CN108921581A - A kind of brush single operation recognition methods, device and computer readable storage medium - Google Patents
A kind of brush single operation recognition methods, device and computer readable storage medium Download PDFInfo
- Publication number
- CN108921581A CN108921581A CN201810792415.5A CN201810792415A CN108921581A CN 108921581 A CN108921581 A CN 108921581A CN 201810792415 A CN201810792415 A CN 201810792415A CN 108921581 A CN108921581 A CN 108921581A
- Authority
- CN
- China
- Prior art keywords
- refresh operation
- feature
- sample
- orders
- single operation
- 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.)
- Granted
Links
Classifications
-
- 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
- G06Q30/00—Commerce
- G06Q30/018—Certifying business or products
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
- G06F18/232—Non-hierarchical techniques
- G06F18/2321—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions
- G06F18/23213—Non-hierarchical techniques using statistics or function optimisation, e.g. modelling of probability density functions with fixed number of clusters, e.g. K-means clustering
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Entrepreneurship & Innovation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Probability & Statistics with Applications (AREA)
- Bioinformatics & Computational Biology (AREA)
- Accounting & Taxation (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Finance (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
- Image Analysis (AREA)
Abstract
The application provides a kind of brush single operation recognition methods, device and computer readable storage medium, method：The refresh operation information of dispatching person in the period for meeting preset condition is obtained, the period for meeting preset condition includes the adjacent period robbed between single operation twice during competition for orders；The adjacent refresh operation feature for robbing rear competition for orders operation in single operation twice according to the refresh operation acquisition of information；The refresh operation feature is input in advance trained brush single operation identification model, obtain it is described after competition for orders operation whether be brush single operation recognition result.The application can influence the problem of brush single operation identifies to avoid due to adulterating the operation of other orders in the single order processing period, it can guarantee the accuracy of brush single operation identification, and foundation can be provided for the supervision and control of subsequent brush single act, and then promote the experience of dispatching person and user.
Description
Technical field
This application involves technical field of data processing more particularly to a kind of brush single operation recognition methods, device and computer
Readable storage medium storing program for executing.
Background technique
It is related in the business dispensed under line at present, certain dispatching persons are to obtain more high yield, externally hung software brush list can be utilized,
To obtain high-quality order (for example, order that dispatching distance is short, dispatching difficulty is small or dispatching unit price is high).This unfair competition
Behavior will lead to the number of long-tail order to be processed (for example, order that dispatching distance is long, dispatching difficulty is big or dispatching unit price is low)
Amount increases, and order is caused to overstock, and dispatching business efficiency reduces, and influences the experience of dispatching person and user.
Summary of the invention
It, can be in view of this, the application provides a kind of brush single operation recognition methods, device and computer readable storage medium
It accurately identifies brush single operation, promotes the experience of dispatching person and user.
Specifically, the application is achieved by the following technical solution：
According to a first aspect of the present application, a kind of brush single operation recognition methods is proposed, including：
Obtain the refresh operation information of dispatching person in the period for meeting preset condition, the time for meeting preset condition
Section includes the adjacent period robbed between single operation twice during competition for orders；
The adjacent refreshing behaviour for robbing rear competition for orders operation in single operation twice according to the refresh operation acquisition of information
Make feature；
The refresh operation feature is input in advance trained brush single operation identification model, obtain it is described after once rob
Single operation whether be brush single operation recognition result, the brush single operation identification model be dispensed according to sample in target time section
What a plurality of sample refresh operation information of member was trained.
In one embodiment, the refresh operation information for obtaining dispatching person in the period for meeting preset condition, including：
If detecting competition for orders operation after adjacent robbed in single operation twice, acquisition met in the period of preset condition
The refresh operation information of dispatching person.
In one embodiment, the brush single operation identification model is trained in advance according to following steps：
The a plurality of sample refresh operation information of sample dispatching person in target time section is obtained, the target time section includes robbing
The adjacent period robbed between single operation twice during list；
A sample robs the sample refresh operation of single operation after according to every sample refresh operation acquisition of information
Feature obtains multiple sample refresh operation features of a plurality of sample refresh operation information；
Classify to the multiple sample refresh operation feature, obtains refreshing behaviour corresponding to the positive sample of non-brush single operation
Make feature and the negative sample refresh operation feature corresponding to brush single operation；
Trained brush single operation is treated using the positive sample refresh operation feature and the negative sample refresh operation feature
Identification model is trained.
In one embodiment, described to classify to the multiple sample refresh operation feature, it obtains corresponding to non-brush list
The positive sample refresh operation feature of operation and negative sample refresh operation feature corresponding to brush single operation, including：
Using preset unsupervised algorithm, behaviour is refreshed to the sample of the first quantity in the multiple sample refresh operation feature
Classify as feature, the positive sample refresh operation obtained in the sample refresh operation feature for characterizing first quantity is special
First classification results of negative sample refresh operation feature of seeking peace；
Based on first classification results, using preset semi-supervised algorithm, to the multiple sample refresh operation feature
In the sample refresh operation feature of the second quantity classify, the sample refresh operation obtained for characterizing second quantity is special
Second classification results of positive sample refresh operation feature and negative sample refresh operation feature in sign；
Wherein, the sample refresh operation feature of second quantity is in the multiple sample refresh operation feature except described
Feature except the sample refresh operation feature of first quantity, and first quantity is less than second quantity.
In one embodiment, the method also includes：
To the positive sample refresh operation feature or negative sample refresh operation of category division mistake in first classification results
Feature is corrected, the first classification results after being corrected；
It is executed based on the first classification results after the correction described in the multiple sample refresh operation feature second
The operation that the sample refresh operation feature of quantity is classified.
In one embodiment, the sample refresh operation feature includes at least one of following：
In the duration of refresh operation number, the target time section in the target time section, the target time section
Time interval between last time refresh operation and the rear competition for orders operation, the refresh operation in the target time section
Update in frequency, the target time section refreshes than the pressure value of terminal device screen, the mesh in, the target time section
Mark the sliding number of terminal device screen in the period；
The update refreshes than for characterizing first number and second several ratio, first number is the target
The number of new order is obtained by refresh operation in period, second number is the refresh operation in the target time section
Number.
In one embodiment, the dimension of the sample refresh operation feature is higher than three-dimensional；
It is described to treat trained brush list using the positive sample refresh operation feature and the negative sample refresh operation feature
Before operation identification model is trained, the method also includes：
PCA dimension-reduction treatment is carried out to the positive sample refresh operation feature and the negative sample refresh operation feature respectively,
It obtains the corresponding positive sample three-dimensional feature of the positive sample refresh operation feature and the negative sample refresh operation feature is corresponding
Negative sample three-dimensional feature；
The positive sample three-dimensional feature and the negative sample three-dimensional feature are projected into preset three-dimensional space, obtained described
The distribution information of positive sample three-dimensional feature and the negative sample three-dimensional feature, the distribution information are described for characterizing
The degree of closeness of positive sample three-dimensional feature and the negative sample three-dimensional feature in the three-dimensional space；
According to the degree of closeness of the positive sample three-dimensional feature and the negative sample three-dimensional feature in the three-dimensional space
Determine brush single operation identification model to be trained.
In one embodiment, the method also includes：
If competition for orders operation is brush single operation after recognition result characterization is described, forbid responding it is described after once rob
Single operation；
If competition for orders operation is not brush single operation after the recognition result characterization is described, allow to respond described rear primary
Rob single operation.
According to a second aspect of the present application, a kind of brush single operation identification device is proposed, including：
Operation information acquisition module, for obtaining the refresh operation information for meeting dispatching person in the period of preset condition,
The period for meeting preset condition includes the adjacent period robbed between single operation twice during competition for orders；
Operating characteristics obtain module, for according to the refresh operation acquisition of information it is adjacent rob in single operation twice after
Once rob the refresh operation feature of single operation；
Recognition result obtains module, identifies mould for the refresh operation feature to be input to brush single operation trained in advance
In type, obtain it is described after competition for orders operation whether be brush single operation recognition result, the brush single operation identification model is root
It is trained according to a plurality of sample refresh operation information of sample dispatching person in target time section.
In one embodiment, the operation information acquisition module be also used to when detect it is adjacent rob twice it is latter in single operation
It is secondary when robbing single operation, obtain the refresh operation information of dispatching person in the period for meeting preset condition.
In one embodiment, described device further includes identification model training module, for being trained in advance according to following steps
Brush single operation identification model；
The identification model training module includes：
Sample information acquiring unit, a plurality of sample refresh operation for obtaining sample dispatching person in target time section are believed
Breath, the target time section includes the adjacent period robbed between single operation twice during competition for orders；
Sample characteristics acquiring unit is robbed for a rear sample according to every sample refresh operation acquisition of information
The sample refresh operation feature of single operation obtains multiple sample refresh operation features of a plurality of sample refresh operation information；
Sample characteristics taxon obtains corresponding to non-for classifying to the multiple sample refresh operation feature
The positive sample refresh operation feature of brush single operation and negative sample refresh operation feature corresponding to brush single operation；
Identification model training unit, for special using the positive sample refresh operation feature and the negative sample refresh operation
Sign is treated trained brush single operation identification model and is trained.
In one embodiment, the sample characteristics taxon is also used to：
Using preset unsupervised algorithm, behaviour is refreshed to the sample of the first quantity in the multiple sample refresh operation feature
Classify as feature, the positive sample refresh operation obtained in the sample refresh operation feature for characterizing first quantity is special
First classification results of negative sample refresh operation feature of seeking peace；
Based on first classification results, using preset semi-supervised algorithm, to the multiple sample refresh operation feature
In the sample refresh operation feature of the second quantity classify, the sample refresh operation obtained for characterizing second quantity is special
Second classification results of positive sample refresh operation feature and negative sample refresh operation feature in sign；
Wherein, the sample refresh operation feature of second quantity is in the multiple sample refresh operation feature except described
Feature except the sample refresh operation feature of first quantity, and first quantity is less than second quantity.
In one embodiment, the sample characteristics taxon is also used to：
To the positive sample refresh operation feature or negative sample refresh operation of category division mistake in first classification results
Feature is corrected, the first classification results after being corrected；
It is executed based on the first classification results after the correction described in the multiple sample refresh operation feature second
The operation that the sample refresh operation feature of quantity is classified.
In one embodiment, the sample refresh operation feature includes at least one of following：
In the duration of refresh operation number, the target time section in the target time section, the target time section
Time interval between last time refresh operation and the rear competition for orders operation, the refresh operation in the target time section
Update in frequency, the target time section refreshes than the pressure value of terminal device screen, the mesh in, the target time section
Mark the sliding number of terminal device screen in the period；
The update refreshes than for characterizing first number and second several ratio, first number is the target
The number of new order is obtained by refresh operation in period, second number is the refresh operation in the target time section
Number.
In one embodiment, the dimension of the sample refresh operation feature is higher than three-dimensional；
Described device further includes：Identification model determining module；
The identification model determining module further includes：
Sample characteristics dimensionality reduction unit, for respectively to the positive sample refresh operation feature and the negative sample refresh operation
Feature carries out PCA dimension-reduction treatment, obtains the corresponding positive sample three-dimensional feature of the positive sample refresh operation feature and the negative sample
The corresponding negative sample three-dimensional feature of this refresh operation feature；
State information acquisition unit, it is pre- for projecting to the positive sample three-dimensional feature and the negative sample three-dimensional feature
If three-dimensional space, obtain the distribution information of the positive sample three-dimensional feature and the negative sample three-dimensional feature, described point
Cloth status information is for characterizing the positive sample three-dimensional feature and negative sample three-dimensional feature the connecing in the three-dimensional space
Short range degree；
Identification model determination unit is used for according to the positive sample three-dimensional feature and the negative sample three-dimensional feature described
Degree of closeness in three-dimensional space determines brush single operation identification model to be trained.
Described device further includes：Competition for orders operation processing module；
The competition for orders operation processing module, including：
Attendant exclusion response unit is brush single operation for competition for orders to operate after recognition result characterization is described
When, forbid responding the rear competition for orders operation；
Operation allows response unit, is not brush single operation for competition for orders to operate after recognition result characterization is described
When, allow to respond the rear competition for orders operation.
According to the third aspect of the application, a kind of computer readable storage medium is proposed, the storage medium is stored with
Computer program, the computer program are used to execute any of the above-described brush single operation recognition methods.
The application meets the refresh operation information of dispatching person in the period of preset condition by acquisition, and according to the brush
Described in new operation information acquisition it is adjacent robbed in single operation twice after competition for orders operation refresh operation feature, and then by the brush
New operating characteristics are input in brush single operation identification model trained in advance, obtain whether the rear competition for orders operation is that brush is single
The recognition result of operation can influence brush single operation to avoid due to adulterating the operation of other orders in the single order processing period
The problem of identification, it is ensured that the accuracy of brush single operation identification, and can be provided for the supervision and control of subsequent brush single act
Foundation, and then promote the experience of dispatching person and user.
Detailed description of the invention
Figure 1A is a kind of flow chart of brush single operation recognition methods shown in one exemplary embodiment of the application；
Fig. 1 C is the schematic diagram of the period for meeting preset condition shown in one exemplary embodiment of the application；
Figure 1B is the schematic diagram of the order processing process shown in one exemplary embodiment of the application；
Fig. 2 is a kind of flow chart of brush single operation recognition methods shown in the application another exemplary embodiment；
Fig. 3 A is how to divide the multiple sample refresh operation feature shown in one exemplary embodiment of the application
The flow chart of class；
Fig. 3 B is the effect classified to multiple sample refresh operation features shown in one exemplary embodiment of the application
Schematic diagram；
Fig. 4 is how to divide the multiple sample refresh operation feature shown in the application another exemplary embodiment
The flow chart of class；
Fig. 5 is a kind of flow chart of brush single operation recognition methods shown in the application another exemplary embodiment；
Fig. 6 is a kind of flow chart of brush single operation recognition methods shown in the application another exemplary embodiment；
Fig. 7 is a kind of structural block diagram of brush single operation identification device shown in one exemplary embodiment of the application；
Fig. 8 is a kind of structural block diagram of brush single operation identification device shown in the application another exemplary embodiment.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all embodiments consistent with the application.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the application.
It is only to be not intended to be limiting the application merely for for the purpose of describing particular embodiments in term used in this application.
It is also intended in the application and the "an" of singular used in the attached claims, " described " and "the" including majority
Form, unless the context clearly indicates other meaning.It is also understood that term "and/or" used herein refers to and wraps
It may be combined containing one or more associated any or all of project listed.
It will be appreciated that though various information, but this may be described using term first, second, third, etc. in the application
A little information should not necessarily be limited by these terms.These terms are only used to for same type of information being distinguished from each other out.For example, not departing from
In the case where the application range, the first information can also be referred to as the second information, and similarly, the second information can also be referred to as
One information.Depending on context, word as used in this " if " can be construed to " ... when " or " when ...
When " or " in response to determination ".
Figure 1A is a kind of flow chart of brush single operation recognition methods shown in one exemplary embodiment of the application；The embodiment
It can be used for server-side (for example, server cluster etc. of a server and multiple servers composition).As shown in Figure 1A, the party
Method includes step S101-S103：
S101：Obtain the refresh operation information of dispatching person in the period for meeting preset condition.
In one embodiment, above-mentioned refresh operation information can set to obtain order to be dispensed in terminal for dispatching person
The information of the refresh operation executed in standby client.
In one embodiment, the period for meeting preset condition includes that above-mentioned same dispatching person holds during competition for orders
The capable adjacent period robbed between single operation twice.
It is worth noting that, it is generally the case that the order processing process in dispatching person's client substantially has following step
Suddenly：1, refresh, 2, check details page, 3, check Distribution path (second level details page), 4, competition for orders, 5, arrive shop, 6, be sent to.And at this
In several steps, refresh, check details page and check that Distribution path operation can exist several times in an order period, and
Competition for orders can only operate once to shop with each order of operation is sent to.Therefore, in order to formulate brush single operation identifying schemes, primary
Task is exactly the research object of determining scheme.
On surface apparently, the research object of scheme should be each order, but actually if using order as research pair
As due to that may include the operation for other orders in an order period, and the operation of dispatching person may also
Corresponding multiple orders, the then data that will lead to an order dimension adulterate the interference information of other orders.Figure 1B is the application one
The schematic diagram of order processing process shown in exemplary embodiment；As shown in Figure 1B, dispatching person, which first refreshes, obtains order A, B, C,
Then it executes and robs order A operation, and refresh and obtained order D, E, and execute and rob order D operation, arrive shop later and take order
A…….As can be seen that being doped into the operation of order D in the process cycle of order A, i.e., order D is doped in the data of order A
Interference information.
And if the data of an order dimension adulterate the interference information of other orders, it will affect the identification effect of the order
Fruit, therefore be directly unreasonable using the complete order period as research object.
In view of the above problem, inventor has found through overtesting, can use and meet matching in the period of preset condition
The refresh operation information for the person of sending is as research object, and it is possible to determine that this meets preset condition using following two mode
Period：
If first way robs between single operation that there are refresh operations adjacent twice, can be by during competition for orders adjacent two
The secondary period robbed between single operation is determined directly as meeting the period of preset condition.
For example, Fig. 1 C is the signal of the period for meeting preset condition shown in one exemplary embodiment of the application
Figure.As shown in Figure 1 C, if the adjacent period robbed between single operation twice is the competition for orders moment T1 of previous order A to latter order
Period S1 between the competition for orders moment T2 of B, then the period for meeting preset condition can be determined as period S1.
The second way will can rob twice if refresh operation is not present in adjacent rob twice between single operation comprising adjacent
Period between single operation, the length longer period is determined as meeting the period of preset condition.
For example, as shown in Figure 1 C, which can be determined as specified between the T1 moment
Moment T0 to the T2 moment period S2.
It is worth noting that the above-mentioned longer period can rule of thumb be set by developer, such as it is arranged
It is 5 minutes, 10 minutes etc., the present embodiment is to this without limiting.For example, if developer analyzes by mass data
Know, 90% or so the adjacent period robbed between single operation twice is smaller than five minutes, and adjacent greater than five minutes
Rob twice the period between single operation (such as between current first time brush single operation and the operation of the last time of the previous day
Period) in, the operation before five minutes of rear competition for orders operation, to the differentiation of brush single operation, there is no too big meanings, then
It can be determined as a period for meeting preset condition for five minutes.
In one embodiment, it can obtain and meet after detecting adjacent robbed in single operation twice when competition for orders operation
The refresh operation information of dispatching person in the period of preset condition, so can determined based on the refresh operation information it is described after
Once rob single operation be brush single operation when, single operation is robbed to this and is forbidden.
S102：The adjacent brush for robbing rear competition for orders operation in single operation twice according to the refresh operation acquisition of information
New operating characteristics.
In one embodiment, when obtain meet the refresh operation information of dispatching person in the period of preset condition after, can be with
The adjacent refresh operation feature for robbing rear competition for orders operation in single operation twice according to the refresh operation acquisition of information.
It is worth noting that the present embodiment be by study each order rob the refresh operation situation before single operation come pair
Determine that this robs whether single operation is brush single operation, thus for it is adjacent rob single operation twice for, rear competition for orders operation is only
Really identification object.
In one embodiment, above-mentioned refresh operation feature can be that can protrude corresponding refresh operation information well
Feature, and the feature that the difference with other refresh operation information can be shown and contacted, with Enhanced feature identification.
, can be according to a kind of feature of refresh operation acquisition of information in an optional embodiment, or obtain simultaneously
(quantity for obtaining feature is opposite with the type of sample characteristics that the identification model of subsequent training is obtained in training process for various features
Answer), the characteristics of sufficiently comprehensively to show the refresh operation information.
S103：The refresh operation feature is input in advance trained brush single operation identification model, obtain it is described after
Once rob single operation whether be brush single operation recognition result.
Wherein, the brush single operation identification model is to refresh behaviour according to a plurality of sample of sample dispatching person in target time section
It is trained as information.
In one embodiment, brush single operation identification model can be trained in advance, for example, can be by obtaining sample dispatching person
A plurality of sample refresh operation information, and feature extraction is carried out according to a plurality of sample refresh operation information, and then according to mentioning
The selected brush single operation identification model of feature and the calibration result training taken.In one embodiment, the brush single operation identification model
Training method may refer to display Fig. 2 embodiment, herein first without be described in detail.
In one embodiment, when according to the refresh operation acquisition of information it is adjacent rob in single operation twice after once rob
After the refresh operation feature of single operation, obtained refresh operation feature can be input to above-mentioned brush single operation trained in advance and known
In other model, with obtain it is described after competition for orders operation whether be brush single operation recognition result.
In one embodiment, it is brush single operation that above-mentioned recognition result, which can be used for characterizing the rear competition for orders operation, or
It is not brush single operation.
Seen from the above description, refreshing behaviour of the present embodiment for the period robbed twice during competition for orders between single operation
Make the corresponding refresh operation feature of acquisition of information, and then can accurately be identified based on brush single operation identification model trained in advance
Rob whether single operation is brush single operation, due to being with the refresh operation in the period between adjacent competition for orders twice for research pair
As, thus asking for brush single operation identification can be influenced to avoid due to adulterating the operation of other orders in the single order processing period
Topic, it is ensured that the accuracy of brush single operation identification, and foundation can be provided for the supervision and control of subsequent brush single act, in turn
Promote the experience of dispatching person and user.
Fig. 2 is a kind of flow chart of brush single operation recognition methods shown in the application another exemplary embodiment；The implementation
Example can be used for server-side (for example, server cluster etc. of a server and multiple servers composition).As shown in Fig. 2, should
Method includes step S201-S207：
S201：Obtain a plurality of sample refresh operation information of sample dispatching person in target time section.
In one embodiment, above-mentioned sample refresh operation information can for dispatching person in order to obtain order to be dispensed and at end
The information of the refresh operation executed in the client of end equipment.
In one embodiment, the target time section includes the adjacent time robbed between single operation twice during competition for orders
Section.
It can be with based on determining the similar reason for meeting the period of preset condition in Figure 1A illustrated embodiment, in the present embodiment
Using the sample refresh operation information of the dispatching person in target time section as research object, and it is possible to using following two
Mode determines the target time section：
If first way robs between single operation that there are refresh operations adjacent twice, can be by during competition for orders adjacent two
The secondary period robbed between single operation is determined directly as target time section.
The second way will can rob twice if refresh operation is not present in adjacent rob twice between single operation comprising adjacent
Period between single operation, the length longer period be determined as target time section.
In one embodiment, in order to ensure the accuracy of trained brush single operation identification model, available a plurality of sample
Refresh operation information, a plurality of sample refresh operation information can derive from multiple dispatching persons.
S202：The sample that a sample robs single operation after according to every sample refresh operation acquisition of information refreshes
Operating characteristics obtain multiple sample refresh operation features of a plurality of sample refresh operation information.
It in one embodiment, can after obtaining a plurality of sample refresh operation information of sample dispatching person in target time section
The sample refresh operation feature of single operation is robbed with a rear sample according to a plurality of sample refresh operation acquisition of information, with
To multiple sample refresh operation features of a plurality of sample refresh operation information.
In one embodiment, above-mentioned sample refresh operation information can refresh behaviour for that can protrude corresponding sample well
The characteristics of making information, and the difference with other sample refresh operation information and the feature contacted can be shown, it is distinguished with Enhanced feature
Knowledge and magnanimity.
It, can be according to a kind of feature of sample refresh operation acquisition of information, or simultaneously in an optional embodiment
Various features are obtained, the characteristics of sufficiently comprehensively to show the sample refresh operation information.
In one embodiment, above-mentioned sample refresh operation feature may include at least one of following：
In the duration of refresh operation number, the target time section in the target time section, the target time section
Time interval between last time refresh operation and the rear competition for orders operation, the refresh operation in the target time section
Update in frequency, the target time section refreshes than the pressure value of terminal device screen, the mesh in, the target time section
Mark the sliding number of terminal device screen in the period；
The update refreshes than for characterizing first number and second several ratio, first number is the target
The number of new order is obtained by refresh operation in period, second number is the refresh operation in the target time section
Number.
S203：Classify to the multiple sample refresh operation feature, obtains the positive sample corresponding to non-brush single operation
Refresh operation feature and negative sample refresh operation feature corresponding to brush single operation.
In one embodiment, after obtaining multiple sample refresh operation features of a plurality of sample refresh operation information,
It can classify to the multiple sample refresh operation feature, i.e., be refreshed by preset sorting algorithm from the multiple sample
Operating characteristics determine that the positive sample refresh operation feature for corresponding to non-brush single operation and the negative sample corresponding to brush single operation refresh
Operating characteristics.
It in one embodiment, can be by developer according to quantity, the classification difficulty etc. of actual sample refresh operation feature
Determine that suitable machine learning algorithm, such as unsupervised learning algorithm (such as k-means clustering algorithm), semi-supervised study are calculated
Method (e.g., K- nearest neighbor algorithm etc.) and other learning algorithms for having supervision, the present embodiment is to this without limiting.
In one embodiment, the above-mentioned mode classified to the multiple sample refresh operation feature is referring also under
Fig. 3 A illustrated embodiment is stated, herein first without being described in detail.
S204：Trained brush list is treated using the positive sample refresh operation feature and the negative sample refresh operation feature
Operation identification model is trained.
In one embodiment, classify when to the multiple sample refresh operation feature, obtain corresponding to non-brush singly behaviour
The positive sample refresh operation feature of work and corresponding to the negative sample refresh operation feature of brush single operation after, can use the positive sample
This refresh operation feature and the negative sample refresh operation feature, to predetermined brush single operation identification model to be trained into
Row training, obtains brush single operation identification model.
In one embodiment, the determination method of above-mentioned brush single operation identification model may refer to following embodiment illustrated in fig. 5,
Herein first without being described in detail.
S205：The refresh operation information of dispatching person in the period for meeting preset condition is obtained, it is described to meet preset condition
Period include the adjacent period robbed between single operation twice during competition for orders.
S206：The adjacent brush for robbing rear competition for orders operation in single operation twice according to the refresh operation acquisition of information
New operating characteristics.
S207：The refresh operation feature is input in advance trained brush single operation identification model, obtain it is described after
Once rob single operation whether be brush single operation recognition result.
Wherein, the relevant explanation explanation of step S205-S207 may refer to above-described embodiment, herein without repeating.
Seen from the above description, the present embodiment refreshes behaviour by obtaining a plurality of sample of sample dispatching person in target time section
Make information, and a rear sample according to every sample refresh operation acquisition of information robs the sample refresh operation of single operation
Feature obtains multiple sample refresh operation features of a plurality of sample refresh operation information, then refreshes to the multiple sample
Operating characteristics are classified, and are obtained corresponding to the positive sample refresh operation feature of non-brush single operation and corresponding to the negative of brush single operation
Sample refresh operation feature, and then instruction is treated using the positive sample refresh operation feature and the negative sample refresh operation feature
Experienced brush single operation identification model is trained, available brush single operation identification model, and then may be implemented subsequent based on this
Brush single operation identification model, which accurately identifies, robs whether single operation is brush single operation, due to be between adjacent competition for orders twice when
Between refresh operation in section be research object, thus can be to avoid due to the behaviour for adulterating other orders in the single order processing period
The problem of making and influencing brush single operation identification, it is ensured that the accuracy of brush single operation identification model is subsequent brush single act
Supervision and control provide foundation, and then promote the experience of dispatching person and user.
Fig. 3 A is how to divide the multiple sample refresh operation feature shown in one exemplary embodiment of the application
The flow chart of class；Fig. 3 B is the effect classified to multiple sample refresh operation features shown in one exemplary embodiment of the application
Fruit schematic diagram.
The present embodiment on the basis of the above embodiments, how to classify to the multiple sample refresh operation feature
For illustrate.
It is worth noting that in order to ensure the accuracy of trained brush single operation identification model, it may be necessary to obtain a large amount of
Sample refresh operation information, and then determine above the average age for marriage sample refresh operation feature.In the case, if only with unsupervised
Algorithm may not can guarantee the accuracy of the classification of sample refresh operation feature, so will affect subsequent training model it is accurate
Property.And if needing to expend a large amount of people only with there is supervision algorithm to classify a large amount of sample refresh operation information
Work cost.
In view of the above problem, inventor has found through overtesting, can mutually be tied using unsupervised algorithm with semi-supervised algorithm
The mode of conjunction classifies to a large amount of sample refresh operation information, such as first using unsupervised algorithm to a small amount of sample characteristics into
Row classification, then classification results based on a small amount of sample characteristics, using semi-supervised algorithm to remaining a large amount of sample characteristics into
Row classification, can not only guarantee classification accuracy, but also can save labour turnover.
Specifically, as shown in Figure 3A, classifying described in step S203 to the multiple sample refresh operation feature, obtaining
It, can to the positive sample refresh operation feature for corresponding to non-brush single operation and corresponding to the negative sample refresh operation feature of brush single operation
To include the following steps S301-S302：
S301：Using preset unsupervised algorithm, to the sample of the first quantity in the multiple sample refresh operation feature
Refresh operation feature is classified, and the positive sample obtained in the sample refresh operation feature for characterizing first quantity refreshes
First classification results of operating characteristics and negative sample refresh operation feature.
It in one embodiment, as shown in Figure 3B, can be from multiple sample after obtaining multiple sample refresh operation features
The sample refresh operation feature of the first quantity is obtained in refresh operation feature.Wherein, which can be multiple sample brushes
The smaller portions (i.e. small sample set) of the total amount of new operating characteristics.
In one embodiment, above-mentioned unsupervised algorithm can be selected by developer according to actual classification required precision
It takes, such as is chosen for K-means clustering algorithm, the present embodiment is to this without limiting.
In one embodiment, using above-mentioned preset unsupervised algorithm to the sample refresh operation feature of first quantity into
After row classification, positive sample refresh operation feature and negative sample can be marked off from the sample refresh operation feature of first quantity
Refresh operation feature (total amount of the positive sample refresh operation feature and negative sample refresh operation feature is first quantity).
S302：Based on first classification results, using preset semi-supervised algorithm, to the multiple sample refresh operation
The sample refresh operation feature of the second quantity is classified in feature, is obtained the sample for characterizing second quantity and is refreshed behaviour
Make the second classification results of the positive sample refresh operation feature and negative sample refresh operation feature in feature.
In one embodiment, the sample refresh operation feature of second quantity is the multiple sample refresh operation feature
In feature in addition to the sample refresh operation feature of first quantity, and first quantity is less than second quantity.
Specifically, when the sample refresh operation feature for obtaining the first quantity from above-mentioned multiple sample refresh operation features
Afterwards, the sample refresh operation feature of available remaining second quantity.Obviously, which can refresh for multiple samples
The major part (i.e. large sample collection) of the total amount of operating characteristics.
In one embodiment, above-mentioned semi-supervised algorithm can be selected by developer according to actual classification required precision
It takes, such as is chosen for Label Propagation algorithm, the present embodiment is to this without limiting.
In one embodiment, using above-mentioned preset semi-supervised algorithm to the sample refresh operation feature of second quantity into
After row classification, positive sample refresh operation feature and negative sample can be marked off from the sample refresh operation feature of second quantity
Refresh operation feature (total amount of the positive sample refresh operation feature and negative sample refresh operation feature is second quantity).
Classify in this way, realizing to the multiple sample refresh operation feature, obtains corresponding to non-brush single operation
Positive sample refresh operation feature and negative sample refresh operation feature corresponding to brush single operation.
Seen from the above description, the present embodiment refreshes the multiple sample and grasps by using preset unsupervised algorithm
The sample refresh operation feature for making the first quantity in feature is classified, and is obtained the sample for characterizing first quantity and is refreshed
First classification results of positive sample refresh operation feature and negative sample refresh operation feature in operating characteristics, and based on described the
One classification results, using preset semi-supervised algorithm, to the sample brush of the second quantity in the multiple sample refresh operation feature
New operating characteristics are classified, and the positive sample obtained in the sample refresh operation feature for characterizing second quantity refreshes behaviour
The second classification results for making feature and negative sample refresh operation feature may be implemented to carry out multiple sample refresh operation features quasi-
Really classification, due to being carried out in such a way that unsupervised algorithm and semi-supervised algorithm combine to a large amount of sample refresh operation information
Classification, thus can guarantee the classification accuracy of great amount of samples refresh operation feature, and can save labour turnover.
Fig. 4 is how to divide the multiple sample refresh operation feature shown in the application another exemplary embodiment
The flow chart of class.The present embodiment on the basis of the above embodiments, with how to the multiple sample refresh operation feature carry out
It is illustrated for classification.As shown in figure 4, being carried out described in step S203 to the multiple sample refresh operation feature
Classification obtains the negative sample refresh operation corresponding to the positive sample refresh operation feature of non-brush single operation and corresponding to brush single operation
Feature may comprise steps of S401-S403：
S401：Using preset unsupervised algorithm, to the sample of the first quantity in the multiple sample refresh operation feature
Refresh operation feature is classified, and the positive sample obtained in the sample refresh operation feature for characterizing first quantity refreshes
First classification results of operating characteristics and negative sample refresh operation feature.
S402：Positive sample refresh operation feature or negative sample to category division mistake in first classification results refresh
Operating characteristics are corrected, the first classification results after being corrected.
It in one embodiment, can be wrong to category division in first classification results after obtaining the first classification results
Positive sample refresh operation feature or negative sample refresh operation feature accidentally is corrected.
In one embodiment, as shown in Figure 3B, school can be carried out to first classification results using artificial selection mode
Just.For example, if in above-mentioned first classification results, sample refresh operation feature A is confirmed as positive sample refresh operation feature,
And learn that sample refresh operation feature A is practical by actual verification or analysis and should be negative sample refresh operation feature, thus can
Its classification results is corrected to " negative sample refresh operation feature " by " positive sample refresh operation feature ".
Similarly, if in above-mentioned first classification results, it is special that sample refresh operation feature B is confirmed as negative sample refresh operation
Sign, and pass through actual verification or analysis learns that sample refresh operation feature B is practical and should be positive sample refresh operation feature, thus
Its classification results can be corrected to " positive sample refresh operation feature " by " negative sample refresh operation feature ".
Thus, when special using positive sample refresh operation of the aforesaid way to category division mistake in first classification results
After sign or negative sample refresh operation feature are corrected, the first classification results after available correction, and then can be based on should
The first classification results after correction execute following step S402, and then obtain in the sample refresh operation feature of the second quantity just
Second classification results of sample refresh operation feature and negative sample refresh operation feature.
S402：Based on the first classification results after the correction, using preset semi-supervised algorithm, to the multiple sample
The sample refresh operation feature of the second quantity is classified in refresh operation feature, obtains the sample for characterizing second quantity
Second classification results of positive sample refresh operation feature and negative sample refresh operation feature in this refresh operation feature.
Wherein, step S401, the relevant explanation of S403 and explanation may refer to above-described embodiment, herein without repeating.
Seen from the above description, the present embodiment refreshes the multiple sample and grasps by using preset unsupervised algorithm
The sample refresh operation feature for making the first quantity in feature is classified, and obtains the first classification results, and tie to first classification
Fruit is corrected, the first classification results after being corrected, and then uses preset semi-supervised algorithm, to the multiple sample brush
The sample refresh operation feature of the second quantity is classified in new operating characteristics, is obtained the second classification results, be may be implemented to more
A sample refresh operation feature carries out Accurate classification, due in such a way that unsupervised algorithm and semi-supervised algorithm combine, and
And the classification results of unsupervised algorithm are corrected, thus can be improved and a large amount of sample refresh operation information is divided
The accuracy of class is further ensured that the classification accuracy of great amount of samples refresh operation feature.
Fig. 5 is a kind of flow chart of brush single operation recognition methods shown in the application another exemplary embodiment；The implementation
Example can be used for server-side (for example, server cluster etc. of a server and multiple servers composition).Fig. 5 is the application one
Drop shadow effect's schematic diagram of positive and negative samples refresh operation feature after dimensionality reduction shown in exemplary embodiment.
As shown in figure 5, the method comprising the steps of S501-S510：
S501：Obtain a plurality of sample refresh operation information of sample dispatching person in target time section, the target time section
Including the period robbed between single operation twice adjacent during competition for orders.
S502：The sample that a sample robs single operation after according to every sample refresh operation acquisition of information refreshes
Operating characteristics obtain multiple sample refresh operation features of a plurality of sample refresh operation information.
S503：Classify to the multiple sample refresh operation feature, obtains the positive sample corresponding to non-brush single operation
Refresh operation feature and negative sample refresh operation feature corresponding to brush single operation.
S504：PCA dimensionality reduction is carried out to the positive sample refresh operation feature and the negative sample refresh operation feature respectively
Processing, obtains the corresponding positive sample three-dimensional feature of the positive sample refresh operation feature and the negative sample refresh operation feature pair
The negative sample three-dimensional feature answered.
Wherein, above-mentioned positive sample refresh operation feature is identical with the dimension of negative sample refresh operation feature.
It is worth noting that can be determined based on classification results after classifying to multiple sample refresh operation features
The complexity of classification problem (the brush single operation identification model problem to be solved of i.e. subsequent training), and then choose suitable mould
Type is trained.In one embodiment, the complexity of principal component analysis PCA algorithm research classification problem can be used.PCA algorithm
It is mainly used in two aspects：First aspect is relatively common Data Dimensionality Reduction, i.e., high-dimensional feature space is projected to low-dimensional sky
Between, to reduce computation complexity, improve the whole efficiency of algorithm；The second aspect is data visualization, due to feature space
When dimension is very high, the spatial distribution of data can not be observed, thus can be by Feature Mapping to three-dimensional or two-dimensional space, to pass through
Image shows feature, and then can intuitively determine the sample distribution of classification problem.
In one embodiment, if the dimension of positive sample refresh operation feature and negative sample refresh operation feature is higher than three-dimensional,
Then PCA dimension-reduction treatment can be carried out to the positive sample refresh operation feature and the negative sample refresh operation feature respectively, obtained
It is corresponding negative to the corresponding positive sample three-dimensional feature of the positive sample refresh operation feature and the negative sample refresh operation feature
Sample three-dimensional feature.
It wherein, is that three-dimensional is at three-dimensional reason by positive sample refresh operation feature and negative sample refresh operation Feature Dimension Reduction
The highest latitude that human eye is observed that can more intuitively show the distribution of sample characteristics, have compared to two dimension better
Bandwagon effect.
S505：The positive sample three-dimensional feature and the negative sample three-dimensional feature are projected into preset three-dimensional space, obtained
To the distribution information of the positive sample three-dimensional feature and the negative sample three-dimensional feature, the distribution information is used for table
Levy the degree of closeness of the positive sample three-dimensional feature and the negative sample three-dimensional feature in the three-dimensional space.
It in one embodiment, can be by the positive sample after obtaining positive sample three-dimensional feature and the negative sample three-dimensional feature
This three-dimensional feature and negative sample three-dimensional feature project to preset three-dimensional space.
It is worth noting that the preset three-dimensional space be based on dimensionality reduction after three features establish, i.e., with dimensionality reduction after
Three features be reference axis, establish the three-dimensional space.
In one embodiment, when the positive sample three-dimensional feature and negative sample three-dimensional feature are projected to preset three-dimensional space
Between, the distribution information of available the positive sample three-dimensional feature and the negative sample three-dimensional feature.In one embodiment,
The distribution information can be in the three-dimensional space, and the positive sample three-dimensional feature and the negative sample three-dimensional feature connect
Short range degree (e.g., the distance of positive sample three-dimensional feature and the negative sample three-dimensional feature).
S506：It is close in the three-dimensional space according to the positive sample three-dimensional feature and the negative sample three-dimensional feature
Degree determines brush single operation identification model to be trained.
For example, if above-mentioned positive sample three-dimensional feature and negative sample three-dimensional feature are projected to preset three-dimensional space
Afterwards, the position of the positive sample three-dimensional feature in the three-dimensional space and negative sample three-dimensional feature is very close, is not distributed across one
The two sides of straight line indicate the problem of being classified based on the positive sample three-dimensional feature and negative sample three-dimensional feature, be not letter
The classification problem that single linear classification model is able to solve.Further, since the quantity of positive and negative sample characteristics is relatively more, thus can be with
Using a kind of relative complex disaggregated model, as (Gradient Boosting Decision Tree, is called GBDT
Multiple Additive Regression Tree, is a kind of decision Tree algorithms of iteration), RF etc..
In one embodiment, it is contemplated that GBDT model is more difficult overfitting problem occurs, thus GBDT model is determined
It is trained for brush single operation identification model to be trained.
S507：Trained brush list is treated using the positive sample refresh operation feature and the negative sample refresh operation feature
Operation identification model is trained.
S508：The refresh operation information of dispatching person in the period for meeting preset condition is obtained, it is described to meet preset condition
Period include the adjacent period robbed between single operation twice during competition for orders.
S509：The adjacent brush for robbing rear competition for orders operation in single operation twice according to the refresh operation acquisition of information
New operating characteristics.
S510：The refresh operation feature is input in advance trained brush single operation identification model, obtain it is described after
Once rob single operation whether be brush single operation recognition result.
Wherein, step S501-S503, the relevant explanation of S507-S510 and explanation may refer to above-described embodiment, herein not
It is repeated.
Seen from the above description, the present embodiment passes through respectively to the positive sample refresh operation feature and the negative sample brush
New operating characteristics carry out PCA dimension-reduction treatment, obtain the corresponding positive sample three-dimensional feature of the positive sample refresh operation feature and institute
State the corresponding negative sample three-dimensional feature of negative sample refresh operation feature, and by the positive sample three-dimensional feature and the negative sample three
Dimensional feature projects to preset three-dimensional space, obtains the distribution shape of the positive sample three-dimensional feature and the negative sample three-dimensional feature
State information, so according to the positive sample three-dimensional feature and the negative sample three-dimensional feature in the three-dimensional space close to journey
Degree determines brush single operation identification model to be trained, and may be implemented reasonably to select brush single operation identification model to be trained, is
Subsequent accurately identified based on trained brush single operation identification model robs whether single operation for brush single operation provides basis, Ke Yibao
Demonstrate,prove the accuracy of brush single operation identification model.
Fig. 6 is a kind of flow chart of brush single operation recognition methods shown in the application another exemplary embodiment；The implementation
Example can be used for server-side (for example, server cluster etc. of a server and multiple servers composition).As shown in fig. 6, should
Method includes step S601-S606：
S601：The refresh operation information of dispatching person in the period for meeting preset condition is obtained, it is described to meet preset condition
Period include the adjacent period robbed between single operation twice during competition for orders.
S602：The adjacent brush for robbing rear competition for orders operation in single operation twice according to the refresh operation acquisition of information
New operating characteristics.
S603：The refresh operation feature is input in advance trained brush single operation identification model, obtain it is described after
Once rob single operation whether be brush single operation recognition result.
Wherein, the relevant explanation of step S601-S603 and explanation may refer to above-described embodiment, no longer go to live in the household of one's in-laws on getting married herein
It states.
S604：Judging whether the recognition result characterizes the rear competition for orders operation is brush single operation；If so, executing
Step S605；Otherwise, step S606 is executed；
S605：Then forbid responding the rear competition for orders operation；
S606：Allow to respond the rear competition for orders operation.
In the related technology, after dispatching person triggers described on the terminal device after competition for orders operation, server-side can be with
Single operation is robbed in response to this, corresponding order is handed down to the terminal device of the dispatching person.And in the present embodiment, in order to avoid certain
A little dispatching persons obtain high-quality order using externally hung software brush list, thus when robbing single operation described in dispatching person's triggering, this is robbed
Single operation is identified, to determine that this robs whether single operation is brush single operation according to recognition result.In turn, when recognition result characterizes
This robs single operation when being brush single operation, forbids responding this and robs single operation, i.e., the order is not handed down to dispatching person's terminal of competition for orders
Equipment；And when it is brush single operation that recognition result, which characterizes this to rob single operation not, allowing to respond described this robs single operation, i.e., orders this
Singly it is handed down to dispatching person's terminal device of competition for orders.
In one embodiment, above-mentioned to forbid the mode for responding the rear competition for orders operation be by developer according to reality
Border needs to be configured, such as can be set to ignore this and rob single operation (detect the operation also as do not detect), or
Person can be set to pop-up verification information and rob single act etc. with interrupt the dispatching person, and the present embodiment is to this without limiting.
Seen from the above description, the present embodiment is that brush is single by the way that competition for orders operates after the recognition result characterizes described
Operation is forbidden responding the rear competition for orders operation, and competition for orders operation is not brush after recognition result characterization is described
When single operation, allows to respond the rear competition for orders operation, may be implemented to supervise and control based on the recognition result for robbing single operation
The brush single act of dispatching person, and then can be overstock to avoid order, dispatching business efficiency is promoted, and then promote dispatching person and user
Experience.
Corresponding with preceding method embodiment, present invention also provides the embodiments of corresponding device.
Fig. 7 is a kind of structural block diagram of brush single operation identification device shown in one exemplary embodiment of the application；Such as Fig. 7 institute
Show, which includes：Operation information acquisition module 110, operating characteristics obtain module 120 and recognition result obtains module 130,
Wherein：
Operation information acquisition module 110, for obtaining the refresh operation letter for meeting dispatching person in the period of preset condition
Breath, the period for meeting preset condition includes the adjacent period robbed between single operation twice during competition for orders；
Operating characteristics obtain module 120, rob single operation twice for adjacent according to the refresh operation acquisition of information
In after a competition for orders operation refresh operation feature；
Recognition result obtains module 130, knows for the refresh operation feature to be input to brush single operation trained in advance
In other model, obtain it is described after competition for orders operation whether be brush single operation recognition result
Seen from the above description, refreshing behaviour of the present embodiment for the period robbed twice during competition for orders between single operation
Make the corresponding refresh operation feature of acquisition of information, and then can accurately be identified based on brush single operation identification model trained in advance
Rob whether single operation is brush single operation, due to being with the refresh operation in the period between adjacent competition for orders twice for research pair
As, thus asking for brush single operation identification can be influenced to avoid due to adulterating the operation of other orders in the single order processing period
Topic, it is ensured that the accuracy of brush single operation identification, and foundation can be provided for the supervision and control of subsequent brush single act, in turn
Promote the experience of dispatching person and user.
Fig. 8 is a kind of structural block diagram of brush single operation identification device shown in the application another exemplary embodiment；Wherein,
Operation information acquisition module 230, operating characteristics obtain module 240 and recognition result obtains shown in module 250 and earlier figures 7 in fact
Apply the operation information acquisition module 110 in example, operating characteristics obtain module 120 and recognition result obtains the function of module 130
It is identical, herein without repeating.As shown in figure 8, operation information acquisition module 230 can be also used for adjacent robbing twice when detecting
After in single operation when competition for orders operation, the refresh operation information of dispatching person in the period for meeting preset condition is obtained.
In one embodiment, described device can also include identification model training module 210, for pre- according to following steps
First train brush single operation identification model；
The identification model training module 210 may include：
Sample information acquiring unit 211, for obtaining a plurality of sample refresh operation of sample dispatching person in target time section
Information, the target time section include the adjacent period robbed between single operation twice during competition for orders；
Sample characteristics acquiring unit 212, for a rear sample according to every sample refresh operation acquisition of information
Originally the sample refresh operation feature for robbing single operation, the multiple sample refresh operations for obtaining a plurality of sample refresh operation information are special
Sign；
Sample characteristics taxon 213 is corresponded to for classifying to the multiple sample refresh operation feature
The positive sample refresh operation feature of non-brush single operation and negative sample refresh operation feature corresponding to brush single operation；
Identification model training unit 214, for refreshing behaviour using the positive sample refresh operation feature and the negative sample
Trained brush single operation identification model is treated as feature to be trained.
In one embodiment, sample characteristics taxon 213 can be also used for：
Using preset unsupervised algorithm, behaviour is refreshed to the sample of the first quantity in the multiple sample refresh operation feature
Classify as feature, the positive sample refresh operation obtained in the sample refresh operation feature for characterizing first quantity is special
First classification results of negative sample refresh operation feature of seeking peace；
Based on first classification results, using preset semi-supervised algorithm, to the multiple sample refresh operation feature
In the sample refresh operation feature of the second quantity classify, the sample refresh operation obtained for characterizing second quantity is special
Second classification results of positive sample refresh operation feature and negative sample refresh operation feature in sign；
Wherein, the sample refresh operation feature of second quantity is in the multiple sample refresh operation feature except described
Feature except the sample refresh operation feature of first quantity, and first quantity is less than second quantity.
In one embodiment, sample characteristics taxon 213 can be also used for：
To the positive sample refresh operation feature or negative sample refresh operation of category division mistake in first classification results
Feature is corrected, the first classification results after being corrected；
It is executed based on the first classification results after the correction described in the multiple sample refresh operation feature second
The operation that the sample refresh operation feature of quantity is classified.
In one embodiment, the sample refresh operation feature includes at least one of following：
In the duration of refresh operation number, the target time section in the target time section, the target time section
Time interval between last time refresh operation and the rear competition for orders operation, the refresh operation in the target time section
Update in frequency, the target time section refreshes than the pressure value of terminal device screen, the mesh in, the target time section
Mark the sliding number of terminal device screen in the period；
The update refreshes than for characterizing first number and second several ratio, first number is the target
The number of new order is obtained by refresh operation in period, second number is the refresh operation in the target time section
Number.
In one embodiment, the dimension of the sample refresh operation feature is higher than three-dimensional；
Described device further includes：Identification model determining module 220；
Identification model determining module 220 can also include：
Sample characteristics dimensionality reduction unit 221, for refreshing respectively to the positive sample refresh operation feature and the negative sample
Operating characteristics carry out PCA dimension-reduction treatment, obtain the corresponding positive sample three-dimensional feature of the positive sample refresh operation feature and described
The corresponding negative sample three-dimensional feature of negative sample refresh operation feature；
State information acquisition unit 222, for projecting the positive sample three-dimensional feature and the negative sample three-dimensional feature
To preset three-dimensional space, the distribution information of the positive sample three-dimensional feature and the negative sample three-dimensional feature, institute are obtained
Distribution information is stated for characterizing the positive sample three-dimensional feature and the negative sample three-dimensional feature in the three-dimensional space
Degree of closeness；
Identification model determination unit 223, for being existed according to the positive sample three-dimensional feature and the negative sample three-dimensional feature
Degree of closeness in the three-dimensional space determines brush single operation identification model to be trained.
In one embodiment, described device can also include：Competition for orders operation processing module 260；
The competition for orders operation processing module 260 may include：
Attendant exclusion response unit 261 is the single behaviour of brush for competition for orders to operate after recognition result characterization is described
When making, forbid responding the rear competition for orders operation；
Operation allows response unit 262, is not that brush is single for competition for orders to operate after recognition result characterization is described
When operation, allow to respond the rear competition for orders operation.
It is worth noting that all the above alternatives, can form the optional reality of the disclosure using any combination
Example is applied, this is no longer going to repeat them.
On the other hand, present invention also provides a kind of computer readable storage medium, storage medium is stored with computer journey
Sequence, computer program are used to execute the brush single operation recognition methods that above-mentioned Figure 1A~embodiment illustrated in fig. 6 provides.
For device embodiment, since it corresponds essentially to embodiment of the method, so related place is referring to method reality
Apply the part explanation of example.The apparatus embodiments described above are merely exemplary, wherein described be used as separation unit
The unit of explanation may or may not be physically separated, and component shown as a unit can be or can also be with
It is not physical unit, it can it is in one place, or may be distributed over multiple network units.It can be according to actual
The purpose for needing to select some or all of the modules therein to realize application scheme.Those of ordinary skill in the art are not paying
Out in the case where creative work, it can understand and implement.
Those skilled in the art after considering the specification and implementing the invention disclosed here, will readily occur to its of the application
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the application, these modifications, purposes or
Person's adaptive change follows the general principle of the application and including the undocumented common knowledge in the art of the application
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the application are by following
Claim is pointed out.
It should also be noted that, the terms "include", "comprise" or its any other variant are intended to nonexcludability
It include so that the process, method, commodity or the equipment that include a series of elements not only include those elements, but also to wrap
Include other elements that are not explicitly listed, or further include for this process, method, commodity or equipment intrinsic want
Element.In the absence of more restrictions, the element limited by sentence "including a ...", it is not excluded that including described want
There is also other identical elements in the process, method of element, commodity or equipment.
The foregoing is merely the preferred embodiments of the application, not to limit the application, all essences in the application
Within mind and principle, any modification, equivalent substitution, improvement and etc. done be should be included within the scope of the application protection.
Claims (17)
1. a kind of brush single operation recognition methods, which is characterized in that including：
Obtain the refresh operation information of dispatching person in the period for meeting preset condition, the period packet for meeting preset condition
Include the adjacent period robbed between single operation twice during competition for orders；
The adjacent refresh operation for robbing rear competition for orders operation in single operation twice according to the refresh operation acquisition of information is special
Sign；
The refresh operation feature is input in brush single operation identification model trained in advance, obtains the rear competition for orders behaviour
Whether be brush single operation recognition result, the brush single operation identification model be according to sample dispatching person in target time section
What a plurality of sample refresh operation information was trained.
2. the method according to claim 1, wherein described obtain dispatching person in the period for meeting preset condition
Refresh operation information, including：
If detecting competition for orders operation after adjacent robbed in single operation twice, dispatching in the period for meeting preset condition is obtained
The refresh operation information of member.
3. the method according to claim 1, wherein the brush single operation identification model is preparatory according to following steps
Training：
The a plurality of sample refresh operation information of sample dispatching person in target time section is obtained, the target time section includes competition for orders
The adjacent period robbed between single operation twice in journey；
A sample robs the sample refresh operation feature of single operation after according to every sample refresh operation acquisition of information,
Obtain multiple sample refresh operation features of a plurality of sample refresh operation information；
Classify to the multiple sample refresh operation feature, obtains special corresponding to the positive sample refresh operation of non-brush single operation
The negative sample refresh operation feature sought peace corresponding to brush single operation；
Trained brush single operation is treated using the positive sample refresh operation feature and the negative sample refresh operation feature to identify
Model is trained.
4. according to the method described in claim 3, it is characterized in that, described divide the multiple sample refresh operation feature
Class, obtains the positive sample refresh operation feature corresponding to non-brush single operation and the negative sample refresh operation corresponding to brush single operation is special
Sign, including：
It is special to the sample refresh operation of the first quantity in the multiple sample refresh operation feature using preset unsupervised algorithm
Sign is classified, obtain positive sample refresh operation feature in the sample refresh operation feature for characterizing first quantity and
First classification results of negative sample refresh operation feature；
Based on first classification results, using preset semi-supervised algorithm, in the multiple sample refresh operation feature
The sample refresh operation feature of two quantity is classified, and is obtained in the sample refresh operation feature for characterizing second quantity
Positive sample refresh operation feature and negative sample refresh operation feature the second classification results；
Wherein, the sample refresh operation feature of second quantity is that described first is removed in the multiple sample refresh operation feature
Feature except the sample refresh operation feature of quantity, and first quantity is less than second quantity.
5. according to the method described in claim 4, it is characterized in that, the method also includes：
To the positive sample refresh operation feature or negative sample refresh operation feature of category division mistake in first classification results
It is corrected, the first classification results after being corrected；
It is executed based on the first classification results after the correction described to the second quantity in the multiple sample refresh operation feature
The operation classified of sample refresh operation feature.
6. according to the method described in claim 3, it is characterized in that, the sample refresh operation feature includes following at least one
?：
It is last in the duration of refresh operation number, the target time section in the target time section, the target time section
Time interval between refresh operation and the rear competition for orders operation, the refresh operation frequency in the target time section
Update in rate, the target time section refreshes pressure value, the target than terminal device screen in, the target time section
The sliding number of terminal device screen in period；
The update refreshes than for characterizing first number and second several ratio, first number is the object time
The number of new order is obtained by refresh operation in section, second number is the refresh operation in the target time section
Number.
7. according to the method described in claim 3, it is characterized in that, the dimension of the sample refresh operation feature is higher than three-dimensional；
It is described to treat trained brush single operation using the positive sample refresh operation feature and the negative sample refresh operation feature
Before identification model is trained, the method also includes：
PCA dimension-reduction treatment is carried out to the positive sample refresh operation feature and the negative sample refresh operation feature respectively, is obtained
The corresponding positive sample three-dimensional feature of positive sample refresh operation feature and the corresponding negative sample of the negative sample refresh operation feature
This three-dimensional feature；
The positive sample three-dimensional feature and the negative sample three-dimensional feature are projected into preset three-dimensional space, obtain the positive sample
The distribution information of this three-dimensional feature and the negative sample three-dimensional feature, the distribution information is for characterizing the positive sample
The degree of closeness of this three-dimensional feature and the negative sample three-dimensional feature in the three-dimensional space；
It is determined according to the degree of closeness of the positive sample three-dimensional feature and the negative sample three-dimensional feature in the three-dimensional space
Brush single operation identification model to be trained.
8. the method according to claim 1, wherein the method also includes：
If competition for orders operation is brush single operation after the recognition result characterization is described, forbid responding the rear competition for orders behaviour
Make；
If competition for orders operation is not brush single operation after the recognition result characterization is described, allow to respond a rear competition for orders
Operation.
9. a kind of brush single operation identification device, which is characterized in that including：
Operation information acquisition module, it is described for obtaining the refresh operation information for meeting dispatching person in the period of preset condition
The period for meeting preset condition includes the adjacent period robbed between single operation twice during competition for orders；
Operating characteristics obtain module, for according to the refresh operation acquisition of information it is adjacent rob in single operation twice after it is primary
Rob the refresh operation feature of single operation；
Recognition result obtains module, for the refresh operation feature to be input to brush single operation identification model trained in advance
In, obtain it is described after competition for orders operation whether be brush single operation recognition result, according to the brush single operation identification model
The a plurality of sample refresh operation information of sample dispatching person is trained in target time section.
10. device according to claim 9, which is characterized in that the operation information acquisition module is also used to work as and detect
It is adjacent rob in single operation twice after a competition for orders operation when, obtain the refresh operation of dispatching person in the period for meeting preset condition
Information.
11. device according to claim 9, which is characterized in that described device further includes identification model training module, is used for
Brush single operation identification model is trained in advance according to following steps；
The identification model training module includes：
Sample information acquiring unit, for obtaining a plurality of sample refresh operation information of sample dispatching person in target time section, institute
Stating target time section includes the adjacent period robbed between single operation twice during competition for orders；
Sample characteristics acquiring unit, for the rear sample competition for orders behaviour according to every sample refresh operation acquisition of information
The sample refresh operation feature of work obtains multiple sample refresh operation features of a plurality of sample refresh operation information；
Sample characteristics taxon obtains corresponding to non-brush list for classifying to the multiple sample refresh operation feature
The positive sample refresh operation feature of operation and negative sample refresh operation feature corresponding to brush single operation；
Identification model training unit, for utilizing the positive sample refresh operation feature and the negative sample refresh operation feature pair
Brush single operation identification model to be trained is trained.
12. device according to claim 11, which is characterized in that the sample characteristics taxon is also used to：
It is special to the sample refresh operation of the first quantity in the multiple sample refresh operation feature using preset unsupervised algorithm
Sign is classified, obtain positive sample refresh operation feature in the sample refresh operation feature for characterizing first quantity and
First classification results of negative sample refresh operation feature；
Based on first classification results, using preset semi-supervised algorithm, in the multiple sample refresh operation feature
The sample refresh operation feature of two quantity is classified, and is obtained in the sample refresh operation feature for characterizing second quantity
Positive sample refresh operation feature and negative sample refresh operation feature the second classification results；
Wherein, the sample refresh operation feature of second quantity is that described first is removed in the multiple sample refresh operation feature
Feature except the sample refresh operation feature of quantity, and first quantity is less than second quantity.
13. device according to claim 12, which is characterized in that the sample characteristics taxon is also used to：
To the positive sample refresh operation feature or negative sample refresh operation feature of category division mistake in first classification results
It is corrected, the first classification results after being corrected；
It is executed based on the first classification results after the correction described to the second quantity in the multiple sample refresh operation feature
The operation classified of sample refresh operation feature.
14. device according to claim 11, which is characterized in that the sample refresh operation feature includes following at least one
?：
It is last in the duration of refresh operation number, the target time section in the target time section, the target time section
Time interval between refresh operation and the rear competition for orders operation, the refresh operation frequency in the target time section
Update in rate, the target time section refreshes pressure value, the target than terminal device screen in, the target time section
The sliding number of terminal device screen in period；
The update refreshes than for characterizing first number and second several ratio, first number is the object time
The number of new order is obtained by refresh operation in section, second number is the refresh operation in the target time section
Number.
15. device according to claim 11, which is characterized in that the dimension of the sample refresh operation feature is higher than three
Dimension；
Described device further includes：Identification model determining module；
The identification model determining module further includes：
Sample characteristics dimensionality reduction unit, for respectively to the positive sample refresh operation feature and the negative sample refresh operation feature
PCA dimension-reduction treatment is carried out, the corresponding positive sample three-dimensional feature of the positive sample refresh operation feature and the negative sample brush are obtained
The corresponding negative sample three-dimensional feature of new operating characteristics；
State information acquisition unit, it is preset for projecting to the positive sample three-dimensional feature and the negative sample three-dimensional feature
Three-dimensional space obtains the distribution information of the positive sample three-dimensional feature and the negative sample three-dimensional feature, the distribution shape
State information be used to characterize the positive sample three-dimensional feature and the negative sample three-dimensional feature in the three-dimensional space close to journey
Degree；
Identification model determination unit is used for according to the positive sample three-dimensional feature and the negative sample three-dimensional feature in the three-dimensional
Degree of closeness in space determines brush single operation identification model to be trained.
16. device according to claim 9, which is characterized in that described device further includes：Competition for orders operation processing module；
The competition for orders operation processing module, including：
Attendant exclusion response unit, for prohibiting when competition for orders operation is brush single operation after recognition result characterization is described
Only respond the rear competition for orders operation；
Operation allows response unit, for when competition for orders operation is not brush single operation after recognition result characterization is described,
Allow to respond the rear competition for orders operation.
17. a kind of computer readable storage medium, which is characterized in that the storage medium is stored with computer program, the meter
Calculation machine program is used to execute any brush single operation recognition methods of the claims 1-8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810792415.5A CN108921581B (en) | 2018-07-18 | 2018-07-18 | Method and device for identifying bill-swiping operation and computer-readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810792415.5A CN108921581B (en) | 2018-07-18 | 2018-07-18 | Method and device for identifying bill-swiping operation and computer-readable storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN108921581A true CN108921581A (en) | 2018-11-30 |
| CN108921581B CN108921581B (en) | 2021-07-02 |
Family
ID=64416350
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810792415.5A Active CN108921581B (en) | 2018-07-18 | 2018-07-18 | Method and device for identifying bill-swiping operation and computer-readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108921581B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112529505A (en) * | 2020-12-21 | 2021-03-19 | 北京顺达同行科技有限公司 | Illegal bill-swiping detection method and device and readable storage medium |
| CN113383362A (en) * | 2019-06-24 | 2021-09-10 | 深圳市欢太科技有限公司 | User identification method and related product |
| CN113742719A (en) * | 2021-08-26 | 2021-12-03 | 深圳依时货拉拉科技有限公司 | Order grabbing and plug-in hanging detection method and computer equipment |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040163135A1 (en) * | 2000-05-25 | 2004-08-19 | Giaccherini Thomas Nello | Method for securely distributing & updating software |
| CN106022826A (en) * | 2016-05-18 | 2016-10-12 | 武汉斗鱼网络科技有限公司 | Cheating user recognition method and system in webcast platform |
| CN107146089A (en) * | 2017-03-29 | 2017-09-08 | 北京三快在线科技有限公司 | The single recognition methods of one kind brush and device, electronic equipment |
-
2018
- 2018-07-18 CN CN201810792415.5A patent/CN108921581B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20040163135A1 (en) * | 2000-05-25 | 2004-08-19 | Giaccherini Thomas Nello | Method for securely distributing & updating software |
| CN106022826A (en) * | 2016-05-18 | 2016-10-12 | 武汉斗鱼网络科技有限公司 | Cheating user recognition method and system in webcast platform |
| CN107146089A (en) * | 2017-03-29 | 2017-09-08 | 北京三快在线科技有限公司 | The single recognition methods of one kind brush and device, electronic equipment |
Non-Patent Citations (1)
| Title |
|---|
| 匿名用户: "现在美团众包手动刷新抢单太难了", 《HTTPS://TIEBA.BAIDU.COM/P/5291901664》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113383362A (en) * | 2019-06-24 | 2021-09-10 | 深圳市欢太科技有限公司 | User identification method and related product |
| CN113383362B (en) * | 2019-06-24 | 2022-05-13 | 深圳市欢太科技有限公司 | User identification method and related product |
| CN112529505A (en) * | 2020-12-21 | 2021-03-19 | 北京顺达同行科技有限公司 | Illegal bill-swiping detection method and device and readable storage medium |
| CN112529505B (en) * | 2020-12-21 | 2024-02-27 | 北京顺达同行科技有限公司 | Method and device for detecting illegal bill, and readable storage medium |
| CN113742719A (en) * | 2021-08-26 | 2021-12-03 | 深圳依时货拉拉科技有限公司 | Order grabbing and plug-in hanging detection method and computer equipment |
| CN113742719B (en) * | 2021-08-26 | 2022-04-15 | 深圳依时货拉拉科技有限公司 | Order grabbing and plug-in hanging detection method and computer equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN108921581B (en) | 2021-07-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Li et al. | The wisdom of minority: Discovering and targeting the right group of workers for crowdsourcing | |
| Dekel et al. | Selective sampling and active learning from single and multiple teachers | |
| KR101853091B1 (en) | Method, apparatus and computer program for providing personalized educational contents through user response prediction framework with machine learning | |
| Sai et al. | Facial age range estimation with extreme learning machines | |
| Lan et al. | Behavior-Based Latent Variable Model for Learner Engagement. | |
| CN108647571B (en) | Video motion classification model training method and device and video motion classification method | |
| CN108921581A (en) | A kind of brush single operation recognition methods, device and computer readable storage medium | |
| KR101895959B1 (en) | Method, apparatus and computer program for interpreting analysis results of machine learning framework | |
| CN107145857A (en) | Face character recognition methods, device and method for establishing model | |
| CN109635833A (en) | A kind of image-recognizing method and system based on cloud platform and model intelligent recommendation | |
| Xiong et al. | Predicting learning status in MOOCs using LSTM | |
| KR20190025873A (en) | Method, apparatus and computer program for providing educational contents | |
| CN108074203A (en) | A kind of teaching readjustment method and apparatus | |
| Kaardal et al. | Identifying functional bases for multidimensional neural computations | |
| Sha et al. | Neural knowledge tracing | |
| Colliri et al. | A network-based high level data classification technique | |
| KR20190049627A (en) | Method, apparatus and computer program for interpreting analysis results of machine learning framework | |
| Guise et al. | A Bayesian model of polychronicity | |
| US11482127B2 (en) | System and method for behavioral analysis and recommendations | |
| Lalor et al. | CIFT: Crowd-informed fine-tuning to improve machine learning ability | |
| EL MEZOUARY et al. | An evaluation of learner clustering based on learning styles in MOOC course | |
| KR20190025871A (en) | Method, apparatus and computer program for providing personalized educational contents | |
| Ponzi et al. | Analysis pre and post covid-19 pandemic rorschach test data of using em algorithms and gmm models | |
| Kavya et al. | Student Personality Analysis In Blended Mode Using Big Five | |
| de Lima et al. | Evisclass: a new evaluation method for image data stream classifiers |
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 | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1258546 Country of ref document: HK |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |