-series feature of the variable and at least one feature of the variable in each target domain. Furthermore, the target time-series feature of any variable may be an M-dimensional vector, where M is a positive integer. The value of M is equal to the sum of the dimension of the initial time-series feature of the variable and the dimension of each feature of the variable in each target domain. For example, assuming that at least one target domain includes the time domain, and at least one feature of any variable in the time domain includes a first time-domain feature and a second time-domain feature, then the value of M is equal to the sum of the dimension of the initial time-series feature of the variable, the dimension of the first time-domain feature, and the dimension of the second time-domain feature.

[0083] For example, such as Figure 5 As shown, when at least one target domain can include the time domain, the spectral domain, and the statistical domain, the initial time series features of any variable can be extracted as the target time series features of any variable. At this time, the target time series features of any variable can include the initial time series features of any variable, at least one feature of any variable in the time domain, at least one feature of any variable in the spectral domain, and at least one feature of any variable in the statistical domain, etc.

[0084] This invention, after obtaining the initial time-series features of each of the multiple basic variables and determining at least one derived variable, calculates the initial time-series features of each of the at least one derived variable based on the initial time-series features of each of the basic variables. Further, for any variable among the multiple basic variables and at least one derived variable, at least one target domain is determined, including at least one of the following: time domain, spectral domain, and statistical domain. Based on the initial time-series features of any variable, at least one feature of any variable in each of the at least one target domain is calculated. Thus, based on the initial time-series features of any variable and at least one feature of any variable in each target domain, a target time-series feature of any variable is generated to generate data to be analyzed. This data supports the model input data for determining the target model, enabling prediction of the model input data using the target model. This invention can abstract the complex and chaotic feature generation process into feature engineering centered on vectors (i.e., initial temporal features), thereby extracting the target temporal features of each basic variable and each derived variable. This makes the subsequent application of the target model more user-friendly and efficient. Through a unified feature application framework, the target model can be efficiently iterated on the basis of existing features, thereby improving model performance and further enhancing the accuracy of prediction results.

[0085] Based on the description of the relevant embodiments of the above data generation method, this invention also proposes a data generation apparatus, which can be a computer program (including program code) running in an electronic device; such as Figure 6 As shown, the data generation apparatus may include an acquisition unit 601 and a processing unit 602. The data generation apparatus can perform... Figure 1 or Figure 3 The data generation method shown, i.e., the data generation device can operate the above-mentioned unit:

[0086] The acquisition unit 601 is used to acquire the initial time series features of each basic variable among multiple basic variables. An initial time series feature includes the features of the corresponding variable on the target time series.

[0087] Processing unit 602 is configured to determine at least one derived variable and, based on the initial time-series characteristics of each of the at least one derived variable, calculate the initial time-series characteristics of each derived variable among the at least one derived variable.

[0088] The processing unit 602 is further configured to generate target time-series features of each basic variable and target time-series features of each derived variable based on the initial time-series features of each basic variable and the initial time-series features of each derived variable, respectively, to generate data to be analyzed. The data to be analyzed supports the determination of model input data for a target model, so as to predict the model input data through the target model.

[0089] In one implementation, when the processing unit 602 calculates the initial time-series characteristics of each derived variable among the at least one derived variable based on the initial time-series characteristics of each of the basic variables, it may specifically be used to:

[0090] For any one of the at least one derived variables, the initial time-series characteristics of each calculated variable in at least one calculated variable corresponding to the derived variable are determined from the initial time-series characteristics of each basic variable. Each calculated variable is a variable among the plurality of basic variables used to calculate the initial time-series characteristics of the derived variable.

[0091] The initial time series characteristics of any derived variable are calculated using the initial time series characteristics of each of the calculated variables.

[0092] In another embodiment, when processing unit 602 calculates the initial time-series characteristics of any derived variable using the initial time-series characteristics of each of the calculated variables, it may specifically be used to:

[0093] Determine the calculation rule corresponding to any of the derived variables, wherein the calculation rule includes any one of the following: addition operation rule, division operation rule, subtraction operation rule, multiplication operation rule, and extraction operation rule;

[0094] Using the initial time-series characteristics of each of the calculated variables, the initial time-series characteristics of any derived variable are calculated according to the calculation rules.

[0095] In another embodiment, when the processing unit 602 generates the target time-series features of each basic variable and the target time-series features of each derived variable based on the initial time-series features of each basic variable and the initial time-series features of each derived variable, it can be specifically used to:

[0096] For any one of the plurality of basic variables and the at least one derived variable, at least one target domain is determined, the at least one target domain including at least one of the following: time domain, spectral domain and statistical domain;

[0097] Based on the initial temporal characteristics of any variable, calculate at least one feature of any variable in each target domain in the at least one target domain;

[0098] Based on the initial temporal features of any variable and at least one feature of any variable in each target domain, the target temporal features of any variable are generated.

[0099] In another embodiment, at least one target domain includes the spectral domain, and the spectral domain includes at least one transformation mode; when the processing unit 602 calculates at least one feature of the any variable under each target domain in the at least one target domain based on the initial temporal characteristics of the any variable, it can be specifically used for:

[0100] Iterate through each of the at least one transformation method and take the currently traversed transformation method as the current transformation method;

[0101] According to the current transformation method, the initial time-series characteristics of any variable are subjected to spectral domain transformation processing to obtain the spectral domain transformation result of any variable under the current transformation method;

[0102] Determine at least one spectral domain variable corresponding to the current transformation method, and based on the spectral domain transformation result, calculate the variable information of each spectral domain variable in the at least one spectral domain variable, and add the variable information of each spectral domain variable to at least one feature of any variable in the spectral domain;

[0103] After traversing all the transformation methods in the at least one transformation method, at least one feature of the variable in the spectral domain is obtained.

[0104] In another implementation, when acquiring the initial time-series characteristics of each basic variable among multiple basic variables, the acquisition unit 601 can be specifically used for:

[0105] For any one of the multiple basic variables, a target time range is obtained. The target time range includes multiple time sub-ranges, and each time sub-range corresponds one-to-one with a sequence point in the target time series.

[0106] The variable collection data is determined, which includes variable data in each time sub-range of the plurality of time sub-ranges. The variable data in a time sub-range includes the characteristics of each basic variable in the corresponding time sub-range.

[0107] From the data collected from the variables, the characteristics of any one basic variable within each time sub-range are determined, and the initial time series characteristics of any one basic variable are generated according to the target time series using the characteristics of any one basic variable within each time sub-range.

[0108] In another implementation, the target model is an object retention model, and the plurality of basic variables are variables under the object retention scenario; the plurality of basic variables include at least one variable under each of the plurality of target categories.

[0109] According to one embodiment of the present invention, Figure 1 or Figure 3 Each step involved in the method shown can be derived from... Figure 6 This is performed by the individual units in the data generation apparatus shown. For example, Figure 1 Step S101 shown can be performed by Figure 6 The acquisition unit 601 shown is executed, and steps S102 and S103 can both be performed by... Figure 6 The processing unit 602 shown executes this. For example, Figure 3 Step S301 shown can be performed by Figure 6 The acquisition unit 601 shown is executed, and steps S302-S305 can all be performed by... Figure 6 The processing unit 602 shown executes, etc.

[0110] According to another embodiment of the present invention, Figure 6 Each unit in the data generation apparatus shown can be individually or entirely merged into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effects of the embodiments of the present invention. The above-mentioned units are based on logical function division. In practical applications, the function of one unit can also be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of the present invention, any data generation apparatus may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0111] According to another embodiment of the present invention, the following can be performed by running on a general-purpose electronic device, such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM). Figure 1 or Figure 3 The computer program (including program code) for each step involved in the corresponding method shown, to construct such... Figure 6 The data generation apparatus shown herein, and the data generation method for implementing embodiments of the present invention, are described. The computer program may be recorded on, for example, a computer storage medium, loaded onto the aforementioned electronic device via the computer storage medium, and run therein.

[0112] In this embodiment of the invention, after obtaining the initial time-series features of each of the multiple basic variables, at least one derived variable can be determined. Based on the initial time-series features of each basic variable, the initial time-series features of each derived variable are calculated. Each initial time-series feature includes the characteristics of the corresponding variable on the target time series. Furthermore, based on the initial time-series features of each basic variable and each derived variable, target time-series features of each basic variable and each derived variable can be generated respectively to generate data to be analyzed. The data to be analyzed supports the model input data for determining the target model, so as to predict the model input data through the target model. Therefore, this embodiment of the invention can generate data to be analyzed with relatively comprehensive feature coverage, thereby improving the accuracy of prediction results. In addition, the data to be analyzed has strong scalability and is easy to maintain.

[0113] Based on the description of the method and apparatus embodiments above, an exemplary embodiment of the present invention also provides an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor. The memory stores a computer program executable by the at least one processor, which, when executed by the at least one processor, causes the electronic device to perform the method according to an embodiment of the present invention.

[0114] An exemplary embodiment of the present invention also provides a non-transitory computer-readable storage medium storing a computer program, wherein the computer program, when executed by a computer's processor, is used to cause the computer to perform a method according to an embodiment of the present invention.

[0115] An exemplary embodiment of the present invention also provides a computer program product, including a computer program, wherein, when executed by a computer's processor, the computer program is used to cause the computer to perform a method according to an embodiment of the present invention.

[0116] refer to Figure 7 The present invention will now be described in the form of a structural block diagram of an electronic device 700 that can serve as a server or client of the present invention, which is an example of a hardware device that can be applied to various aspects of the present invention. The electronic device is intended to represent various forms of digital electronic computer devices, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0117] like Figure 7 As shown, the electronic device 700 includes a computing unit 701, which can perform various appropriate actions and processes based on a computer program stored in a read-only memory (ROM) 702 or a computer program loaded from a storage unit 708 into a random access memory (RAM) 703. The RAM 703 may also store various programs and data required for the operation of the device 700. The computing unit 701, ROM 702, and RAM 703 are interconnected via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0118] Multiple components in electronic device 700 are connected to I / O interface 705, including: input unit 706, output unit 707, storage unit 708, and communication unit 709. Input unit 706 can be any type of device capable of inputting information to electronic device 700. Input unit 706 can receive input digital or character information and generate key signal inputs related to user settings and / or function control of electronic device. Output unit 707 can be any type of device capable of presenting information and may include, but is not limited to, a display, speaker, video / audio output terminal, vibrator, and / or printer. Storage unit 708 may include, but is not limited to, disk and optical disk. Communication unit 709 allows electronic device 700 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks, and may include, but is not limited to, modems, network cards, infrared communication devices, wireless communication transceivers, and / or chipsets, such as Bluetooth™ devices, WiFi devices, WiMax devices, cellular communication devices, and / or the like.

[0119] The computing unit 701 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 701 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 701 performs the various methods and processes described above. For example, in some embodiments, the data generation method may be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 708. In some embodiments, part or all of the computer program may be loaded and / or installed on the electronic device 700 via ROM 702 and / or communication unit 709. In some embodiments, the computing unit 701 may be configured to perform the data generation method by any other suitable means (e.g., by means of firmware).

[0120] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0121] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0122] As used herein, the terms "machine-readable medium" and "computer-readable medium" refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) for providing machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term "machine-readable signal" refers to any signal for providing machine instructions and / or data to a programmable processor.

[0123] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0124] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as a data server), or computing systems that include middleware components (e.g., an application server), or computing systems that include frontend components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with embodiments of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., a communication network). Examples of communication networks include local area networks (LANs), wide area networks (WANs), and the Internet.

[0125] Computer systems can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. Client-server relationships are created by computer programs running on the respective computers and having a client-server relationship with each other.

[0126] Furthermore, it should be understood that the above-disclosed embodiments are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, any equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A data generation method, characterized in that, include: Obtaining the initial time series features of each basic variable among multiple basic variables includes: for any basic variable among the multiple basic variables, obtaining a target time range, wherein the target time range includes multiple time sub-ranges, and the time sub-ranges among the multiple time sub-ranges correspond one-to-one with the sequence points in the target time series; The variable collection data is determined, which includes variable data in each time sub-range of the plurality of time sub-ranges. The variable data in a time sub-range includes the characteristics of each basic variable in the corresponding time sub-range. From the data collected from the variables, the characteristics of any basic variable in each time sub-range are determined, and the initial time series characteristics of any basic variable are generated according to the target time series using the characteristics of any basic variable in each time sub-range. An initial time series characteristic includes the characteristics of the corresponding variable in the target time series. The characteristic of any basic variable at any sequence point is: the value of any basic variable in the time sub-range corresponding to the sequence point. Identify at least one derived variable, and calculate the initial time-series characteristics of each derived variable among the at least one derived variable based on the initial time-series characteristics of each of the basic variables; Based on the initial time-series features of each basic variable and the initial time-series features of each derived variable, target time-series features of each basic variable and target time-series features of each derived variable are generated respectively to generate data to be analyzed. The data to be analyzed supports the determination of the model input data of the target model so as to predict the model input data through the target model. The target model is an object retention model, and the multiple basic variables are variables under the object retention scenario; the multiple basic variables include at least one variable under each of the multiple target categories, and the multiple target categories include: internal loan performance, other internal performance, internal strategy actions, competitor loan performance, competitor strategy actions, and other external behaviors.

2. The method according to claim 1, characterized in that, The step of calculating the initial time-series characteristics of each derived variable among the at least one derived variable based on the initial time-series characteristics of each of the basic variables includes: For any one of the at least one derived variables, the initial time-series characteristics of each calculated variable in at least one calculated variable corresponding to the derived variable are determined from the initial time-series characteristics of each basic variable. Each calculated variable is a variable among the plurality of basic variables used to calculate the initial time-series characteristics of the derived variable. The initial time series characteristics of any derived variable are calculated using the initial time series characteristics of each of the calculated variables.

3. The method according to claim 2, characterized in that, The step of using the initial time-series characteristics of each of the calculated variables to calculate the initial time-series characteristics of any derived variable includes: Determine the calculation rule corresponding to any of the derived variables, wherein the calculation rule includes any one of the following: addition operation rule, division operation rule, subtraction operation rule, multiplication operation rule, and extraction operation rule; Using the initial time-series characteristics of each of the calculated variables, the initial time-series characteristics of any derived variable are calculated according to the calculation rules.

4. The method according to any one of claims 1-3, characterized in that, The process of generating target time-series features for each basic variable and each derived variable based on their initial time-series features includes: For any one of the plurality of basic variables and the at least one derived variable, at least one target domain is determined, the at least one target domain including at least one of the following: time domain, spectral domain and statistical domain; Based on the initial temporal characteristics of any variable, calculate at least one feature of any variable in each target domain in the at least one target domain; Based on the initial temporal features of any variable and at least one feature of any variable in each target domain, the target temporal features of any variable are generated.

5. The method according to claim 4, characterized in that, The at least one target domain includes the spectral domain, and the spectral domain includes at least one transformation mode; the step of calculating at least one feature of any variable in each target domain based on the initial time-series features of any variable includes: Iterate through each of the at least one transformation method and take the currently traversed transformation method as the current transformation method; According to the current transformation method, the initial time-series characteristics of any variable are subjected to spectral domain transformation processing to obtain the spectral domain transformation result of any variable under the current transformation method; Determine at least one spectral domain variable corresponding to the current transformation method, and based on the spectral domain transformation result, calculate the variable information of each spectral domain variable in the at least one spectral domain variable, and add the variable information of each spectral domain variable to at least one feature of any variable in the spectral domain; After traversing all the transformation methods in the at least one transformation method, at least one feature of the variable in the spectral domain is obtained.

6. A data generation apparatus, characterized in that, The device includes: An acquisition unit is used to acquire the initial time-series features of each basic variable among multiple basic variables, including: acquiring a target time range for any basic variable among the multiple basic variables, the target time range including multiple time sub-ranges, and the time sub-ranges of the multiple time sub-ranges correspond one-to-one with the sequence points in the target time series; determining variable acquisition data, the variable acquisition data including variable data in each time sub-range of the multiple time sub-ranges, the variable data in one time sub-range including the features of each basic variable in the corresponding time sub-range; determining the features of any basic variable in each time sub-range from the variable acquisition data, and using the features of any basic variable in each time sub-range to generate the initial time-series feature of any basic variable according to the target time series; an initial time-series feature includes the feature of the corresponding variable in the target time series, and the feature of any basic variable at any sequence point is: the value of any basic variable in the time sub-range corresponding to the any sequence point; A processing unit is configured to determine at least one derived variable and, based on the initial time-series characteristics of each of the at least one derived variable, calculate the initial time-series characteristics of each derived variable among the at least one derived variable. The processing unit is further configured to generate target time-series features of each basic variable and target time-series features of each derived variable based on the initial time-series features of each basic variable and the initial time-series features of each derived variable, respectively, to generate data to be analyzed. The data to be analyzed supports the determination of model input data of the target model, so as to predict the model input data through the target model. The target model is an object retention model, and the multiple basic variables are variables under the object retention scenario; the multiple basic variables include at least one variable under each of the multiple target categories, and the multiple target categories include: internal loan performance, other internal performance, internal strategy actions, competitor loan performance, competitor strategy actions, and other external behaviors.

7. An electronic device, characterized in that, include: processor; as well as Stored program memory, The program includes instructions that, when executed by the processor, cause the processor to perform the method according to any one of claims 1-5.

8. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-5.

Images