CN110275903A - Improve the method and system of the feature formation efficiency of machine learning sample - Google Patents
Improve the method and system of the feature formation efficiency of machine learning sample Download PDFInfo
- Publication number
- CN110275903A CN110275903A CN201910578459.2A CN201910578459A CN110275903A CN 110275903 A CN110275903 A CN 110275903A CN 201910578459 A CN201910578459 A CN 201910578459A CN 110275903 A CN110275903 A CN 110275903A
- Authority
- CN
- China
- Prior art keywords
- flowing water
- water table
- single order
- temporal aspect
- machine learning
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2455—Query execution
- G06F16/24553—Query execution of query operations
- G06F16/24554—Unary operations; Data partitioning operations
- G06F16/24556—Aggregation; Duplicate elimination
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2474—Sequence data queries, e.g. querying versioned data
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Databases & Information Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Fuzzy Systems (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Computing Systems (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
Provide a kind of method and system of feature formation efficiency for improving machine learning sample.The described method includes: obtaining the flowing water table that record has the data record with timing, wherein the corresponding data record of a line of the flowing water table, the one of the flowing water table arrange a corresponding field；The temporal aspect counted for needs, the single order for first carrying out coarseness to the respective field in the flowing water table polymerize and saves single order polymerization result, then obtains the temporal aspect that every data in the flowing water table records corresponding machine learning sample based on the single order polymerization result saved.According to the method and system, the formation efficiency of temporal aspect can be improved.
Description
Technical field
All things considered of the present invention is related to artificial intelligence field, more particularly, is related to a kind of improving machine learning sample
The method and system of feature formation efficiency.
Background technique
Currently, for each pipelined data, needing when carrying out temporal aspect generation for pipelined data with this stream
Water number recalls forward the pipelined data in certain time length according to corresponding timing node, and whole pipelined datas based on backtracking come
Obtain the temporal aspect for this pipelined data.Since the corresponding timing node of every pipelined data very maximum probability can be different,
Therefore, lower for the multiplexing possibility of the temporal aspect of different pipelined datas and storage cost is larger.This to calculate every time
For pipelined data temporal aspect when, require to combine all pipelined datas, and based on be directed to every flowing water number
The temporal aspect for this pipelined data is obtained according to the whole pipelined datas recalled forward, accordingly, there exist calculating costs greatly,
The low problem of temporal aspect formation efficiency.
Summary of the invention
Exemplary embodiment of the present invention is to provide a kind of method of feature formation efficiency for improving machine learning sample
And system, it is able to solve the low problem of the formation efficiency of temporal aspect of the existing technology.
An exemplary embodiment of the present invention provides a kind of side of feature formation efficiency for improving machine learning sample
Method, wherein the described method includes: obtaining the flowing water table that record has the data record with timing, wherein the one of the flowing water table
The corresponding data record of row, the one of the flowing water table arrange a corresponding field；For the temporal aspect that needs count, first to institute
The single order for stating the respective field progress coarseness in flowing water table polymerize and saves single order polymerization result, then based on the single order saved
Every data that polymerization result obtains in the flowing water table records the temporal aspect of corresponding machine learning sample.
Optionally, the flowing water table is bank transaction flowing water table；Alternatively, the flowing water table is Internet user's behavior record
Table.
Optionally, the method also includes: provide a user root feature operator configuration interface；It obtains user and passes through described
Feature operator configures the Gent sign of interface configuration, and the corresponding word for being directed to the temporal aspect is determined based on Gent sign
Section.
Optionally, the method also includes: provide a user at least one of following Aggregation Operator configuration interface: being used for
Configure the configuration interface of the time window of the temporal aspect, the configuration of polymerization methods for configuring the temporal aspect connects
Mouthful, the configuration interface of polymerization methods for configuring single order polymerization, division mode for configuring the coarseness match
Set interface；Obtain the content that user configures interface configuration by the Aggregation Operator, wherein execute to obtain based on the content
Every data in the flowing water table records the step of temporal aspect of corresponding machine learning sample.
Optionally, for the temporal aspect that counts of needs, the single order that the respective field in convection current water meter carries out coarseness is poly-
The step of conjunction includes: respectively for each user identification field value in the flowing water table, to having the use in the flowing water table
Family identification field values and time field value belong to the field value of the respective field of the data record of each coarseness period
Carry out single order polymerization respectively to obtain single order polymerization result.
Optionally, for the temporal aspect that counts of needs, the single order that the respective field in convection current water meter carries out coarseness is poly-
The step of conjunction includes: when the temporal aspect is related to only to the described corresponding of the certain types of data record in the flowing water table
It, should to having in the flowing water table respectively for each user identification field value in the flowing water table when field is counted
User identification field value, time field value belong to each coarseness period and meet the certain types of data record
The field value of the respective field carries out single order polymerization respectively to obtain single order polymerization result.
Optionally, the temporal aspect counted for needs, obtains the flowing water table based on the single order polymerization result saved
In every data the step of recording the temporal aspect of corresponding machine learning sample include: respectively for the flowing water table
In every data record, to the corresponding coarseness period belong to the data record the corresponding period and institute it is right
The identical single order polymerization result that the user identification field value answered and the data record carries out two that second order polymerize, and will obtain
Rank polymerization result records the temporal aspect of corresponding machine learning sample as the data, wherein every data record
The corresponding period are as follows: the period of the specific duration before the time field value of data record.
Optionally, the polymerization methods of the single order polymerization include at least one among following item: summing, are averaging, take
Maximum value is minimized, seeks standard deviation, counts；The polymerization methods of the second order polymerization include at least one among following item:
It sums, be averaging, being maximized, being minimized, seek standard deviation, count.
Optionally, the method also includes: for other temporal aspects for counting of needs, respectively in the flowing water table
Every data record, to time field value belong to the data record corresponding period and user identification field value with
The quantity of the different field value of the field corresponding with other described temporal aspects of the identical data record of data record
It is counted, records other temporal aspects described in corresponding machine learning sample to obtain the data.
Optionally, the feature formation efficiency for improving machine learning sample is executed by executing feature generation script
Method.
In accordance with an alternative illustrative embodiment of the present invention, a kind of feature formation efficiency of raising machine learning sample is provided
System, wherein the system comprises: flowing water table acquisition device, the flowing water for having the data record with timing suitable for obtaining record
Table, wherein the corresponding data record of a line of the flowing water table, the one of the flowing water table arrange a corresponding field；Timing is special
Generating means are levied, suitable for first carrying out coarseness to the respective field in the flowing water table for the temporal aspect for needing to count
Single order polymerize and saves single order polymerization result, then obtains every number in the flowing water table based on the single order polymerization result saved
According to the temporal aspect for recording corresponding machine learning sample.
Optionally, the flowing water table is bank transaction flowing water table；Alternatively, the flowing water table is Internet user's behavior record
Table.
Optionally, the system also includes: configuration interface provide device, suitable for provide a user root feature operator configuration connect
Mouthful；Content acquisition unit is levied suitable for obtaining user by the Gent that described feature operator configures interface configuration；Field determines dress
It sets, suitable for determining the respective field for being directed to the temporal aspect based on Gent sign.
Optionally, the system also includes: configuration interface provides device, suitable for providing a user the configuration of following Aggregation Operator
At least one of interface: for configuring the configuration interface of the time window of the temporal aspect, for configuring the timing spy
The configuration interface, described for configuring of the configuration interface of the polymerization methods of sign, polymerization methods for configuring single order polymerization
The configuration interface of the division mode of coarseness；Content acquisition unit is suitable for obtaining user by the Aggregation Operator and configures interface
The content of configuration, wherein temporal aspect generating means obtain every in the flowing water table suitable for executing based on the content
The temporal aspect of the corresponding machine learning sample of data record.
Optionally, temporal aspect generating means, suitable for respectively be directed to the flowing water table in each user identification field value,
To in the flowing water table with the user identification field value and time field value belongs to the data record of each coarseness period
The respective field field value carry out respectively single order polymerize to obtain single order polymerization result.
Optionally, temporal aspect generating means, suitable for being related to only when the temporal aspect to specific in the flowing water table
When the respective field of the data record of type is counted, respectively for each user identification field in the flowing water table
Value, in the flowing water table there is the user identification field value, time field value to belong to each coarseness period and meet institute
The field value for stating the respective field of certain types of data record carries out single order respectively and polymerize to obtain single order polymerization result.
Optionally, temporal aspect generating means, suitable for respectively be directed to the flowing water table in every data record, to pair
The coarseness period answered belong to record the corresponding period with the data and corresponding user identification field value and this
The identical single order polymerization result of data record carries out second order polymerization, and remembers obtained second order polymerization result as the data
Record the temporal aspect of corresponding machine learning sample, wherein every data records the corresponding period are as follows: in the data
The period of specific duration before the time field value of record.
Optionally, the polymerization methods of the single order polymerization include at least one among following item: summing, are averaging, take
Maximum value is minimized, seeks standard deviation, counts；The polymerization methods of the second order polymerization include at least one among following item:
It sums, be averaging, being maximized, being minimized, seek standard deviation, count.
Optionally, temporal aspect generating means are further adapted for for other temporal aspects for needing to count, respectively for described
Every data record in flowing water table, belongs to time field value and records corresponding period and user identifier with the data
The different field of field value and the field corresponding with other described temporal aspects for the identical data record that the data records
The quantity of value is counted, and records other temporal aspects described in corresponding machine learning sample to obtain the data.
Optionally, it is raw to execute the feature for improving machine learning sample by executing feature generation script for the system
At the method for efficiency.
In accordance with an alternative illustrative embodiment of the present invention, providing a kind of includes that at least one computing device is deposited at least one
The system for storing up the storage device of instruction, wherein described instruction when being run by least one described computing device, promote it is described extremely
The method that a few computing device executes the feature formation efficiency as described above for improving machine learning sample.
In accordance with an alternative illustrative embodiment of the present invention, a kind of computer readable storage medium of store instruction is provided,
In, when described instruction is run by least one computing device, promote at least one described computing device to execute as described above
The method for improving the feature formation efficiency of machine learning sample.
The method and system of the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample, lead to
The temporal aspect generating mode of second order polymerization is carried out again after the single order polymerization for first carrying out coarseness to improve single order polymerization result
Recycling rate of waterused, so as to reduce the calculating cost of temporal aspect, improve the computational efficiency of temporal aspect.In addition, pass through to
User provides configurable operator and configures interface, and user only needs to execute by algorithm configuration interface easily operated, intuitive convenient for reason
The configuration of solution operates, and can be realized and automatically generates the temporal aspect for meeting its demand, has both improved the convenient of building temporal aspect
Property improves the easy to be explanatory of temporal aspect, also reduces the threshold of machine learning.
Part in following description is illustrated into the other aspect and/or advantage of present general inventive concept, there are also one
Dividing will be apparent by description, or can learn by the implementation of present general inventive concept.
Detailed description of the invention
By below with reference to be exemplarily illustrated embodiment attached drawing carry out description, exemplary embodiment of the present it is upper
Stating will become apparent with other purposes and feature, in which:
Fig. 1 shows the method for the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Flow chart；
Fig. 2 shows the feature formation efficiencies of raising machine learning sample in accordance with an alternative illustrative embodiment of the present invention
The flow chart of method；
Fig. 3 shows the system of the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Block diagram.
Specific embodiment
It reference will now be made in detail the embodiment of the present invention, examples of the embodiments are shown in the accompanying drawings, wherein identical mark
Number identical component is referred to always.It will illustrate the embodiment, by referring to accompanying drawing below to explain the present invention.
Fig. 1 shows the method for the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Flow chart.
Referring to Fig.1, in step S10, the flowing water table that record has the data record with timing is obtained, wherein the flowing water
The corresponding data record of a line of table, the one of the flowing water table arrange a corresponding field.
As an example, the flowing water table can be the tables of data for recording the user behavior with timing.As an example,
The flowing water table can be bank transaction flowing water table, each data record in the bank transaction flowing water table can be about
The description of one bank transaction.As an example, the flowing water table can be Internet user's behavior record table, the internet is used
Each data record in the behavior record table of family can be about Internet user's behavior (for example, the use of Website login
Family behavior) description.It should be understood that the tables of data, which is also possible to other kinds of record, the data record with timing
Tables of data.
Coarseness first is carried out to the respective field in the flowing water table for the temporal aspect that needs count in step S20
Single order polymerize and save single order polymerization result, then every in the flowing water table is obtained based on the single order polymerization result saved
The temporal aspect of the corresponding machine learning sample of data record.
Particularly, the temporal aspect counted for needs first can carry out coarse grain to the respective field in the flowing water table
The single order of degree polymerize and saves single order polymerization result, then is based on being saved for every data record in the flowing water table respectively
Single order polymerization result carry out second order polymerization, come obtain every data record corresponding machine learning sample the timing it is special
Sign.Here, the size of size, that is, temporal aspect time window of the granularity of second order polymerization, also, the grain of second order polymerization
The size of degree is greater than the size of the granularity of single order polymerization.It should be understood that subsequent again for described in newly-increased data record generation
When temporal aspect, the single order polymerization result used if necessary has been saved, then does not need to repeatedly generate, and direct use is
Can, only generating needs that be used and unsaved a part of single order polymerization result.
Temporal aspect is with the time is slide dimension a kind of feature, can by same main body different time nodes it
Information in preceding certain time length (that is, time window) is counted to obtain.
As an example, carrying out coarseness to the respective field in the flowing water table for the temporal aspect for needing to count
The step of single order polymerize can include: respectively for each user identification field value in the flowing water table, in the flowing water table
With the user identification field value and time field value belongs to the respective field of the data record of each coarseness period
Field value carry out respectively single order polymerize to obtain single order polymerization result.In other words, in the flowing water table have the same use
Family identification field values and time field value belong to the field of the respective field of the data record of the same coarseness period
Value carries out single order polymerization to obtain a single order polymerization result, that is, using user identification field and time field as polymerization major key
Carry out polymerization calculating, each single order polymerization result corresponds to a user identification field value and a coarseness period.
As an example, when the flowing water table is the tables of data for recording the user behavior with timing, the user
Identification field can be used to indicate behavior generate main body field, the time field can be used to indicate behavior generate when
Between field.
As another example, the temporal aspect counted for needs carries out coarse grain to the respective field in the flowing water table
The step of single order polymerization of degree can include: when the temporal aspect is related to only remembering the certain types of data in the flowing water table
When the respective field of record is counted, respectively for each user identification field value in the flowing water table, to the stream
In water meter there is the user identification field value, time field value to belong to each coarseness period and meet the specific type
Data record the respective field field value carry out respectively single order polymerize to obtain single order polymerization result.That is, by user
Identification field and time field carry out single order polymerization meter as polymerization major key again after being screened according to the specific type
It calculates.
As an example, directly the specific type whether can be met according to the field value of the specific fields of data record, come
Judge whether the data record meets the specific type.For example, when the specific type is POS (point-of-sale terminal) type of transaction
When, it can directly judge whether the field value of the transaction channel field of data record indicates POS type of transaction.
As another example, flag bit field can be increased in the flowing water table, and according to the flag bit word of data record
The field value of section judges whether data record meets the specific type, wherein the field value of flag bit field is based on institute
State the field value generation of specific fields.For example, remembering when the specific type is POS type of transaction for each data
Record enables the mark of data record when the field value of the transaction channel field of data record indicates POS type of transaction
Bit field value is 1, is otherwise 0, thus, it may be determined that the data record that flag bit field value is 1 meets the specific type.
As an example, the polymerization methods of the single order polymerization may include at least one among following item: summing, ask flat
, it is maximized, is minimized, seek standard deviation, count.
As an example, can be according to the division mode of coarseness appropriate, to determine each coarseness period.For example, thick
The division mode of granularity can be to divide by consecutive days or divide by the hour.
As an example, the modes appropriate such as hdfs (distributed file system), which can be used, is saved in phase for single order polymerization result
In the storage medium answered.
As an example, obtaining the flowing water based on the single order polymerization result saved for the temporal aspect for needing to count
Every data in table records the step of temporal aspect of corresponding machine learning sample can include: is directed to the stream respectively
In water meter every data record, to the corresponding coarseness period belong to the data record the corresponding period and
Corresponding user identification field value carries out second order with the identical single order polymerization result that the data records and polymerize, and will obtain
Second order polymerization result the temporal aspect of corresponding machine learning sample is recorded as the data, wherein every data
Record the corresponding period are as follows: the period of the specific duration before the time field value of data record.That is, by user
Identification field and time field to carry out single order polymerization result second order polymerization to calculate as polymerization major key.
As an example, the polymerization methods of the second order polymerization may include at least one among following item: summing, ask flat
, it is maximized, is minimized, seek standard deviation, count.
It should be understood that different time windows (that is, specific duration), which can be used, carries out second order polymerization to obtain the timing
Feature, for example, the size of time window can be 30 days, 150 days etc..
An exemplary embodiment of the present invention generates temporal aspect using two-stage calculating, by saving coarseness
Single order polymerization result for fine-grained second order polymerize, come avoid computing repeatedly with reduce calculate cost, improve computational efficiency；
Also, stored single order polymerization result can be also repeated in the subsequent extraction temporal aspect for the data record increased newly to be made
With only needing to obtain single order polymerization result for corresponding increment for newly-increased data record and save, to pass through
Incremental update can easily obtain corresponding temporal aspect.
As an example, the method for the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
It may also include that and provide a user root feature operator configuration interface；It obtains user and interface configuration is configured by described feature operator
Gent sign, and based on the Gent sign determine be directed to the temporal aspect the respective field.
As an example, based on Gent sign other than determining the respective field for the temporal aspect, also
It can determine the specific type.
As an example, when the flowing water table is bank transaction flowing water table, if user is connect by the configuration of root feature operator
The content of mouth input are as follows: enchashment transaction, then the root feature obtained is: enchashment transaction can determine that described specific based on Gent sign
Type are as follows: encashment service type, and the specific fields corresponding with the specific type are as follows: enchashment transaction ID field；With
The corresponding field of the temporal aspect for needing to count includes: transaction amount field.Wherein, enchashment transaction ID field is used to indicate friendship
It whether is easily enchashment transaction.
As another example, when the flowing water table is bank transaction flowing water table, if user is matched by root feature operator
Set the content of interface input are as follows: number/amount of money of enchashment transaction, then the root feature obtained is: enchashment transaction can specifically incite somebody to action
User is configured in the content of interface input by root feature operator for describing transaction attribute itself (for example, type of service etc.)
Contents extraction, which comes out, is used as Gent to levy.It should be understood that the content of which type can be used as Gent sign can be according to specific applied field
Scape is preset, for example, being related in bank transaction flowing water table when application scenarios are anti money washing application scenarios for describing to hand over
The content of the field value of the field of easy attribute itself can be used as Gent sign.
As an example, user can configure interface by described feature operator, Gent sign is configured by mode appropriate.
For example, Gent sign can be configured by SQL (Structured Query Language, structured query language) sentence.
As an example, the method for the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
May also include that and provide a user at least one of following Aggregation Operator configuration interface: for configure the temporal aspect when
Between the configuration interface of window, polymerization methods for configuring the temporal aspect configuration interface, poly- for configuring the single order
The configuration interface of the configuration interface of the polymerization methods of conjunction, division mode for configuring the coarseness；It obtains user and passes through institute
State the content of Aggregation Operator configuration interface configuration, wherein step S20 is executed based on the content.
An exemplary embodiment of the present invention, by providing a user configurable algorithm configuration interface, user is only needed
Configuration operation that is easily operated, being intuitively easy to understand is executed by algorithm configuration interface, can automatically generate and meet user demand
Machine learning sample temporal aspect, on the one hand, the convenience of significant increase building temporal aspect simultaneously improves timing spy
That levies is easy to be explanatory；On the other hand, do not have the business personnel of professional ability relevant to machine learning also can complete independently,
The threshold of machine learning is greatly reduced, and Feature Engineering can be also an apprentice of in the study to the business of target domain and be liberated
Out, it puts into more professional production work.
Other than the above-mentioned temporal aspect that can be obtained based on corresponding single order polymerization result, it may also need to obtain
Others can not be based on corresponding single order polymerization result come the temporal aspect of obtained machine learning sample, that is, can not be based on two
Secondary other temporal aspects for polymerizeing to obtain, therefore, as an example, raising machine learning according to an exemplary embodiment of the present invention
The method of the feature formation efficiency of sample may also include that for other temporal aspects for needing to count, and be directed to the flowing water respectively
Every data record in table, belongs to time field value and records corresponding period and user identification field with the data
It is worth and the different field value of the field corresponding with other described temporal aspects for the identical data record that the data records
Quantity is counted, and records other temporal aspects described in corresponding machine learning sample to obtain the data.
In addition, as an example, the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Method may also include that when needing other temporal aspects for counting to be related to only remembering the data of the specified type in the flowing water table
When the respective field of record is counted, respectively for every data record in the flowing water table, to time field value category
In and that the data records the corresponding period, user identification field value and the data record is identical and meet specified class
The quantity of the different field value of the field corresponding with other described temporal aspects of the data record of type is counted, to be somebody's turn to do
Data records other described temporal aspects of corresponding machine learning sample.
In addition, as an example, the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Method may also include that the operator provided a user for configuring other temporal aspects configures interface, obtain user and pass through
The content of the operator configuration interface configuration, and execute to obtain every data record in the flowing water table based on the content
The step of other described temporal aspects of corresponding machine learning sample.For example, the operator configuration interface can be used for configuring institute
State the time windows of other temporal aspects, field corresponding with other described temporal aspects, at least one in the specified type
?.
In an embodiment of the present invention, the temporal aspect obtained based on corresponding single order polymerization result, and it is based on it
His method obtains other than other temporal aspects, can also carry out other behaviors of user to the Various types of data table including flowing water table
The extraction of feature, and finally carry out feature summarize, and by obtained each category feature by major key (generally user identifier) into
Row association, obtains user behavior characteristics summary sheet, and every a line in the table is corresponding with a user, the corresponding feature of each column.
In the following, providing referring to Fig. 2 when application scenarios are the machine learning sample generated for training money laundering ancestor's prediction model
Exemplary embodiment when this temporal aspect.
In step S101, bank transaction flowing water table is obtained, wherein the corresponding number of a line of the bank transaction flowing water table
According to record, the one of the flowing water table arranges a corresponding field, and each data record is for describing a bank transaction.
In step S102, root feature operator configuration interface is provided a user, user is obtained and is matched by described feature operator
Set the Gent sign of interface configuration, and the determining respective field that the corresponding temporal aspect for needing to count is levied with the Gent.
In step S103, for the temporal aspect, first the respective field in bank transaction flowing water table is carried out thick
The single order of granularity polymerize and saves single order polymerization result.
In step S104, the record of every data in bank transaction flowing water table is obtained based on the single order polymerization result saved
The temporal aspect of corresponding machine learning sample.
As an example, when the user obtained is by the root feature that root feature operator configures interface configuration: when POS trades,
It can determine that levying the temporal aspect that corresponding needs count with the Gent is related to only to POS type of transaction based on Gent sign
The respective field of data record is counted, and specific fields corresponding with the specific type are as follows: type of transaction field；
The respective field includes: transaction amount field.
Further, as an example, when the division mode of coarseness is to divide by consecutive days, step S103 can include:
Respectively for each user identification field value in bank transaction flowing water table, to having the user identifier in bank transaction flowing water table
Field value, time field value belong to each consecutive days and meet the word of the transaction amount field of the data record of POS type of transaction
Segment value carries out single order polymerization respectively to obtain single order polymerization result.As an example, when the mode of single order polymerization is summation, one
Rank polymerization result indicates user's odd-numbered day POS transaction total amount；When the mode of single order polymerization is to be maximized, single order polymerization knot
Fruit indicates user's odd-numbered day single POS transaction Maximum Amount；When the mode of single order polymerization is to count, the instruction of single order polymerization result
User's odd-numbered day POS transaction count.It is used it should be understood that instruction can also be obtained by the counting statistics of the field value to other fields
The single order polymerization result of family odd-numbered day POS transaction count.
Further, as an example, step S104 can include: respectively for every data note in bank transaction flowing water table
Record, to belong to corresponding consecutive days record the corresponding period with the data and corresponding user identification field value with
The identical single order polymerization result of data record carries out second order polymerization, and using obtained second order polymerization result as this number
According to the temporal aspect for recording corresponding machine learning sample.Wherein, every data records the corresponding period are as follows: at this
The period of specific duration before the time field value of data record, the size of the specific duration, that is, time window.As
Example, when the single order polymerization mode be summation and second order polymerization mode be also summation when, second order polymerization result indicate user
In this POS transaction total amount before corresponding timing node in specific duration (for example, 7 days and 30 days) of trading.
For example, when a user on January this 10 days 10, -2019 years on the 1st January in 2019 only on January 3rd, 2019 this
It generates 10 POS transaction, and generates 8 POS transaction in this day on January 5 in 2019, then can in step S103, obtain with
The user and corresponding POS transaction this single order polymerization result of total amount in 3 days January 2019 consecutive days and the user and nature
Day on January 5th, 2019, corresponding POS traded this single order polymerization result of total amount；It can be obtained in step S104 for data record
Take in 10 days POS transaction this temporal aspect of total amount when, when a data record correspond to the user and corresponding friendship
42 when dividing 8 seconds this timing nodes when the easy time is 11 days 7 January in 2019, can obtain January 1 in 2019 corresponding with the user
Day on January 10th, 1 in this 10 days odd-numbered day POS transaction total amount is summarized, i.e., to the user and 2019 1 consecutive days
Months 3 days corresponding POS transaction this single order polymerization result of total amount and corresponding with the user and January 5 2019 consecutive days
POS transaction this single order polymerization result of total amount is summarized, and temporal aspect corresponding with the data record can be obtained.
As an example, can also provide a user Aggregation Operator configuration interface before step S103, obtain user and pass through institute
The content of Aggregation Operator configuration interface configuration is stated, and executes step S103 and step S104 based on the content.
As an example, can also be for other temporal aspects for needing to count, respectively for every in bank transaction flowing water table
Data record, belongs to time field value and records corresponding period and user identification field value and this with the data
The quantity of the different field value of the field corresponding with other described temporal aspects of the identical data record of data record carries out
Statistics, records other temporal aspects described in corresponding machine learning sample to obtain the data.
For example, being to record transaction different in the corresponding period from every data when needing other temporal aspects counted
When the quantity of opponent, field corresponding with other described temporal aspects correspondingly can distinguish needle for counterparty's identification field
To in bank transaction flowing water table every data record, to time field value belong to the data record the corresponding period,
And the different field value of counterparty's identification field of user identification field value and the identical data record of data record
Quantity counted, record other temporal aspects described in corresponding machine learning sample to obtain the data.
As an example, raising machine according to an exemplary embodiment of the present invention can be executed by executing feature generation script
The method of the feature formation efficiency of learning sample.For example, the feature generation script can be for example, by the script modes such as spark reality
It is existing.An exemplary embodiment of the present invention generates the application programming interface of the programming language exposure of script by feature
(API) and attribute dynamic sensing and can adapt to the variation of upstream operator.
Traditional data warehouse technology (ETL) method (for example, SQL query) is not due to having field adaptation mechanism to cope with
The variation of upstream data field, if upstream field is caused to change (for example, newly-increased or deletion), subsequent all corresponding operators
It requires to remodify, there is a problem of that inflexible for use, excessive artificial work is easy to be mixed into mistake.It is according to the present invention to show
Example property embodiment, configured using automatized script, can by the variation of adaptation mechanism automatic sensing upstream operator, and after
Continuous feature product process can adaptively change upstream, for example, change upstream operator data field is not necessarily to change downstream operator, under
The newest data schema in upstream can be obtained by API by swimming operator, to reach automation, reduce what manual operation may introduce
Mistake.
Fig. 3 shows the system of the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Block diagram.
As shown in figure 3, the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample is
System includes: flowing water table acquisition device 10 and temporal aspect generating means 20.
Particularly, flowing water table acquisition device 10 is suitable for obtaining the flowing water table that record has the data record with timing,
In, the corresponding data record of a line of the flowing water table, the one of the flowing water table arranges a corresponding field.
As an example, the flowing water table can be bank transaction flowing water table；Alternatively, the flowing water table can be internet use
Family behavior record table.
Temporal aspect generating means 20 are suitable for for the temporal aspect for needing to count, first to the corresponding word in the flowing water table
The single order of Duan Jinhang coarseness polymerize and saves single order polymerization result, then obtains the stream based on the single order polymerization result saved
Every data in water meter records the temporal aspect of corresponding machine learning sample.
As an example, temporal aspect generating means 20 may be adapted to respectively for each user identifier word in the flowing water table
Segment value, in the flowing water table with the user identification field value and time field value belongs to the number of each coarseness period
Carry out single order polymerization respectively according to the field value of the respective field of record to obtain single order polymerization result.
As an example, temporal aspect generating means 20 may be adapted to be related to only when the temporal aspect in the flowing water table
When the respective field of certain types of data record is counted, respectively for each user identifier in the flowing water table
Field value belongs to each coarseness period and symbol with the user identification field value, time field value in the flowing water table
The field value for closing the respective field of the certain types of data record carries out single order polymerization respectively to obtain single order polymerization
As a result.
As an example, temporal aspect generating means 20 may be adapted to record for every data in the flowing water table respectively,
The corresponding coarseness period is belonged to and records the corresponding period with the data and corresponding user identification field value
It carries out second order with the identical single order polymerization result of data record to polymerize, and using obtained second order polymerization result as this
The temporal aspect of the corresponding machine learning sample of data record, wherein every data records the corresponding period are as follows: at this
The period of specific duration before the time field value of data record.
As an example, the polymerization methods of the single order polymerization may include at least one among following item: summing, ask flat
, it is maximized, is minimized, seek standard deviation, count；The polymerization methods of second order polymerization may include among following item extremely
One item missing: it sums, be averaging, being maximized, being minimized, seek standard deviation, count.
As an example, the system of the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
It may also include that configuration interface provides device (not shown), content acquisition unit (not shown) and field determining device (not shown).
Particularly, configuration interface provides device suitable for providing a user root feature operator configuration interface.
Content acquisition unit is suitable for obtaining the Gent sign that user configures interface configuration by described feature operator.
Field determining device is suitable for determining the respective field for being directed to the temporal aspect based on Gent sign.
It is also adapted for providing a user in following Aggregation Operator configuration interface extremely as an example, configuration interface provides device
It is one few: the configuration interface for configuring the time window of the temporal aspect, the polymerization side for configuring the temporal aspect
The configuration interface of formula, the configuration interface for configuring the polymerization methods that the single order polymerize, stroke for configuring the coarseness
The configuration interface for the mode of dividing.Content acquisition unit is also adapted for acquisition user and configures the interior of interface configuration by the Aggregation Operator
Hold.Temporal aspect generating means 20 may be adapted to execute to obtain based on the content record pair of every data in the flowing water table
The temporal aspect for the machine learning sample answered.
As an example, temporal aspect generating means 20 are also adapted for for other temporal aspects for needing to count, difference needle
To every data record in the flowing water table, time field value is belonged to and records the corresponding period with the data and uses
Family identification field values and the field corresponding with other described temporal aspects for the identical data record that the data records are not
Quantity with field value is counted, and records the spy of other timing described in corresponding machine learning sample to obtain the data
Sign.
As an example, the system of the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
The method that the feature formation efficiency for improving machine learning sample can be executed by executing feature generation script.
It should be understood that the system of the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Specific implementation may be incorporated by reference the related specific implementation that Fig. 1 and Fig. 2 are described to realize, details are not described herein.
Included by the system of the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Device can be individually configured to execute any combination of the software of specific function, hardware, firmware or above-mentioned item.For example, these are filled
It sets and can correspond to dedicated integrated circuit, can also correspond to pure software code, also correspond to software and combined with hardware
Module.In addition, the one or more functions that these devices are realized can also be by physical entity equipment (for example, processor, client
End or server etc.) in component seek unity of action.
It should be understood that the method for the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample can
It is realized by the program being recorded in computer-readable media, for example, an exemplary embodiment of the present invention, it is possible to provide a kind of
Improve the computer-readable medium of the feature formation efficiency of machine learning sample, wherein remember on the computer-readable medium
Record has the computer program for executing following methods step: the flowing water table that record has the data record with timing is obtained,
In, the corresponding data record of a line of the flowing water table, the one of the flowing water table arranges a corresponding field；For needing to count
Temporal aspect, first in the flowing water table respective field carry out coarseness single order polymerize and save single order polymerization result,
Every data in the flowing water table is obtained based on the single order polymerization result saved again and records corresponding machine learning sample
The temporal aspect.
Computer program in above-mentioned computer-readable medium can be in client, host, agent apparatus, server etc.
Run in the environment disposed in computer equipment, it should be noted that the computer program can also be used in execute in addition to above-mentioned steps with
Outer additional step or execute when executing above-mentioned steps more specifically handles, these additional steps and is further processed
Content is described referring to Figures 1 and 2, here in order to avoid repetition will be repeated no longer.
It should be noted that the system of the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample can
The operation of computer program is completely dependent on to realize corresponding function, that is, in the function structure of each device and computer program
It is corresponding to each step, so that whole system is called by special software package (for example, the library lib), to realize corresponding function
Energy.
On the other hand, the system of the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Included each device can also be realized by hardware, software, firmware, middleware, microcode or any combination thereof.When with
When software, firmware, middleware or microcode are realized, program code or code segment for executing corresponding operating be can store
In the computer-readable medium of such as storage medium, so that processor can be by reading and running corresponding program code or generation
Code section executes corresponding operation.
For example, exemplary embodiment of the present invention is also implemented as computing device, which includes storage unit
And processor, set of computer-executable instructions conjunction is stored in storage unit, when the set of computer-executable instructions is closed by institute
When stating processor execution, the method for realizing the feature formation efficiency for improving machine learning sample is executed.
Particularly, the computing device can be deployed in server or client, can also be deployed in distributed network
On node apparatus in network environment.In addition, the computing device can be PC computer, board device, personal digital assistant, intelligence
Energy mobile phone, web are applied or other are able to carry out the device of above-metioned instruction set.
Here, the computing device is not necessarily single computing device, can also be it is any can be alone or in combination
Execute the device of above-metioned instruction (or instruction set) or the aggregate of circuit.Computing device can also be integrated control system or system
A part of manager, or can be configured to Local or Remote (for example, via wireless transmission) with the portable of interface inter-link
Formula electronic device.
In the computing device, processor may include central processing unit (CPU), graphics processor (GPU), may be programmed and patrol
Collect device, dedicated processor systems, microcontroller or microprocessor.As an example, not a limit, processor may also include simulation
Processor, digital processing unit, microprocessor, multi-core processor, processor array, network processing unit etc..
Described in the method for the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Certain operations can realize that certain operations can be realized by hardware mode by software mode, in addition, can also be by soft or hard
Part in conjunction with mode realize these operations.
Processor can run the instruction being stored in one of storage unit or code, wherein the storage unit can be with
Storing data.Instruction and data can be also sent and received via Network Interface Unit and by network, wherein the network connects
Any of transport protocol can be used in mouth device.
Storage unit can be integral to the processor and be integrated, for example, RAM or flash memory are arranged in integrated circuit microprocessor etc.
Within.In addition, storage unit may include independent device, such as, external dish driving, storage array or any Database Systems can
Other storage devices used.Storage unit and processor can be coupled operationally, or can for example by the port I/O,
Network connection etc. communicates with each other, and enables a processor to read the file being stored in storage unit.
In addition, the computing device may also include video display (such as, liquid crystal display) and user's interactive interface is (all
Such as, keyboard, mouse, touch input device etc.).The all components of computing device can be connected to each other via bus and/or network.
Involved in the method for the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Operation can be described as the functional block or function diagram of various interconnections or coupling.However, these functional blocks or function diagram can quilts
It is equably integrated into single logic device or is operated according to non-exact boundary.
For example, as described above, the feature formation efficiency according to an exemplary embodiment of the present invention for improving machine learning sample
Computing device may include storage unit and processor, wherein be stored in storage unit set of computer-executable instructions conjunction, when
When the set of computer-executable instructions is closed by processor execution, execute following step: obtaining record has with timing
The flowing water table of data record, wherein the corresponding data record of a line of the flowing water table, a column corresponding one of the flowing water table
A field；For the temporal aspect that needs count, the single order for first carrying out coarseness to the respective field in the flowing water table polymerize
And single order polymerization result is saved, then the record pair of every data in the flowing water table is obtained based on the single order polymerization result saved
The temporal aspect for the machine learning sample answered.
The foregoing describe each exemplary embodiments of the invention, it should be appreciated that foregoing description is merely exemplary, and exhaustive
Property, the present invention is not limited to disclosed each exemplary embodiments.Without departing from the scope and spirit of the invention, right
Many modifications and changes are obvious for those skilled in the art.Therefore, protection of the invention
Range should be subject to the scope of the claims.
Claims (10)
1. a kind of method for the feature formation efficiency for improving machine learning sample, wherein the described method includes:
Obtain the flowing water table that record has the data record with timing, wherein the corresponding data note of a line of the flowing water table
Record, the one of the flowing water table arrange a corresponding field；
For the temporal aspect that needs count, the single order for first carrying out coarseness to the respective field in the flowing water table polymerize and protects
Deposit single order polymerization result, then based on the single order polymerization result saved obtain every data in the flowing water table record it is corresponding
The temporal aspect of machine learning sample.
2. the method for claim 1, wherein
The flowing water table is bank transaction flowing water table；
Alternatively, the flowing water table is Internet user's behavior record table.
3. the method for claim 1, wherein the method also includes:
Provide a user root feature operator configuration interface；
It obtains user and configures the Gent sign of interface configuration by described feature operator, and determined based on Gent sign and be directed to institute
State the respective field of temporal aspect.
4. the method for claim 1, wherein the method also includes:
Provide a user at least one of following Aggregation Operator configuration interface: for configuring the time window of the temporal aspect
Configuration interface, polymerization methods for configuring the temporal aspect configuration interface, for configuring the poly- of the single order polymerization
The configuration interface of the configuration interface of conjunction mode, division mode for configuring the coarseness；
The content that user configures interface configuration by the Aggregation Operator is obtained,
Wherein, it executes to obtain every data in the flowing water table based on the content and records corresponding machine learning sample
The step of temporal aspect.
5. being the method for claim 1, wherein directed to the temporal aspect for needing to count, the respective field in convection current water meter
Carry out coarseness single order polymerization the step of include:
Respectively for each user identification field value in the flowing water table, to having the user identification field in the flowing water table
The field value that value and time field value belong to the respective field of the data record of each coarseness period carries out one respectively
Rank polymerize to obtain single order polymerization result.
6. method as claimed in claim 5, wherein the respective field for the temporal aspect that counts of needs, in convection current water meter
Carry out coarseness single order polymerization the step of include:
When the temporal aspect is related to only carrying out the respective field of the certain types of data record in the flowing water table
When statistics, respectively for each user identification field value in the flowing water table, to having the user identifier in the flowing water table
Field value, time field value belong to each coarseness period and meet the described corresponding of the certain types of data record
The field value of field carries out single order polymerization respectively to obtain single order polymerization result.
7. such as method described in claim 5 or 6, wherein poly- based on the single order saved for the temporal aspect that needs count
It closes result and obtains the step of every data in the flowing water table records the temporal aspect of corresponding machine learning sample packet
It includes:
Respectively for every data record in the flowing water table, the corresponding coarseness period is belonged to and is remembered with the data
It records the corresponding period and the identical single order polymerization result of corresponding user identification field value and data record carries out
Second order polymerize, and the timing spy of corresponding machine learning sample is recorded using obtained second order polymerization result as the data
Sign,
Wherein, every data records the corresponding period are as follows: the specific duration before the time field value of data record
Period.
8. a kind of system for the feature formation efficiency for improving machine learning sample, wherein the system comprises:
Flowing water table acquisition device, the flowing water table for having the data record with timing suitable for obtaining record, wherein the flowing water table
The corresponding data record of a line, the one of the flowing water table arrange a corresponding field；
Temporal aspect generating means, suitable for for the temporal aspect that counts is needed, first to the respective field in the flowing water table into
The single order of row coarseness polymerize and saves single order polymerization result, then obtains the flowing water table based on the single order polymerization result saved
In every data record the temporal aspect of corresponding machine learning sample.
9. a kind of system including at least one computing device He the storage device of at least one store instruction, wherein the finger
It enables when being run by least one described computing device, at least one described computing device is promoted to execute as in claim 1 to 7
Any claim described in raising machine learning sample feature formation efficiency method.
10. a kind of computer readable storage medium of store instruction, wherein when described instruction is run by least one computing device
When, promote at least one described computing device to execute the raising engineering as described in any claim in claim 1 to 7
The method for practising the feature formation efficiency of sample.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910578459.2A CN110275903A (en) | 2019-06-28 | 2019-06-28 | Improve the method and system of the feature formation efficiency of machine learning sample |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910578459.2A CN110275903A (en) | 2019-06-28 | 2019-06-28 | Improve the method and system of the feature formation efficiency of machine learning sample |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110275903A true CN110275903A (en) | 2019-09-24 |
Family
ID=67963104
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910578459.2A Pending CN110275903A (en) | 2019-06-28 | 2019-06-28 | Improve the method and system of the feature formation efficiency of machine learning sample |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110275903A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110955659A (en) * | 2019-11-28 | 2020-04-03 | 第四范式（北京）技术有限公司 | Method and system for processing data table |
| CN111143433A (en) * | 2019-12-10 | 2020-05-12 | 中国平安财产保险股份有限公司 | Method and device for counting data of data bins |
| CN113806354A (en) * | 2020-06-12 | 2021-12-17 | 第四范式（北京）技术有限公司 | Method and device for realizing time sequence feature extraction |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB0912846D0 (en) * | 2009-07-24 | 2009-08-26 | Hewlett Packard Development Co | Deduplication of encoded data |
| CN103605651A (en) * | 2013-08-28 | 2014-02-26 | 杭州顺网科技股份有限公司 | Data processing showing method based on on-line analytical processing (OLAP) multi-dimensional analysis |
| CN107766413A (en) * | 2017-09-05 | 2018-03-06 | 珠海宇能云企科技有限公司 | A kind of implementation method of real-time stream aggregate query |
| CN109284833A (en) * | 2018-08-22 | 2019-01-29 | 中国平安人寿保险股份有限公司 | Method, equipment and the storage medium of characteristic are obtained for machine learning model |
-
2019
- 2019-06-28 CN CN201910578459.2A patent/CN110275903A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB0912846D0 (en) * | 2009-07-24 | 2009-08-26 | Hewlett Packard Development Co | Deduplication of encoded data |
| CN103605651A (en) * | 2013-08-28 | 2014-02-26 | 杭州顺网科技股份有限公司 | Data processing showing method based on on-line analytical processing (OLAP) multi-dimensional analysis |
| CN107766413A (en) * | 2017-09-05 | 2018-03-06 | 珠海宇能云企科技有限公司 | A kind of implementation method of real-time stream aggregate query |
| CN109284833A (en) * | 2018-08-22 | 2019-01-29 | 中国平安人寿保险股份有限公司 | Method, equipment and the storage medium of characteristic are obtained for machine learning model |
Non-Patent Citations (1)
| Title |
|---|
| 徐雪琪: "《基于统计视角的数据挖掘研究》", 30 December 2010 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110955659A (en) * | 2019-11-28 | 2020-04-03 | 第四范式（北京）技术有限公司 | Method and system for processing data table |
| CN110955659B (en) * | 2019-11-28 | 2023-02-21 | 第四范式（北京）技术有限公司 | Method and system for processing data table |
| CN111143433A (en) * | 2019-12-10 | 2020-05-12 | 中国平安财产保险股份有限公司 | Method and device for counting data of data bins |
| CN113806354A (en) * | 2020-06-12 | 2021-12-17 | 第四范式（北京）技术有限公司 | Method and device for realizing time sequence feature extraction |
| CN113806354B (en) * | 2020-06-12 | 2023-12-22 | 第四范式（北京）技术有限公司 | Method and device for realizing time sequence feature extraction |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11416456B2 (en) | Method, apparatus, and computer program product for data quality analysis | |
| EP3537325B1 (en) | Interactive user interfaces | |
| US20190012683A1 (en) | Method for predicting purchase probability based on behavior sequence of user and apparatus for the same | |
| CN107424069A (en) | A kind of generation method of air control feature, risk monitoring and control method and apparatus | |
| CN107909178B (en) | Electronic device, loss of association repair rate prediction method, and computer-readable storage medium | |
| CN110275903A (en) | Improve the method and system of the feature formation efficiency of machine learning sample | |
| US20170351821A1 (en) | System and method for dynamic healthcare insurance claims decision support | |
| CN110502516A (en) | List data analytic method, device, computer equipment and storage medium | |
| US10977266B2 (en) | Ad-hoc analytical query of graph data | |
| US20210089331A1 (en) | Machine-learning models applied to interaction data for determining interaction goals and facilitating experience-based modifications to interface elements in online environments | |
| CN107609890A (en) | A kind of method and apparatus of order tracking | |
| US11775412B2 (en) | Machine learning models applied to interaction data for facilitating modifications to online environments | |
| CN107862425B (en) | Wind control data acquisition method, device and system and readable storage medium | |
| WO2014159296A1 (en) | A method, system and program product for matching of transaction records | |
| WO2022083093A1 (en) | Probability calculation method and apparatus in graph, computer device and storage medium | |
| CN109189931A (en) | A kind of screening technique and device of object statement | |
| CN109636497A (en) | A kind of Products Show method and apparatus | |
| CN108647064A (en) | The method and device of courses of action navigation | |
| CN111737922A (en) | Data processing method, device, equipment and medium based on recurrent neural network | |
| CN108197002A (en) | Mobile equipment is without burying point data statistical method, system, terminal and medium | |
| CN107239387A (en) | A kind of data exception detection method and terminal | |
| CN105574091B (en) | Information-pushing method and device | |
| US20220114607A1 (en) | Method, apparatus and computer readable storage medium for data processing | |
| CN108461127A (en) | Medical data relationship image acquiring method, device, terminal device and storage medium | |
| CN104636422A (en) | Method and system for mining of patterns in a data set |
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 |