CN111539532A - Model construction-oriented automatic feature derivation method - Google Patents
Model construction-oriented automatic feature derivation method Download PDFInfo
- Publication number
- CN111539532A CN111539532A CN202010251633.5A CN202010251633A CN111539532A CN 111539532 A CN111539532 A CN 111539532A CN 202010251633 A CN202010251633 A CN 202010251633A CN 111539532 A CN111539532 A CN 111539532A
- Authority
- CN
- China
- Prior art keywords
- variable
- derivation
- template
- operator
- basic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000009795 derivation Methods 0.000 title claims abstract description 64
- 238000000034 method Methods 0.000 title claims abstract description 51
- 238000013507 mapping Methods 0.000 claims description 11
- 230000002776 aggregation Effects 0.000 claims description 10
- 238000004220 aggregation Methods 0.000 claims description 10
- 238000010276 construction Methods 0.000 claims description 6
- 238000012935 Averaging Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 6
- 238000010801 machine learning Methods 0.000 description 4
- 101100503482 Arabidopsis thaliana FTSH5 gene Proteins 0.000 description 2
- 101150082136 VAR1 gene Proteins 0.000 description 2
- 230000001419 dependent effect Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000011835 investigation Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 101100013558 Arabidopsis thaliana FTSH2 gene Proteins 0.000 description 1
- 101100540153 Arabidopsis thaliana VAR3 gene Proteins 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002194 synthesizing effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q40/00—Finance; Insurance; Tax strategies; Processing of corporate or income taxes
- G06Q40/03—Credit; Loans; Processing thereof
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Marketing (AREA)
- Economics (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Technology Law (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Development Economics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Stored Programmes (AREA)
Abstract
The invention discloses a model-construction-oriented automatic feature derivation method, which can be used for automatically deriving a large number of new features from original small-dimension detail data. The automatic feature derivation method provided by the invention is used for realizing automatic generation of a large number of new features, increases the possibility of model effect, greatly shortens the modeling time, quickly realizes the service requirement and accelerates the iteration speed of the model.
Description
Technical Field
The invention belongs to the technical field of machine learning, and particularly relates to a feature automatic derivation method.
Background
In the field of machine learning technology, "data and features determine the upper limit of machine learning, and models are only approaching the upper limit," that is, good data and features are the premise that all models are brought into effect. The characteristic is information which is extracted from data and is useful for result prediction, the characteristic engineering is a process of constructing new characteristics and acquiring an efficient and accurate model based on the existing data according to knowledge and experience in the professional field, and the process is the key of machine learning.
With the continuous development of computer technology and big data application, more and more technical fields can be modeled based on big data so as to realize accurate classification and regression prediction of user groups. The key to modeling is to derive as many variables as possible from the limited business data so that valid key variables can be found therein. However, in daily modeling, a modeler usually needs to spend a lot of effort and time in processing features and deriving new features from the features, and relies on personal domain knowledge, intuition and data manipulation, which is a tedious manual process and may be very boring, and the resulting features are limited by human subjectivity and time.
Therefore, in the prior art, the manual processing mode of the feature engineering is very limited, and on one hand, limited experience of people can derive limited variables; on the other hand, to segment each dimension and form a new variable, a lot of time and effort are needed to write and maintain codes, and a lot of finally obtained features cannot necessarily bring great improvement to the model, which finally attacks the enthusiasm of modelers, and becomes one of important factors that the modelers do not want to put more effort and spend more time to perfect so as to derive the new variable.
Disclosure of Invention
In order to solve the problems, the invention provides a feature automatic derivation method oriented to model construction, which is used for realizing automatic generation of a large number of new features.
In order to achieve the purpose, the technical scheme of the invention is as follows:
a method for automatically deriving features oriented to model construction comprises the following steps:
s1: constructing a basic operator according to a derivation mode of common data, and reserving a relevant configuration corresponding to the basic operator;
s2: one or more basic operators are used for constructing a template, the configurable items corresponding to the basic operators in the template complete the setting of operator parameters, and finally a variable mapping interface is generated, wherein each basic operator is different;
s3: mapping the relation between the data and the template to realize the calling of the template interface;
s4: and calling a template interface to complete the internal mapping of the data and the basic operator, and realizing variable derivation by the basic operator so as to generate new characteristics.
In the invention, the automatic characteristic derivation method can be used for automatically deriving a large number of new characteristics from original small-dimension detail data, the method is based on a database, basic operators are constructed by adopting a derivation mode of common data, the basic operators are integrated on a template, and operator parameters are correspondingly set, so that the realization of subsequent steps is facilitated.
Specifically, in step S1, a basic operator is constructed according to the currently defined operator content, where the basic operator includes an operation mode, a variable configuration, an operation method, a time derivative setting (optional), and a derivative rule, and these five parts are independent and cooperate with each other to finally determine the variable to be derived, and the time derivative setting is optional.
Specifically, the operation mode includes a direct operation mode including generating a new variable from a direct operation between variables and a merge mode. In the invention, the direct operation mode is mainly used for 1:1 scenes, data combination is not carried out, and new characteristics are directly derived according to one column or a plurality of columns.
Specifically, the merging mode includes that the variables are converged to the dimension of the unique mark variable through aggregation operation to generate new variables, such as the total amount of all transactions and the total transaction number of a certain client are calculated.
Specifically, the variable configuration is mainly used for specifying calling variables to be used when the basic operator is called, and the calling variables include a unique flag variable (such as a customer unique number), a date variable (such as a transaction date), an aggregation variable (such as an amount) and an automatic derivation rule variable (such as a channel and a transaction type). The variable configuration can be equivalent to a database for storing variables and new variables, and required data is called from the variable configuration when operation is performed, and is finally used for forming the template external interface.
In the present invention, the aggregation variable may be a single value, or may be a result variable after a plurality of values are operated. For example, for transaction pipelining data, a large number of pipelining features may be derived; aiming at the credit investigation data, a large number of wind control characteristics can be derived, such as overdue total amount of the unsettled account in about 12 months, total amount of payment due to the unsettled account and the like.
Specifically, the operation method generally includes four arithmetic operations, addition, subtraction, multiplication and division; and a convergence operation, the convergence operation comprising: summing, averaging a maximum value and a minimum value; when the time derivation setting exists, the ratio, the difference value and the like of the current time slice to the previous time slice are calculated according to the difference of the convergence time slices after the conventional convergence operation based on the time.
In the operation method, in the process of generating a new variable, selecting a corresponding operation method according to an adopted operation mode; when the operation mode is a combination mode (1: N), the sum, the total times, the mean, the maximum value, the minimum value and the like can be selected; when the operation mode is a direct operation mode (1:1), the operation formula can be defined by user, such as calculating the ratio of the remaining principal to the loan amount; or may be derived based on the currently only supported configuration, such as an identity card drawing a birthday.
Specifically, the derivation rules mainly include the following two modes of derivation rules: one is the derivation of a person for a given derivation rule, such as a variable that defines the value of a derived variable as "consume", "cash out"; the other is an automatic derivation rule, and if only one or more derivation variables need to be specified, the system automatically adopts an exhaustion method to find out all rule combinations and realize the derivation of the variables under different rules.
In the invention, an operation mode determines whether data should be converged or directly derived, a variable configuration defines an external interface of an operator, and an operation method, time derivative setting and a derivative rule determine which variable needs to be derived.
Specifically, in step S2, the specific steps include:
s21: according to the application scene requirements, the template is correspondingly integrated with one or more different basic operators with the same operation mode;
s22: setting parameters corresponding to variable configuration, an operation method, time derivative setting and derivative rule contents required by a basic operator in a template, and then generating a template interface for a user to call;
s23: after the configuration of one or more basic operators is completed, the variable configuration of all the basic operators is automatically gathered, and a template interface and a calling parameter are synthesized.
In the invention, after the template completes the configuration of five parts in each basic operator, the basic operators have dependent variable configuration, the basic operators can be collected and de-duplicated to form a template interface, and when the template is finally applied, the database data of a user calls the template interface, so that a large amount of derivation of the data in the database can be completed. For example:
a time slicing operator needs to depend on a user ID variable, a time variable, a convergence variable and a derivative variable;
the partial proportion operator needs to depend on user ID variables, aggregation variables and derivative variables;
the template A is configured with the two operators, and finally only four variables of the user ID variable, the time variable, the convergence variable and the derivative variable need to be mapped.
The method of the invention greatly shortens the modeling time, can realize the business requirement more quickly and accelerates the iteration speed of the model.
Specifically, in step S3, the data is derived from one or more data tables, and when the data is derived from multiple data tables, the association relationship between the multiple data tables is configured to merge into one initial table, and after the initial table is formed, the template interface is called.
In the invention, when a template interface is called, all operators in the template are traversed and variable derivation is realized one by one, when a single operator operates and derives a new characteristic, data is required to be converted into an original variable required by the operator, and the operator parameters set by the template are used, so that a large amount of derivation of the characteristic is realized and a high-dimensional data table is output based on the operation logic of multi-dimensional combination. This can be extended specifically as follows: the total amount of consumption transactions in the last month, the total number of times of cash-out transactions in the last year and the like are constructed by time slicing and automation rules.
The invention has the advantages that:
compared with the prior art, the automatic feature derivation method oriented to model construction is used for automatically generating a large number of new features, can not only separate modeling personnel from the complexity of data derivation, but also increase the possibility of model effect, greatly shortens modeling time, realizes business requirements more quickly, and accelerates the iteration speed of the model.
Drawings
FIG. 1 is a schematic block diagram of an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The technical scheme of the invention is as follows:
referring to fig. 1, the invention implements a model-oriented feature automated derivation method, comprising the following steps:
s1: constructing a basic operator according to a derivation mode of common data, and reserving a relevant configuration corresponding to the basic operator; the basic operator is constructed according to the currently defined operator content, the basic operator comprises an operation mode, variable configuration, an operation method, time derivative setting (optional) and derivative rules, the five parts are independent and matched with each other, and finally, the variable to be derived is determined;
s2: one or more different basic operators are used for constructing a template, and in the template, the configurable items corresponding to the basic operators complete the setting of operator parameters, and finally a variable mapping interface is generated;
in step S2, the method specifically includes the following steps:
s21: according to the application scene requirements, the template is correspondingly integrated with one or more different basic operators with the same operation mode;
s22: setting parameters corresponding to variable configuration, an operation method, time derivative setting and derivative rule contents required by a basic operator in a template, and then generating a template interface for a user to call;
s23: after the configuration of one or more basic operators is completed, automatically converging the variable configuration of all the basic operators, synthesizing template interfaces and calling parameters, such as CUST _ ID, TRAN _ DT, CHANNEL and TRAN _ TYPE;
s3: mapping the relation between the data and the template to realize the calling of the template interface;
in step S3, the data is derived from one or more data tables, and when the data is derived from multiple data tables, the association relationship between multiple data tables is configured to merge into one initial table, and after the initial table is formed, the template interface is called;
s4: and calling a template interface to complete the internal mapping of the data and the basic operator, and realizing variable derivation by the basic operator so as to generate new characteristics.
The method is based on a database, adopts a derivation mode of common data to construct basic operators, integrates the basic operators on a template, and correspondingly sets operator parameters, so that the realization of subsequent steps is facilitated.
In the invention, parameters preset in the template, such as the configuration of variables, the unique identifier of which is CUST _ ID, the date variable of which is TRAN _ DT, the automatic derivation rule variables of which are CHANNEL and TRAN _ TYPE, the time derivation mode of which is derivation according to natural month slices, the operation method specifies the summation and the total number of times of summation and the ratio, the difference, the week average, the month average, the number of days with data and the like which need to be calculated according to the time slices, and the derivation rule specifies the values of the two variables of CHANNEL and TRAN _ TYPE for the automatic derivation rule.
In this embodiment, the operation mode includes a direct operation mode and a merge mode, and the direct operation mode includes generating a new variable, such as a + B, from a direct operation between the variables.
In the invention, the direct operation mode is mainly used for 1:1 scenes, data combination is not carried out, new characteristics are directly derived according to one or more columns, for example, in wind control, the number of days of difference between loan application time and account opening time is calculated; for another example, the year, month, sex and area of birth are extracted from the ID card number; for another example, the ratio of remaining principal to loan amount is calculated from the credit data.
In the embodiment, the merging mode includes that the variables are converged to the dimension of the unique mark variable through aggregation operation to generate new variables, such as the total amount of all transactions and the total transaction number of a certain client.
For example: and when the configuration is needed according to the operation mode of the basic operator, a plurality of new features are calculated.
When the operation mode is the merging mode (1: N), one user has a plurality of transaction data;
variable required settings (variable configuration): user ID, transaction type, transaction amount, transaction date;
operation method (operation method): summing up;
time derivation: selecting the slice data six months before backtracking;
derivation rules: obtaining rules according to transaction types;
and (3) specific operation: selecting the transaction data of the previous [ N months ], and for the data of the transaction TYPE [ TYPE ], according to the transaction amount of a user [ SUM/COUNT/AVG/MAX/MIN ]; n is 1,2,3,4,5, 6;
the TYPE comprises consumption, cash withdrawal and the like, is obtained according to a data use exhaustion method, and extracts part of rules according to statistics to serve as final derivative rules.
In the specific implementation method, the variable configuration is mainly used for specifying calling variables to be used when the basic operator is called, and the calling variables include unique flag variables (such as a unique customer number), date variables (such as a transaction date), aggregation variables (such as an amount) and automatically derived rule variables (such as a channel and a transaction type). The variable configuration can be equivalent to a database for storing variables and new variables, and required data is called from the variable configuration when operation is performed, and is finally used for forming the template external interface.
In the present invention, the aggregation variable may be a single value, or may be a result variable after a plurality of values are operated. For example, for transaction pipelining data, a large number of pipelining features may be derived; aiming at the credit investigation data, a large number of wind control characteristics can be derived, such as overdue total amount of the unsettled account in about 12 months, total amount of payment due to the unsettled account and the like.
In the present embodiment, the operation method generally includes four arithmetic operations, addition, subtraction, multiplication and division; and a convergence operation, the convergence operation comprising: summing, averaging a maximum value and a minimum value; when the time derivation setting exists, the ratio, the difference value and the like of the current time slice to the previous time slice are calculated according to the difference of the convergence time slices after the conventional convergence operation based on the time.
In the operation method, in the process of generating a new variable, selecting a corresponding operation method according to an adopted operation mode; when the operation mode is a combination mode (1: N), the sum, the total times, the mean, the maximum value, the minimum value and the like can be selected; when the operation mode is a direct operation mode (1:1), the operation formula can be defined by user, such as calculating the ratio of the remaining principal to the loan amount; or may be derived based on the currently only supported configuration, such as an identity card drawing a birthday.
In this embodiment, the derivation rules mainly include the following two modes of derivation rules: one is the derivation of a person for a given derivation rule, such as a variable that defines the value of a derived variable as "consume", "cash out"; the other is an automatic derivation rule, and if only one or more derivation variables need to be specified, the system automatically adopts an exhaustion method to find out all rule combinations and realize the derivation of the variables under different rules.
In the invention, an operation mode determines whether data should be converged or directly derived, a variable configuration defines an external interface of an operator, and an operation method, time derivative setting and a derivative rule determine which variable needs to be derived. For example: the marketing model features require the calculation of the total amount of the previous monthly consumption transaction;
the operation mode is a merging mode (1: N), and one user has a plurality of transaction data;
variables need to be set: user ID, transaction type, transaction amount, transaction date;
the operation mode is as follows: summing up;
time derivation: selecting time slice data one month before backtracking;
derivation rules: obtaining rules according to transaction types;
and (3) specific operation: selecting the transaction data before [ one month ], and for the data with the transaction type of [ consumption ], carrying out the transaction amount according to the summary of the user.
In the present embodiment, as shown in fig. 1, application scene selection: automated feature derivation of bank debit card transaction data.
First, two basic operators are constructed based on the currently defined operator content: a time slicing operator, a partial proportion operator; the time slicing operator is responsible for slicing according to different time to obtain new variables related to time, such as the total transaction number consumed in the previous 6 months; and the partial proportion operator is responsible for calculating new variables related to the proportion under different derivative rules, such as the total consumption transaction number on the line of the channel to the percentage of the total transaction number.
And then, using the time slice operator and the part proportion operator to form a transaction data derivative template, after the template completes the configuration of five parts in each operator, both operators have dependent variable configuration, and a template interface is formed after summarizing and de-duplicating, and when the template is finally applied, the database data of a user calls the template interface, so that a large amount of derivative of the data in the database can be completed.
In the process, the time slice operator needs to depend on a user ID variable, a time variable, a derivative variable and a convergence variable; the partial proportion operator needs to depend on user ID variables, aggregation variables and derivative variables; the trading data derivative template is configured with the two operators, and finally only four variables of a user ID variable, a time variable, a convergence variable and a derivative variable need to be mapped.
And because the data used is from different tables, table association is required to form the initial table (result data table) before mapping. And mapping the user ID, the transaction time, the channel/transaction type and the transaction amount to a user ID variable, a time variable, a convergence variable and a derivative variable respectively from the result data sheet, so that template calling can be completed, and operator derivation is used for completing variable derivation.
According to the above, assuming that several ten variables of TABLE in the initial TABLE are VAR1 and VAR2.. VAR10, when calling an interface, it is only necessary to set VAR1 corresponding to the CUST _ ID, VAR3 corresponding to TRAN _ DT, VAR10 corresponding to CHANNEL, and VAR4 corresponding to TRAN _ TYPE.
In the invention, when a template interface is called, all operators in the template are traversed and variable derivation is realized one by one, when a single operator operates and derives a new characteristic, data is required to be converted into an original variable required by the operator, and the operator parameters set by the template are used, so that a large amount of derivation of the characteristic is realized and a high-dimensional data table is output based on the operation logic of multi-dimensional combination. This can be extended specifically as follows: the total amount of consumption transactions in the last month, the total number of times of cash-out transactions in the last year and the like are constructed by time slicing and automation rules.
The invention has the advantages that:
compared with the prior art, the automatic feature derivation method for model construction is used for automatically generating a large number of new features, can not only separate modeling personnel from the complexity of data derivation, but also increase the possibility of model effect, greatly shortens the modeling time, and can meet business requirements (if feature derivation is needed on transaction data of a bank debit card) more quickly and accelerate the iteration speed of the model.
The above description is illustrative of the preferred embodiment of the present invention and is not to be construed as limiting the invention, but rather as encompassing all the modifications, equivalents, and improvements made within the spirit and principles of the invention.
Claims (8)
1. A method for automatically deriving features oriented to model construction is characterized by comprising the following steps:
s1: constructing a basic operator according to a derivation mode of common data, and reserving a relevant configuration corresponding to the basic operator;
s2: one or more basic operators are used for constructing a template, the configurable items corresponding to the basic operators in the template complete the setting of operator parameters, and finally a variable mapping interface is generated, wherein each basic operator is different;
s3: mapping the relation between the data and the template to realize the calling of the template interface;
s4: and calling a template interface to complete the internal mapping of the data and the basic operator, and realizing variable derivation by the basic operator so as to generate new characteristics.
2. The method for automatically deriving features oriented to model construction according to claim 1, wherein in step S1, a basic operator is constructed according to the currently defined operator content, the basic operator includes an operation mode, a variable configuration, an operation method, a time-derived setting and a derivation rule, and the five parts are independent and cooperate with each other to finally determine the variables to be derived, and the time-derived setting is an option.
3. The model-build oriented feature automated derivation method of claim 2, wherein the operational modes include a direct operational mode and a merge mode, the direct operational mode including generating new variables from direct operations between variables; the merging mode comprises the step of generating new variables in a mode that the variables are converged to the dimension of the unique mark variable through aggregation operation.
4. The model-building-oriented feature automated derivation method of claim 2, wherein the variable configuration comprises call variables for specifying a base operator to be used when called, the call variables comprising a unique flag variable, a date variable, an aggregation variable, and an automatically derived rule variable.
5. The model-building-oriented automated feature derivation method according to claim 2, wherein the operational method comprises a four-way operation and a convergence operation, the convergence operation comprising: summing, averaging a maximum value and a minimum value; when the time derivation setting exists, convergence operation is carried out according to different convergence time slices, and the ratio and the difference value of the current time slice to the previous time slice are calculated.
6. The model-build oriented feature automated derivation method of claim 2, wherein the derivation rules comprise manually specified derivation rules and automated derivation rules.
7. The method for automatically deriving features oriented to model building according to claim 2, wherein in step S2, the specific steps include:
s21: according to the application scene requirements, the template is correspondingly integrated with one or more different basic operators with the same operation mode;
s22: setting parameters corresponding to variable configuration, an operation method, time derivative setting and derivative rule contents required by a basic operator in a template, and then generating a template interface for a user to call;
s23: after the configuration of one or more basic operators is completed, the variable configuration of all the basic operators is automatically gathered, and a template interface and a calling parameter are synthesized.
8. The method for automatically deriving features oriented to model building according to claim 1, wherein in step S3, data is derived from one or more data tables, and when derived from a plurality of data tables, the association relationship between the plurality of data tables is configured to be combined into an initial table, and after the initial table is formed, the template interface is called.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010251633.5A CN111539532A (en) | 2020-04-01 | 2020-04-01 | Model construction-oriented automatic feature derivation method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010251633.5A CN111539532A (en) | 2020-04-01 | 2020-04-01 | Model construction-oriented automatic feature derivation method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111539532A true CN111539532A (en) | 2020-08-14 |
Family
ID=71978487
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010251633.5A Pending CN111539532A (en) | 2020-04-01 | 2020-04-01 | Model construction-oriented automatic feature derivation method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111539532A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112037013A (en) * | 2020-08-25 | 2020-12-04 | 成都榕慧科技有限公司 | Pedestrian credit variable derivation method and device |
| CN114218929A (en) * | 2022-02-22 | 2022-03-22 | 之江实验室 | Multi-platform operator intelligent development system and method based on meta-operator fusion |
| CN114553395A (en) * | 2022-04-24 | 2022-05-27 | 蓝象智联(杭州)科技有限公司 | Longitudinal federal feature derivation method in wind control scene |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017202006A1 (en) * | 2016-05-25 | 2017-11-30 | 腾讯科技（深圳）有限公司 | Data processing method and device, and computer storage medium |
| CN109977151A (en) * | 2019-03-28 | 2019-07-05 | 北京九章云极科技有限公司 | A kind of data analysing method and system |
| WO2019154108A1 (en) * | 2018-02-12 | 2019-08-15 | 阿里巴巴集团控股有限公司 | Method and apparatus for processing transaction data |
| CN110475370A (en) * | 2019-07-31 | 2019-11-19 | 苏州研数信息科技有限公司 | The derivative structure's variable method of intelligence in data modeling |
| US20200089650A1 (en) * | 2018-09-14 | 2020-03-19 | Software Ag | Techniques for automated data cleansing for machine learning algorithms |
-
2020
- 2020-04-01 CN CN202010251633.5A patent/CN111539532A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2017202006A1 (en) * | 2016-05-25 | 2017-11-30 | 腾讯科技（深圳）有限公司 | Data processing method and device, and computer storage medium |
| WO2019154108A1 (en) * | 2018-02-12 | 2019-08-15 | 阿里巴巴集团控股有限公司 | Method and apparatus for processing transaction data |
| US20200089650A1 (en) * | 2018-09-14 | 2020-03-19 | Software Ag | Techniques for automated data cleansing for machine learning algorithms |
| CN109977151A (en) * | 2019-03-28 | 2019-07-05 | 北京九章云极科技有限公司 | A kind of data analysing method and system |
| CN110475370A (en) * | 2019-07-31 | 2019-11-19 | 苏州研数信息科技有限公司 | The derivative structure's variable method of intelligence in data modeling |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112037013A (en) * | 2020-08-25 | 2020-12-04 | 成都榕慧科技有限公司 | Pedestrian credit variable derivation method and device |
| CN114218929A (en) * | 2022-02-22 | 2022-03-22 | 之江实验室 | Multi-platform operator intelligent development system and method based on meta-operator fusion |
| CN114218929B (en) * | 2022-02-22 | 2022-05-17 | 之江实验室 | Multi-platform operator intelligent development system and method based on meta-operator fusion |
| CN114553395A (en) * | 2022-04-24 | 2022-05-27 | 蓝象智联(杭州)科技有限公司 | Longitudinal federal feature derivation method in wind control scene |
| CN114553395B (en) * | 2022-04-24 | 2022-07-26 | 蓝象智联(杭州)科技有限公司 | Longitudinal federal feature derivation method in wind control scene |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111539532A (en) | Model construction-oriented automatic feature derivation method | |
| JP2001522100A5 (en) | ||
| CN109035060A (en) | A kind of structured processes management system and management method | |
| CN110263307A (en) | Report form generation method, server and computer readable storage medium | |
| CN113642920B (en) | Quick construction method for digital application in industrial manufacturing field | |
| CN102496083B (en) | Credit rating report manuscript manufacturing method | |
| CN106980689B (en) | Method for realizing data visualization through voice interaction | |
| CN110851667A (en) | Integrated analysis method and tool for multi-source large data | |
| US6959429B1 (en) | System for developing data collection software applications | |
| CN107665457A (en) | A kind of method and system of commodity editor restocking | |
| US20120253997A1 (en) | Method for multi-dimensional accounting of business transactions and system therefor | |
| CN104346698A (en) | Catering member big data analysis and checking system based on cloud computing and data mining | |
| CN112527250A (en) | Software development platform based on visualization | |
| CN114547170A (en) | Method for standardized extraction and modeling of unstructured financial statement data | |
| CN107341685A (en) | Data analysing method and device | |
| CN113282623A (en) | Data processing method and device | |
| CN112785427A (en) | Enterprise credit analysis system based on electric power data | |
| CN108268615A (en) | A kind of data processing method, device and system | |
| CN113268496A (en) | Client list screening method and device | |
| CN108595727A (en) | Design method, device and the realization device that case becomes | |
| CN107423905A (en) | A kind of energy performance appraisal method and system | |
| Clarete et al. | CGE models and development policy analysis: problems, pitfalls, and challenges | |
| WO2021169094A1 (en) | Commodity transaction method and system for transaction platform | |
| CN109584047A (en) | A kind of credit method, system, computer equipment and medium | |
| CN112215454B (en) | Intelligent decision-making system for power grid enterprise management |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |