CN110069617A - Management method, device, server and the computer readable storage medium of forum's note - Google Patents
Management method, device, server and the computer readable storage medium of forum's note Download PDFInfo
- Publication number
- CN110069617A CN110069617A CN201711100303.0A CN201711100303A CN110069617A CN 110069617 A CN110069617 A CN 110069617A CN 201711100303 A CN201711100303 A CN 201711100303A CN 110069617 A CN110069617 A CN 110069617A
- Authority
- CN
- China
- Prior art keywords
- model
- note
- smart
- forum
- plus
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/335—Filtering based on additional data, e.g. user or group profiles
- G06F16/337—Profile generation, learning or modification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
- G06F16/353—Clustering; Classification into predefined classes
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides management method, device, server and the computer readable storage mediums of a kind of forum's note, wherein the management method of forum's note includes: that the input of current forum's note is added smart model automatically;According to the output result of automatic plus smart model, it is determined whether executed to current forum's note and add essence operation.Technical solution of the present invention, on the one hand, the automatic plus essence function to forum's note is realized, user is saved and adds smart operation manually, improve the usage experience of user, on the other hand, add smart model automatically by being arranged, smart model can be added according to different setting rules, training automatically, add smart standard to determine, so as to meet different use occasions.
Description
Technical field
The present invention relates to field of computer technology, management method, a kind of forum's note in particular to a kind of forum's note
Managing device, a kind of server and a kind of computer readable storage medium.
Background technique
Nowadays, internet has become the important channel that user obtains information, communication, and forum is (for example, a certain city
Car owner's circle) issued and alternatively, the network user can see note in forum, post and money order receipt to be signed and returned to the sender as different types of information,
While sharing information, useful information is absorbed, forum's note includes the types such as pure words note, picture patch and video patch, with picture
For patch, the model comprising pictures such as beauty, landscape, pets, user's access turns note, money order receipt to be signed and returned to the sender, the amount for chasing after patch, thumbing up and can compare
It is more, and then become popular note, after forum administrator can generally add essence or top set to handle to facilitate user these popular exchange premium rows
Continuous access, current adds essence operation to have the following deficiencies:
Forum administrator generally requires browses each model one by one, then judges whether plus essence or the top set model, this is needed
Many human costs are put into, and influence whole efficiency.
Summary of the invention
The present invention is directed to solve at least one of the technical problems existing in the prior art or related technologies.
For this purpose, it is an object of the present invention to provide a kind of management methods of forum's note.
It is another object of the present invention to provide a kind of devices of the management of forum's note.
Yet another object of the invention is that providing a kind of server.
Yet another object of the invention is that providing a kind of computer readable storage medium.
To achieve the goals above, the technical solution of first aspect present invention provides a kind of management method of forum's note,
It include: that the input of current forum's note is added into smart model automatically;According to the output result of automatic plus smart model, it is determined whether to current opinion
Altar note executes plus essence operation.
In the technical scheme, by the way that the input of current forum's note automatically plus in smart model, is added smart model so that basis is automatic
Output result, it is determined whether current forum's note is carried out plus essence, on the one hand, realize to forum's note automatic plus essence function,
It saves user and adds smart operation manually, improve the usage experience of user, it on the other hand, can by the way that automatic plus smart model is arranged
With according to different setting rules, training is automatic plus smart model, to determine plus smart standard, different uses field so as to meet
It closes.
It is understood that preset automatic plus smart model, can be assessed, really based on the quality to forum's note itself
Fixed automatic plus smart model plus smart standard can also determine plus smart standard according to the feedback of user behavior, can be combined with forum
The quality of note itself and the feedback of user behavior determine and add smart standard.
In the above-mentioned technical solutions, it is preferable that before by the automatic plus smart model of current forum's note input, further includes: collect
Training sample set;Model training is executed to the initial model prestored according to training sample set, to be repaired in real time to initial model
Just;Detecting ratio that training sample is obtained according to revised initial model plus smart more than or equal to preset ratio threshold value
When, revised initial model is determined as to add smart model automatically.
In the technical scheme, by collecting training sample set, to be carried out according to training sample to preset initial model
Model training finds the process of model parameter, using training sample set as defeated during model training according to given data
Enter value, according to the degree of optimization of output valve, initial model is corrected in real time, (adds smart ratio big obtaining optimal output valve
In or equal to preset ratio threshold value) when, it determines corresponding mode parameter, completes training process, realize high-precision and high precision
Property braking plus smart model method of determination, and it is adaptable.
It should be noted that the training sample for training should be with forum's note building frame having the same, it can
Characteristic information identical with forum's note is got, on the one hand the computational accuracy of automatic plus smart model depends on training sample concentration
The correlation of training sample and current forum's note, on the other hand also depends on the radix of training sample.
In addition, initial model can be generated according to one or more history essence forum's note, training sample can also be passed through
The training sample of concentration generates, and can also be generated by directly default.
For example, training sample can be the history essence note derived from this forum (for example, the car owner in a certain city encloses), it can also
To be derived from the high-quality model of other platforms of network, training sample can be determined directly by system, can also pass through artificial screening
It determines, such as some essence notes containing particular picture theme (for example, landscape theme, pet-theme) of artificial screening.
In the above-mentioned technical solutions, it is preferable that model training is executed to the initial model prestored according to training sample set, with
Real-time training pattern is obtained, specifically includes the following steps: extracting the characteristic information for the training sample that training sample is concentrated;It determines just
Multiple model parameters corresponding with characteristic information in beginning model;Multiple model parameters are corrected in real time according to characteristic information,
Until revised multiple model parameters meet plus smart ratio is greater than or equal to preset ratio threshold value.
In the technical scheme, by extracting the characteristic information of training sample, add essence operation to determine according to characteristic information
With the mapping relations before characteristic information, and mapping relations are especially by the reality of multiple model parameters corresponding with characteristic information
Value determines, under the premise of given one or more specific output valves, is corrected in real time to model parameter, is further promoted
The output accuracy of automatic plus smart model.
Specifically, the model training according to training sample set to the initial model prestored can pass through a variety of training algorithms
It realizes, for example using GBDT (Gradient Boosting Decision Tree, gradient promote decision tree), GBDT has more
A learner, and there are sequencing between learner, each training sample has weighted value, and it is initial in training, each
The weighted value of sample is equal, is trained first using first learner to training sample, after the completion of training, increases wrong sample
This weight, while reducing the weight of correct sample, then learnt using second learner, finally obtains b study
After device, merge training result, adds smart ratio to be greater than or equal to default plus essence ratio to realize, wherein learner is specially CART
(post-class processing), is trained using CART, and the quantity for heating training sample is m, and training sample can characterize are as follows:
{(X(1),y(1)),(X(2),y(2)),…,(X(m),y(m))}
Wherein, X(i)For n-dimensional vector, the characteristic information of i-th of sample, y are characterized(i)For model parameter.
Furthermore it is also possible to execute model training, detailed process using neural network (Nerual Network) algorithm are as follows: mind
It is formed by multiple logic units according to different hierarchical organizations through network model, the input that each layer of output variable is next layer
Training sample set is configured to a training matrix, using training sample set as input value, smart ratio will be added to be greater than or wait by variable
Model training is finally completed with Optimized model parameter as output valve condition in preset ratio threshold value.
In the above-mentioned technical solutions, it is preferable that characteristic information includes behavioural characteristic and content characteristic, and behavioural characteristic includes visiting
The amount of asking, money order receipt to be signed and returned to the sender amount turn note amount, chase after patch amount and at least one in the amount of thumbing up, and when forum's note is that picture pastes, content characteristic includes
At least one of in picture theme, picture number and picture clarity, wherein multiple model parameters include corresponding with behavioural characteristic
At least one first weighted value, and at least one second weighted value corresponding with content characteristic.
In the technical scheme, behavioural characteristic include amount of access, money order receipt to be signed and returned to the sender amount, turn note amount, chase after patch amount and the amount of thumbing up in extremely
One item missing, when forum's note is that picture pastes, content characteristic includes at least one in picture theme, picture number and picture clarity
, when behavioural characteristic includes above-mentioned all information, amount of access, money order receipt to be signed and returned to the sender amount turn that note amount, to chase after patch amount one corresponding with the amount of thumbing up
First weighted value, when content characteristic includes all of the above information, picture theme, picture number and picture clarity corresponding one
A second weighted value, the process of model training determine the process of the first weighted value and the second weighted value, according to adding smart standard
Just, it moves and adjusts preset ratio threshold value.
When adding smart standard broadest, preset ratio threshold value is 1, i.e., whole training samples are to add smart forum note.
In the above-mentioned technical solutions, it is preferable that multiple model parameters are corrected in real time according to characteristic information, are also wrapped
Include: when training sample includes history essence forum's note of multiple forums, determine history essence forum note adds the essence operation moment,
And add essence operation moment corresponding behavioural characteristic;When according to adding essence operation moment corresponding behavioural characteristic and add essence operation
It carves, determines the average behavior rate of behavioural characteristic;The first weighted value is modified according to average behavior rate.
In the technical scheme, by determining behavioural characteristic according to smart time and behavioural characteristic corresponding when adding smart is added
Average behavior rate, using behavior rate as the input value being modified to the first weighted value, compared with behavioural characteristic, behavior speed
The measurement to the smart operating time is added is increased in rate, i.e., within a specified time whether the magnitude of behavioural characteristic reaches preset value, or
The magnitude of behavioural characteristic reaches the time used in preset value, adds the smart time by determining, during model training, additionally it is possible to
Extra garbage is rejected, the more efficient of model training is made.
Specifically, by taking amount of access as an example, it is assumed that the amount of access of a certain training sample executes when reaching 500 times plus essence is grasped
Make, at this time from the model publication moment, have passed through 40 minutes, that is, can determine that corresponding average behavior rate is 12.5 beats/min
Clock.
It should be noted that plus essence operation the moment refer to from model issue the moment, to execute plus essence operate it is experienced when
Between.
In the above-mentioned technical solutions, it is preferable that according to the output result of automatic plus smart model, it is determined whether to current forum
Note executes plus essence operation, specifically includes the following steps: obtaining the defeated of automatic plus smart model according to the characteristic information of current forum's note
It is worth out;It is detecting that output valve is greater than or equal to plus when smart threshold value, is executing plus essence operation, and current forum is posted on note team, forum
Top set operation is executed in column;When detecting output valve less than smart threshold value is added, does not execute plus essence operates.
In the technical scheme, add smart threshold value by being arranged, to add automatically by the characteristic information input of current forum's note
After smart model, output valve is compared with smart threshold value is added, when output valve is greater than or equal to and adds smart threshold value, execution plus essence are grasped
Make, when output valve is less than and adds smart threshold value, does not then execute and add essence operation, under the premise of meeting reliability, calculated model
Journey is simpler, wherein smart threshold value is added to be referred to the setting of preset ratio threshold value.
The technical solution of the second aspect of the present invention provides a kind of managing device of forum's note, comprising: input unit is used
Add smart model automatically in inputting current forum's note;Output unit, the output for the automatic plus smart model of basis is as a result, determination is
No execute to current forum's note adds essence operation.
In the technical scheme, by the way that the input of current forum's note automatically plus in smart model, is added smart model so that basis is automatic
Output result, it is determined whether current forum's note is carried out plus essence, on the one hand, realize to forum's note automatic plus essence function,
It saves user and adds smart operation manually, improve the usage experience of user, it on the other hand, can by the way that automatic plus smart model is arranged
With according to different setting rules, training is automatic plus smart model, to determine plus smart standard, different uses field so as to meet
It closes.
It is understood that preset automatic plus smart model, can be assessed, really based on the quality to forum's note itself
Fixed automatic plus smart model plus smart standard can also determine plus smart standard according to the feedback of user behavior, can be combined with forum
The quality of note itself and the feedback of user behavior determine and add smart standard.
In the above-mentioned technical solutions, it is preferable that further include: collector unit, for collecting training sample set;Training unit,
For executing model training to the initial model prestored according to training sample set, to be corrected in real time to initial model;It determines
Unit, for detecting ratio that training sample is obtained according to revised initial model plus smart more than or equal to preset ratio
When threshold value, revised initial model is determined as to add smart model automatically.
In the technical scheme, by collecting training sample set, to be carried out according to training sample to preset initial model
Model training finds the process of model parameter, using training sample set as defeated during model training according to given data
Enter value, according to the degree of optimization of output valve, initial model is corrected in real time, (adds smart ratio big obtaining optimal output valve
In or equal to preset ratio threshold value) when, it determines corresponding mode parameter, completes training process, realize high-precision and high precision
Property braking plus smart model method of determination, and it is adaptable.
It should be noted that the training sample for training should be with forum's note building frame having the same, it can
Characteristic information identical with forum's note is got, on the one hand the computational accuracy of automatic plus smart model depends on training sample concentration
The correlation of training sample and current forum's note, on the other hand also depends on the radix of training sample.
In addition, initial model can be generated according to one or more history essence forum's note, training sample can also be passed through
The training sample of concentration generates, and can also be generated by directly default.
For example, training sample can be the history essence note derived from this forum (for example, the car owner in a certain city encloses), it can also
To be derived from the high-quality model of other platforms of network, training sample can be determined directly by system, can also pass through artificial screening
It determines, such as some essence notes containing particular picture theme (for example, landscape theme, pet-theme) of artificial screening.
In the above-mentioned technical solutions, it is preferable that further include: extraction unit, for extracting the training sample of training sample concentration
This characteristic information;Determination unit is also used to: determining multiple model parameters corresponding with characteristic information in initial model;Management dress
It sets further include: amending unit, for being corrected in real time according to characteristic information to multiple model parameters, until revised multiple
Model parameter meets plus smart ratio is greater than or equal to preset ratio threshold value.
In the technical scheme, by extracting the characteristic information of training sample, add essence operation to determine according to characteristic information
With the mapping relations before characteristic information, and mapping relations are especially by the reality of multiple model parameters corresponding with characteristic information
Value determines, under the premise of given one or more specific output valves, is corrected in real time to model parameter, is further promoted
The output accuracy of automatic plus smart model.
Specifically, the model training according to training sample set to the initial model prestored can pass through a variety of training algorithms
It realizes, for example using GBDT (Gradient Boosting Decision Tree, gradient promote decision tree), GBDT has more
A learner, and there are sequencing between learner, each training sample has weighted value, and it is initial in training, each
The weighted value of sample is equal, is trained first using first learner to training sample, after the completion of training, increases wrong sample
This weight, while reducing the weight of correct sample, then learnt using second learner, finally obtains b study
After device, merge training result, adds smart ratio to be greater than or equal to default plus essence ratio to realize, wherein learner is specially CART
(post-class processing), is trained using CART, and the quantity for heating training sample is m, and training sample can characterize are as follows:
{(X(1),y(1)),(X(2),y(2)),…,(X(m),y(m))}
Wherein, X(i)For n-dimensional vector, the characteristic information of i-th of sample, y are characterized(i)For model parameter.
Furthermore it is also possible to execute model training, detailed process using neural network (Nerual Network) algorithm are as follows: mind
It is formed by multiple logic units according to different hierarchical organizations through network model, the input that each layer of output variable is next layer
Training sample set is configured to a training matrix, using training sample set as input value, smart ratio will be added to be greater than or wait by variable
Model training is finally completed with Optimized model parameter as output valve condition in preset ratio threshold value.
In the above-mentioned technical solutions, it is preferable that characteristic information includes behavioural characteristic and content characteristic, and behavioural characteristic includes visiting
The amount of asking, money order receipt to be signed and returned to the sender amount turn note amount, chase after patch amount and at least one in the amount of thumbing up, and when forum's note is that picture pastes, content characteristic includes
At least one of in picture theme, picture number and picture clarity, wherein multiple model parameters include corresponding with behavioural characteristic
At least one first weighted value, and at least one second weighted value corresponding with content characteristic.
In the technical scheme, behavioural characteristic include amount of access, money order receipt to be signed and returned to the sender amount, turn note amount, chase after patch amount and the amount of thumbing up in extremely
One item missing, when forum's note is that picture pastes, content characteristic includes at least one in picture theme, picture number and picture clarity
, when behavioural characteristic includes above-mentioned all information, amount of access, money order receipt to be signed and returned to the sender amount turn that note amount, to chase after patch amount one corresponding with the amount of thumbing up
First weighted value, when content characteristic includes all of the above information, picture theme, picture number and picture clarity corresponding one
A second weighted value, the process of model training determine the process of the first weighted value and the second weighted value, according to adding smart standard
Just, it moves and adjusts preset ratio threshold value.
When adding smart standard broadest, preset ratio threshold value is 1, i.e., whole training samples are to add smart forum note.
In the above-mentioned technical solutions, it is preferable that determination unit is also used to: including the history essence of multiple forums in training sample
When magnificent forum's note, determine history essence forum note adds the essence operation moment, and essence is added to operate moment corresponding behavioural characteristic;Really
Order member is also used to: according to adding essence operation moment corresponding behavioural characteristic and adding the essence operation moment, determining the flat of behavioural characteristic
Equal behavior rate;Amending unit is also used to: being modified according to average behavior rate to the first weighted value.
In the technical scheme, by determining behavioural characteristic according to smart time and behavioural characteristic corresponding when adding smart is added
Average behavior rate, using behavior rate as the input value being modified to the first weighted value, compared with behavioural characteristic, behavior speed
The measurement to the smart operating time is added is increased in rate, i.e., within a specified time whether the magnitude of behavioural characteristic reaches preset value, or
The magnitude of behavioural characteristic reaches the time used in preset value, adds the smart time by determining, during model training, additionally it is possible to
Extra garbage is rejected, the more efficient of model training is made.
Specifically, by taking amount of access as an example, it is assumed that the amount of access of a certain training sample executes when reaching 500 times plus essence is grasped
Make, at this time from the model publication moment, have passed through 40 minutes, that is, can determine that corresponding average behavior rate is 12.5 beats/min
Clock.
It should be noted that plus essence operation the moment refer to from model issue the moment, to execute plus essence operate it is experienced when
Between.
In the above-mentioned technical solutions, it is preferable that output unit is also used to: according to the characteristic information of current forum's note, obtaining
The output valve of automatic plus smart model;Managing device further include: operating unit, for detecting that output valve is smart more than or equal to adding
When threshold value, executes plus essence operates, and current forum is posted in forum's note queue and executes top set operation;Operating unit is also used to:
When detecting output valve less than smart threshold value is added, does not execute plus essence operates.
In the technical scheme, add smart threshold value by being arranged, to add automatically by the characteristic information input of current forum's note
After smart model, output valve is compared with smart threshold value is added, when output valve is greater than or equal to and adds smart threshold value, execution plus essence are grasped
Make, when output valve is less than and adds smart threshold value, does not then execute and add essence operation, under the premise of meeting reliability, calculated model
Journey is simpler, wherein smart threshold value is added to be referred to the setting of preset ratio threshold value.
The technical solution of the third aspect of the present invention provides a kind of server, comprising: memory, processor and is stored in
On memory and the computer program that can run on a processor, processor realize any of the above-described opinion when executing computer program
The step of management method of altar note limits, and/or the managing device of forum's note including any of the above-described.
The technical solution of the fourth aspect of the present invention provides a kind of computer readable storage medium, is stored thereon with calculating
Machine program realizes the step of management method of any of the above-described forum note limits when computer program is executed by processor.
Advantages of the present invention will provide in following description section, partially will become apparent from the description below, or
Practice through the invention is recognized.
Detailed description of the invention
Above-mentioned and/or additional aspect of the invention and advantage will become from the description of the embodiment in conjunction with the following figures
Obviously and it is readily appreciated that, in which:
Fig. 1 shows the schematic flow diagram of the management method of forum's note of embodiment according to the present invention;
Fig. 2 shows the schematic block diagrams of the managing device of forum's note of embodiment according to the present invention;
Fig. 3 shows the schematic block diagram of the server of embodiment according to the present invention;
Fig. 4 shows the schematic flow diagram of the model training method of embodiment according to the present invention.
Specific embodiment
To better understand the objects, features and advantages of the present invention, with reference to the accompanying drawing and specific real
Applying mode, the present invention is further described in detail.It should be noted that in the absence of conflict, the implementation of the application
Feature in example and embodiment can be combined with each other.
In the following description, numerous specific details are set forth in order to facilitate a full understanding of the present invention, still, the present invention may be used also
To be implemented using other than the one described here other modes, therefore, protection scope of the present invention is not by described below
Specific embodiment limitation.
Fig. 1 shows the schematic flow diagram of the management method of forum's note according to an embodiment of the invention.
As shown in Figure 1, the communication means of the WLAN of embodiment according to the present invention, comprising: step S102 will work as
Preceding forum's note input is automatic to add smart model;Step S104, according to the output result of automatic plus smart model, it is determined whether to current opinion
Altar note executes plus essence operation.
In this embodiment, by inputting current forum's note automatically plus in smart model, add smart model automatically with basis
Export result, it is determined whether carry out to current forum's note plus smart, on the one hand, realize the automatic plus essence function to forum's note, section
It has saved user and has added smart operation manually, improved the usage experience of user, it on the other hand, can be with by the way that automatic plus smart model is arranged
According to different setting rules, training is automatic plus smart model, to determine plus smart standard, different uses field so as to meet
It closes.
It is understood that preset automatic plus smart model, can be assessed, really based on the quality to forum's note itself
Fixed automatic plus smart model plus smart standard can also determine plus smart standard according to the feedback of user behavior, can be combined with forum
The quality of note itself and the feedback of user behavior determine and add smart standard.
In the above embodiment, it is preferable that before by the automatic plus smart model of current forum's note input, further includes: collect instruction
Practice sample set;Model training is executed to the initial model prestored according to training sample set, to be corrected in real time to initial model;
It, will when detecting that ratio that training sample is obtained according to revised initial model plus smart is greater than or equal to preset ratio threshold value
Revised initial model is determined as adding smart model automatically.
In this embodiment, by collecting training sample set, to carry out mould to preset initial model according to training sample
Type training finds the process of model parameter during model training according to given data, using training sample set as input
Value, according to the degree of optimization of output valve, corrects initial model in real time, (smart ratio is added to be greater than obtaining optimal output valve
Or it is equal to preset ratio threshold value) when, it determines corresponding mode parameter, completes training process, realize high-precision and high accuracy
Braking plus smart model method of determination, and it is adaptable.
It should be noted that the training sample for training should be with forum's note building frame having the same, it can
Characteristic information identical with forum's note is got, on the one hand the computational accuracy of automatic plus smart model depends on training sample concentration
The correlation of training sample and current forum's note, on the other hand also depends on the radix of training sample.
In addition, initial model can be generated according to one or more history essence forum's note, training sample can also be passed through
The training sample of concentration generates, and can also be generated by directly default.
For example, training sample can be the history essence note derived from this forum (for example, the car owner in a certain city encloses), it can also
To be derived from the high-quality model of other platforms of network, training sample can be determined directly by system, can also pass through artificial screening
It determines, such as some essence notes containing particular picture theme (for example, landscape theme, pet-theme) of artificial screening.
In the above embodiment, it is preferable that model training is executed to the initial model prestored according to training sample set, with
To real-time training pattern, specifically includes the following steps: extracting the characteristic information for the training sample that training sample is concentrated;It determines initial
Multiple model parameters corresponding with characteristic information in model;Multiple model parameters are corrected in real time according to characteristic information, directly
Meet to revised multiple model parameters plus smart ratio is greater than or equal to preset ratio threshold value.
In this embodiment, by extract training sample characteristic information, with according to characteristic information determine plus essence operation with
Mapping relations before characteristic information, and mapping relations are especially by the actual value of multiple model parameters corresponding with characteristic information
It determines, under the premise of given one or more specific output valves, model parameter is corrected in real time, is further improved
The output accuracy of automatic plus smart model.
Specifically, the model training according to training sample set to the initial model prestored can pass through a variety of training algorithms
It realizes, for example using GBDT (Gradient Boosting Decision Tree, gradient promote decision tree), GBDT has more
A learner, and there are sequencing between learner, each training sample has weighted value, and it is initial in training, each
The weighted value of sample is equal, is trained first using first learner to training sample, after the completion of training, increases wrong sample
This weight, while reducing the weight of correct sample, then learnt using second learner, finally obtains b study
After device, merge training result, adds smart ratio to be greater than or equal to default plus essence ratio to realize, wherein learner is specially CART
(post-class processing), is trained using CART, and the quantity for heating training sample is m, and training sample can characterize are as follows:
{(X(1),y(1)),(X(2),y(2)),…,(X(m),y(m))}
Wherein, X(i)For n-dimensional vector, the characteristic information of i-th of sample, y are characterized(i)For model parameter.
Furthermore it is also possible to execute model training, detailed process using neural network (Nerual Network) algorithm are as follows: mind
It is formed by multiple logic units according to different hierarchical organizations through network model, the input that each layer of output variable is next layer
Training sample set is configured to a training matrix, using training sample set as input value, smart ratio will be added to be greater than or wait by variable
Model training is finally completed with Optimized model parameter as output valve condition in preset ratio threshold value.
In the above embodiment, it is preferable that characteristic information includes behavioural characteristic and content characteristic, behavioural characteristic includes access
Amount, money order receipt to be signed and returned to the sender amount turn note amount, chase after patch amount and at least one in the amount of thumbing up, and when forum's note is that picture pastes, content characteristic includes scheming
At least one of in piece theme, picture number and picture clarity, wherein multiple model parameters include corresponding with behavioural characteristic
At least one first weighted value, and at least one second weighted value corresponding with content characteristic.
In this embodiment, behavioural characteristic include amount of access, money order receipt to be signed and returned to the sender amount, turn note amount, chase after patch amount and the amount of thumbing up at least
One, when forum's note is that picture pastes, content characteristic includes at least one in picture theme, picture number and picture clarity
, when behavioural characteristic includes above-mentioned all information, amount of access, money order receipt to be signed and returned to the sender amount turn that note amount, to chase after patch amount one corresponding with the amount of thumbing up
First weighted value, when content characteristic includes all of the above information, picture theme, picture number and picture clarity corresponding one
A second weighted value, the process of model training determine the process of the first weighted value and the second weighted value, according to adding smart standard
Just, it moves and adjusts preset ratio threshold value.
When adding smart standard broadest, preset ratio threshold value is 1, i.e., whole training samples are to add smart forum note.
In the above embodiment, it is preferable that being corrected in real time according to characteristic information to multiple model parameters, further includes:
When training sample includes history essence forum's note of multiple forums, determine history essence forum note adds the essence operation moment, with
And add essence operation moment corresponding behavioural characteristic;According to adding essence operation moment corresponding behavioural characteristic and add the essence operation moment,
Determine the average behavior rate of behavioural characteristic;The first weighted value is modified according to average behavior rate.
In this embodiment, by determining the flat of behavioural characteristic according to smart time and behavioural characteristic corresponding when adding smart is added
Equal behavior rate, using behavior rate as the input value being modified to the first weighted value, compared with behavioural characteristic, behavior rate
In increase measurement to the smart operating time is added, i.e., within a specified time whether the magnitude of behavioural characteristic reaches preset value, or row
The magnitude being characterized reaches the time used in preset value, adds the smart time by determining, during model training, additionally it is possible to pick
Except extra garbage, make the more efficient of model training.
Specifically, by taking amount of access as an example, it is assumed that the amount of access of a certain training sample executes when reaching 500 times plus essence is grasped
Make, at this time from the model publication moment, have passed through 40 minutes, that is, can determine that corresponding average behavior rate is 12.5 beats/min
Clock.
It should be noted that plus essence operation the moment refer to from model issue the moment, to execute plus essence operate it is experienced when
Between.
In the above embodiment, it is preferable that according to the output result of automatic plus smart model, it is determined whether to current forum's note
It executes plus essence operates, specifically includes the following steps: obtaining the output of automatic plus smart model according to the characteristic information of current forum's note
Value;It is detecting that output valve is greater than or equal to plus when smart threshold value, is executing plus essence operation, and current forum is posted on forum's note queue
Middle execution top set operation;When detecting output valve less than smart threshold value is added, does not execute plus essence operates.
In this embodiment, by setting plus smart threshold value, the characteristic information of current forum's note is being inputted automatic plus essence
After model, output valve is compared with smart threshold value is added, when output valve is greater than or equal to and adds smart threshold value, execution plus essence are operated,
When output valve is less than and adds smart threshold value, does not then execute and add essence operation, under the premise of meeting reliability, make model calculating process more
Simply, wherein smart threshold value is added to be referred to the setting of preset ratio threshold value.
Fig. 2 shows the schematic block diagrams of the managing device of forum's note of embodiment according to the present invention.
As shown in Fig. 2, the managing device 200 of forum's note of embodiment according to the present invention, comprising: input unit 202 is used
Add smart model automatically in inputting current forum's note;Output unit 204, for the output according to automatic plus smart model as a result, really
Fixed whether execute to current forum's note adds essence operation.
In this embodiment, by inputting current forum's note automatically plus in smart model, add smart model automatically with basis
Export result, it is determined whether carry out to current forum's note plus smart, on the one hand, realize the automatic plus essence function to forum's note, section
It has saved user and has added smart operation manually, improved the usage experience of user, it on the other hand, can be with by the way that automatic plus smart model is arranged
According to different setting rules, training is automatic plus smart model, to determine plus smart standard, different uses field so as to meet
It closes.
It is understood that preset automatic plus smart model, can be assessed, really based on the quality to forum's note itself
Fixed automatic plus smart model plus smart standard can also determine plus smart standard according to the feedback of user behavior, can be combined with forum
The quality of note itself and the feedback of user behavior determine and add smart standard.
In the above embodiment, it is preferable that further include: collector unit 206, for collecting training sample set;Training unit
208, for executing model training to the initial model prestored according to training sample set, to be corrected in real time to initial model;
Determination unit 210, for detecting that ratio that training sample is obtained according to revised initial model plus smart is greater than or equal to
When preset ratio threshold value, revised initial model is determined as to add smart model automatically.
In this embodiment, by collecting training sample set, to carry out mould to preset initial model according to training sample
Type training finds the process of model parameter during model training according to given data, using training sample set as input
Value, according to the degree of optimization of output valve, corrects initial model in real time, (smart ratio is added to be greater than obtaining optimal output valve
Or it is equal to preset ratio threshold value) when, it determines corresponding mode parameter, completes training process, realize high-precision and high accuracy
Braking plus smart model method of determination, and it is adaptable.
It should be noted that the training sample for training should be with forum's note building frame having the same, it can
Characteristic information identical with forum's note is got, on the one hand the computational accuracy of automatic plus smart model depends on training sample concentration
The correlation of training sample and current forum's note, on the other hand also depends on the radix of training sample.
In addition, initial model can be generated according to one or more history essence forum's note, training sample can also be passed through
The training sample of concentration generates, and can also be generated by directly default.
For example, training sample can be the history essence note derived from this forum (for example, the car owner in a certain city encloses), it can also
To be derived from the high-quality model of other platforms of network, training sample can be determined directly by system, can also pass through artificial screening
It determines, such as some essence notes containing particular picture theme (for example, landscape theme, pet-theme) of artificial screening.
In the above embodiment, it is preferable that further include: extraction unit 212, for extracting the training sample of training sample concentration
This characteristic information;Determination unit 210 is also used to: determining multiple model parameters corresponding with characteristic information in initial model;Pipe
Manage device 200 further include: amending unit 214, for being corrected in real time according to characteristic information to multiple model parameters, until repairing
Multiple model parameters after just meet plus smart ratio is greater than or equal to preset ratio threshold value.
In this embodiment, by extract training sample characteristic information, with according to characteristic information determine plus essence operation with
Mapping relations before characteristic information, and mapping relations are especially by the actual value of multiple model parameters corresponding with characteristic information
It determines, under the premise of given one or more specific output valves, model parameter is corrected in real time, is further improved
The output accuracy of automatic plus smart model.
In the above embodiment, it is preferable that characteristic information includes behavioural characteristic and content characteristic, behavioural characteristic includes access
Amount, money order receipt to be signed and returned to the sender amount turn note amount, chase after patch amount and at least one in the amount of thumbing up, and when forum's note is that picture pastes, content characteristic includes scheming
At least one of in piece theme, picture number and picture clarity, wherein multiple model parameters include corresponding with behavioural characteristic
At least one first weighted value, and at least one second weighted value corresponding with content characteristic.
Specifically, the model training according to training sample set to the initial model prestored can pass through a variety of training algorithms
It realizes, for example using GBDT (Gradient Boosting Decision Tree, gradient promote decision tree), GBDT has more
A learner, and there are sequencing between learner, each training sample has weighted value, and it is initial in training, each
The weighted value of sample is equal, is trained first using first learner to training sample, after the completion of training, increases wrong sample
This weight, while reducing the weight of correct sample, then learnt using second learner, finally obtains b study
After device, merge training result, adds smart ratio to be greater than or equal to default plus essence ratio to realize, wherein learner is specially CART
(post-class processing), is trained using CART, and the quantity for heating training sample is m, and training sample can characterize are as follows:
{(X(1),y(1)),(X(2),y(2)),…,(X(m),y(m))}
Wherein, X(i)For n-dimensional vector, the characteristic information of i-th of sample, y are characterized(i)For model parameter.
Furthermore it is also possible to execute model training, detailed process using neural network (Nerual Network) algorithm are as follows: mind
It is formed by multiple logic units according to different hierarchical organizations through network model, the input that each layer of output variable is next layer
Training sample set is configured to a training matrix, using training sample set as input value, smart ratio will be added to be greater than or wait by variable
Model training is finally completed with Optimized model parameter as output valve condition in preset ratio threshold value.
In this embodiment, behavioural characteristic include amount of access, money order receipt to be signed and returned to the sender amount, turn note amount, chase after patch amount and the amount of thumbing up at least
One, when forum's note is that picture pastes, content characteristic includes at least one in picture theme, picture number and picture clarity
, when behavioural characteristic includes above-mentioned all information, amount of access, money order receipt to be signed and returned to the sender amount turn that note amount, to chase after patch amount one corresponding with the amount of thumbing up
First weighted value, when content characteristic includes all of the above information, picture theme, picture number and picture clarity corresponding one
A second weighted value, the process of model training determine the process of the first weighted value and the second weighted value, according to adding smart standard
Just, it moves and adjusts preset ratio threshold value.
When adding smart standard broadest, preset ratio threshold value is 1, i.e., whole training samples are to add smart forum note.
In the above embodiment, it is preferable that determination unit 210 is also used to: including the history of multiple forums in training sample
When essence forum note, determine history essence forum note adds the essence operation moment, and essence is added to operate moment corresponding behavioural characteristic;
Determination unit 210 is also used to: according to adding essence operation moment corresponding behavioural characteristic and adding the essence operation moment, determining behavioural characteristic
Average behavior rate;Amending unit 214 is also used to: being modified according to average behavior rate to the first weighted value.
In this embodiment, by determining the flat of behavioural characteristic according to smart time and behavioural characteristic corresponding when adding smart is added
Equal behavior rate, using behavior rate as the input value being modified to the first weighted value, compared with behavioural characteristic, behavior rate
In increase measurement to the smart operating time is added, i.e., within a specified time whether the magnitude of behavioural characteristic reaches preset value, or row
The magnitude being characterized reaches the time used in preset value, adds the smart time by determining, during model training, additionally it is possible to pick
Except extra garbage, make the more efficient of model training.
Specifically, by taking amount of access as an example, it is assumed that the amount of access of a certain training sample executes when reaching 500 times plus essence is grasped
Make, at this time from the model publication moment, have passed through 40 minutes, that is, can determine that corresponding average behavior rate is 12.5 beats/min
Clock.
It should be noted that plus essence operation the moment refer to from model issue the moment, to execute plus essence operate it is experienced when
Between.
In the above embodiment, it is preferable that output unit 204 is also used to: according to the characteristic information of current forum's note, obtaining
The output valve of automatic plus smart model;Managing device 200 further include: operating unit 216, for detecting that output valve is greater than or waits
When Yu Jiajing threshold value, executes plus essence operates, and current forum is posted in forum's note queue and executes top set operation;Operating unit
216 are also used to: when detecting output valve less than smart threshold value is added, not executing plus essence operates.
In this embodiment, by setting plus smart threshold value, the characteristic information of current forum's note is being inputted automatic plus essence
After model, output valve is compared with smart threshold value is added, when output valve is greater than or equal to and adds smart threshold value, execution plus essence are operated,
When output valve is less than and adds smart threshold value, does not then execute and add essence operation, under the premise of meeting reliability, make model calculating process more
Simply, wherein smart threshold value is added to be referred to the setting of preset ratio threshold value.
Fig. 3 shows the schematic block diagram of the server of embodiment according to the present invention.
As shown in figure 3, the server 300 of embodiment according to the present invention, comprising: memory 302, processor 304 and deposit
The computer program that can be run on memory 302 and on a processor is stored up, processor is realized above-mentioned when executing computer program
The step of management method of any one forum note limits, and/or the managing device 200 of forum's note including any of the above-described.
According to an embodiment of the invention, also proposed a kind of computer readable storage medium, it is stored thereon with computer journey
Realization when computer program is executed by processor: the input of current forum's note is added smart model by sequence automatically;According to automatic plus smart model
Output result, it is determined whether to current forum's note execute plus essence operation.
In this embodiment, by inputting current forum's note automatically plus in smart model, add smart model automatically with basis
Export result, it is determined whether carry out to current forum's note plus smart, on the one hand, realize the automatic plus essence function to forum's note, section
It has saved user and has added smart operation manually, improved the usage experience of user, it on the other hand, can be with by the way that automatic plus smart model is arranged
According to different setting rules, training is automatic plus smart model, to determine plus smart standard, different uses field so as to meet
It closes.
It is understood that preset automatic plus smart model, can be assessed, really based on the quality to forum's note itself
Fixed automatic plus smart model plus smart standard can also determine plus smart standard according to the feedback of user behavior, can be combined with forum
The quality of note itself and the feedback of user behavior determine and add smart standard.
In the above embodiment, it is preferable that before by the automatic plus smart model of current forum's note input, further includes: collect instruction
Practice sample set;Model training is executed to the initial model prestored according to training sample set, to be corrected in real time to initial model;
It, will when detecting that ratio that training sample is obtained according to revised initial model plus smart is greater than or equal to preset ratio threshold value
Revised initial model is determined as adding smart model automatically.
In this embodiment, by collecting training sample set, to carry out mould to preset initial model according to training sample
Type training finds the process of model parameter during model training according to given data, using training sample set as input
Value, according to the degree of optimization of output valve, corrects initial model in real time, (smart ratio is added to be greater than obtaining optimal output valve
Or it is equal to preset ratio threshold value) when, it determines corresponding mode parameter, completes training process, realize high-precision and high accuracy
Braking plus smart model method of determination, and it is adaptable.
It should be noted that the training sample for training should be with forum's note building frame having the same, it can
Characteristic information identical with forum's note is got, on the one hand the computational accuracy of automatic plus smart model depends on training sample concentration
The correlation of training sample and current forum's note, on the other hand also depends on the radix of training sample.
In addition, initial model can be generated according to one or more history essence forum's note, training sample can also be passed through
The training sample of concentration generates, and can also be generated by directly default.
For example, training sample can be the history essence note derived from this forum (for example, the car owner in a certain city encloses), it can also
To be derived from the high-quality model of other platforms of network, training sample can be determined directly by system, can also pass through artificial screening
It determines, such as some essence notes containing particular picture theme (for example, landscape theme, pet-theme) of artificial screening.
In the above embodiment, it is preferable that model training is executed to the initial model prestored according to training sample set, with
To real-time training pattern, specifically includes the following steps: extracting the characteristic information for the training sample that training sample is concentrated;It determines initial
Multiple model parameters corresponding with characteristic information in model;Multiple model parameters are corrected in real time according to characteristic information, directly
Meet to revised multiple model parameters plus smart ratio is greater than or equal to preset ratio threshold value.
In this embodiment, by extract training sample characteristic information, with according to characteristic information determine plus essence operation with
Mapping relations before characteristic information, and mapping relations are especially by the actual value of multiple model parameters corresponding with characteristic information
It determines, under the premise of given one or more specific output valves, model parameter is corrected in real time, is further improved
The output accuracy of automatic plus smart model.
Specifically, the model training according to training sample set to the initial model prestored can pass through a variety of training algorithms
It realizes, for example using GBDT (Gradient Boosting Decision Tree, gradient promote decision tree), GBDT has more
A learner, and there are sequencing between learner, each training sample has weighted value, and it is initial in training, each
The weighted value of sample is equal, is trained first using first learner to training sample, after the completion of training, increases wrong sample
This weight, while reducing the weight of correct sample, then learnt using second learner, finally obtains b study
After device, merge training result, adds smart ratio to be greater than or equal to default plus essence ratio to realize, wherein learner is specially CART
(post-class processing), is trained using CART, and the quantity for heating training sample is m, and training sample can characterize are as follows:
{(X(1),y(1)),(X(2),y(2)),…,(X(m),y(m))}
Wherein, X(i)For n-dimensional vector, the characteristic information of i-th of sample, y are characterized(i)For model parameter.
Furthermore it is also possible to execute model training, detailed process using neural network (Nerual Network) algorithm are as follows: mind
It is formed by multiple logic units according to different hierarchical organizations through network model, the input that each layer of output variable is next layer
Training sample set is configured to a training matrix, using training sample set as input value, smart ratio will be added to be greater than or wait by variable
Model training is finally completed with Optimized model parameter as output valve condition in preset ratio threshold value.
In the above embodiment, it is preferable that characteristic information includes behavioural characteristic and content characteristic, behavioural characteristic includes access
Amount, money order receipt to be signed and returned to the sender amount turn note amount, chase after patch amount and at least one in the amount of thumbing up, and when forum's note is that picture pastes, content characteristic includes scheming
At least one of in piece theme, picture number and picture clarity, wherein multiple model parameters include corresponding with behavioural characteristic
At least one first weighted value, and at least one second weighted value corresponding with content characteristic.
In this embodiment, behavioural characteristic include amount of access, money order receipt to be signed and returned to the sender amount, turn note amount, chase after patch amount and the amount of thumbing up at least
One, when forum's note is that picture pastes, content characteristic includes at least one in picture theme, picture number and picture clarity
, when behavioural characteristic includes above-mentioned all information, amount of access, money order receipt to be signed and returned to the sender amount turn that note amount, to chase after patch amount one corresponding with the amount of thumbing up
First weighted value, when content characteristic includes all of the above information, picture theme, picture number and picture clarity corresponding one
A second weighted value, the process of model training determine the process of the first weighted value and the second weighted value, according to adding smart standard
Just, it moves and adjusts preset ratio threshold value.
When adding smart standard broadest, preset ratio threshold value is 1, i.e., whole training samples are to add smart forum note.
In the above embodiment, it is preferable that being corrected in real time according to characteristic information to multiple model parameters, further includes:
When training sample includes history essence forum's note of multiple forums, determine history essence forum note adds the essence operation moment, with
And add essence operation moment corresponding behavioural characteristic;According to adding essence operation moment corresponding behavioural characteristic and add the essence operation moment,
Determine the average behavior rate of behavioural characteristic;The first weighted value is modified according to average behavior rate.
In this embodiment, by determining the flat of behavioural characteristic according to smart time and behavioural characteristic corresponding when adding smart is added
Equal behavior rate, using behavior rate as the input value being modified to the first weighted value, compared with behavioural characteristic, behavior rate
In increase measurement to the smart operating time is added, i.e., within a specified time whether the magnitude of behavioural characteristic reaches preset value, or row
The magnitude being characterized reaches the time used in preset value, adds the smart time by determining, during model training, additionally it is possible to pick
Except extra garbage, make the more efficient of model training.
Specifically, by taking amount of access as an example, it is assumed that the amount of access of a certain training sample executes when reaching 500 times plus essence is grasped
Make, at this time from the model publication moment, have passed through 40 minutes, that is, can determine that corresponding average behavior rate is 12.5 beats/min
Clock.
It should be noted that plus essence operation the moment refer to from model issue the moment, to execute plus essence operate it is experienced when
Between.
In the above embodiment, it is preferable that according to the output result of automatic plus smart model, it is determined whether to current forum's note
It executes plus essence operates, specifically includes the following steps: obtaining the output of automatic plus smart model according to the characteristic information of current forum's note
Value;It is detecting that output valve is greater than or equal to plus when smart threshold value, is executing plus essence operation, and current forum is posted on forum's note queue
Middle execution top set operation;When detecting output valve less than smart threshold value is added, does not execute plus essence operates.
In this embodiment, by setting plus smart threshold value, the characteristic information of current forum's note is being inputted automatic plus essence
After model, output valve is compared with smart threshold value is added, when output valve is greater than or equal to and adds smart threshold value, execution plus essence are operated,
When output valve is less than and adds smart threshold value, does not then execute and add essence operation, under the premise of meeting reliability, make model calculating process more
Simply, wherein smart threshold value is added to be referred to the setting of preset ratio threshold value.
Fig. 4 shows the schematic flow diagram of the model training method of embodiment according to the present invention.
As shown in figure 4, the model training method of embodiment according to the present invention, comprising: step 402, collect training sample
Collection;Step 404, the characteristic information for the training sample that training sample is concentrated is extracted;Step 406, according to characteristic information to multiple moulds
Shape parameter is corrected in real time;Step 408, ratio that training sample set is obtained according to revised initial model plus smart is being detected
When example is greater than or equal to preset ratio threshold value, revised initial model is determined as to add smart model automatically.
Have been described in detail above with reference to the accompanying drawings the embodiment of the present invention, it is contemplated that the manual plus smart manpower that the relevant technologies propose
The technical problems such as low efficiency at high cost, the invention proposes a kind of management methods of forum's note, by inputting current forum's note
In automatic plus smart model, with the output result of the automatic plus smart model of basis, it is determined whether carried out to current forum's note plus smart, a side
Face realizes the automatic plus essence function to forum's note, saves user and add smart operation manually, improve the usage experience of user,
On the other hand, add smart model automatically by being arranged, can be according to different setting rules, training adds smart model automatically, with determination
Add smart standard, so as to meet different use occasions.
The foregoing is only a preferred embodiment of the present invention, is not intended to restrict the invention, for the skill of this field
For art personnel, the invention may be variously modified and varied.All within the spirits and principles of the present invention, made any to repair
Change, equivalent replacement, improvement etc., should all be included in the protection scope of the present invention.
Claims (14)
1. a kind of management method of forum's note is suitable for server characterized by comprising
The input of current forum's note is added into smart model automatically;
According to the output result of the automatic plus smart model, it is determined whether executed to current forum's note and add essence operation.
2. management method according to claim 1, which is characterized in that described that the input of current forum's note is being added smart mould automatically
Before type, further includes:
Collect training sample set;
Model training is executed to the initial model prestored according to the training sample set, to be repaired in real time to the initial model
Just;
Detecting that it is pre- that ratio that the training sample set is obtained according to the revised initial model plus smart is greater than or equal to
If when proportion threshold value, the revised initial model is determined as described automatic plus smart model.
3. management method according to claim 2, which is characterized in that described first to what is prestored according to the training sample set
Beginning model executes model training, to obtain real-time training pattern, specifically includes the following steps:
Extract the characteristic information for the training sample that the training sample is concentrated;
Determine multiple model parameters corresponding with the characteristic information in the initial model;
The multiple model parameter is corrected in real time according to the characteristic information, until revised the multiple model ginseng
Number meets described plus smart ratio and is greater than or equal to the preset ratio threshold value.
4. management method according to claim 3, which is characterized in that
The characteristic information includes behavioural characteristic and content characteristic, the behavioural characteristic include amount of access, money order receipt to be signed and returned to the sender amount, turn note amount,
At least one in patch amount and the amount of thumbing up is chased after, when forum's note is that picture pastes, the content characteristic includes picture theme, figure
At least one of in piece quantity and picture clarity, wherein the multiple model parameter includes corresponding with the behavioural characteristic
At least one first weighted value, and at least one second weighted value corresponding with the content characteristic.
5. management method according to claim 4, which is characterized in that it is described according to the characteristic information to the multiple mould
Shape parameter is corrected in real time, further includes:
When the training sample includes history essence forum's note of multiple forums, determine history essence forum note adds essence
Operate moment and the described plus essence operation moment corresponding behavioural characteristic;
According to the described plus essence operation moment corresponding behavioural characteristic and described plus essence operation moment, determine that the behavior is special
The behavior rate feature of sign;
First weighted value is modified according to the behavior rate feature.
6. management method according to any one of claim 3 to 5, which is characterized in that described according to described automatic plus essence
The output result of model, it is determined whether current forum's note is executed plus essence operates, specifically includes the following steps:
According to the characteristic information of current forum's note, the output valve of the automatic plus smart model is obtained;
It is detecting that the output valve is greater than or equal to plus when smart threshold value, is executing described plus essence operation, and by the current forum
It is posted in forum's note queue and executes top set operation;
When detecting that the output valve is less than described plus smart threshold value, described plus essence operation is not executed.
7. a kind of managing device of forum's note is suitable for server characterized by comprising
Input unit, for the input of current forum's note to be added smart model automatically;
Output unit, for the output result according to the automatic plus smart model, it is determined whether executed to current forum's note
Add smart operation.
8. managing device according to claim 7, which is characterized in that further include:
Collector unit, for collecting training sample set;
Training unit, for executing model training to the initial model prestored according to the training sample set, to described initial
Model is corrected in real time;
Determination unit, for detecting ratio that the training sample set is obtained according to the revised initial model plus smart
When more than or equal to preset ratio threshold value, the revised initial model is determined as described automatic plus smart model.
9. managing device according to claim 8, which is characterized in that further include:
Extraction unit, for extracting the characteristic information for the training sample that the training sample is concentrated;
The determination unit is also used to: determining multiple model parameters corresponding with the characteristic information in the initial model;
The managing device is also reported:
Amending unit, for being corrected in real time according to the characteristic information to the multiple model parameter, until revised
The multiple model parameter meets described plus smart ratio and is greater than or equal to the preset ratio threshold value.
10. managing device according to claim 9, which is characterized in that
The characteristic information includes behavioural characteristic and content characteristic, the behavioural characteristic include amount of access, money order receipt to be signed and returned to the sender amount, turn note amount,
At least one in patch amount and the amount of thumbing up is chased after, when forum's note is that picture pastes, the content characteristic includes picture theme, figure
At least one of in piece quantity and picture clarity, wherein the multiple model parameter includes corresponding with the behavioural characteristic
At least one first weighted value, and at least one second weighted value corresponding with the content characteristic.
11. managing device according to claim 10, which is characterized in that
The determination unit is also used to: when the training sample includes history essence forum's note of multiple forums, described in determination
History essence forum note plus smart operation moment and the described plus essence operation moment corresponding behavioural characteristic;
The determination unit is also used to: according to the described plus essence operation moment corresponding behavioural characteristic and described plus essence operation
Moment determines the behavior rate feature of the behavioural characteristic;
The amending unit is also used to: being modified according to the behavior rate feature to first weighted value.
12. the managing device according to any one of claim 9 to 11, which is characterized in that the output unit is also used to:
According to the characteristic information of current forum's note, the output valve of the automatic plus smart model is obtained;
The managing device further include:
Operating unit, for detecting that the output valve is greater than or equal to plus when smart threshold value, executes described plus essence operation, and will
The current forum, which is posted in forum's note queue, executes top set operation;
The operating unit is also used to: when detecting that the output valve is less than described plus smart threshold value, not executing described plus essence behaviour
Make.
13. a kind of server, including memory, processor and it is stored on the memory and can runs on the processor
Computer program, which is characterized in that the processor is realized when executing the computer program as appointed in claim 1 to 6
The step of one management method limits, and/or include the managing device as described in any one of claim 7 to 12.
14. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program
The step of limiting such as any one of claims 1 to 6 management method is realized when being executed by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711100303.0A CN110069617A (en) | 2017-11-09 | 2017-11-09 | Management method, device, server and the computer readable storage medium of forum's note |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711100303.0A CN110069617A (en) | 2017-11-09 | 2017-11-09 | Management method, device, server and the computer readable storage medium of forum's note |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110069617A true CN110069617A (en) | 2019-07-30 |
Family
ID=67364556
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711100303.0A Pending CN110069617A (en) | 2017-11-09 | 2017-11-09 | Management method, device, server and the computer readable storage medium of forum's note |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110069617A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111028584A (en) * | 2019-12-31 | 2020-04-17 | 北京网梯科技发展有限公司 | User space-based personalized teaching interaction method and device |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101482888A (en) * | 2009-02-23 | 2009-07-15 | 阿里巴巴集团控股有限公司 | Website caller value computing method and system |
CN105279556A (en) * | 2015-11-05 | 2016-01-27 | 国家卫星海洋应用中心 | Enteromorpha detection method and enteromorpha detection device |
CN106294363A (en) * | 2015-05-15 | 2017-01-04 | 厦门美柚信息科技有限公司 | A kind of forum postings evaluation methodology, Apparatus and system |
CN106682118A (en) * | 2016-12-08 | 2017-05-17 | 华中科技大学 | Social network site false fan detection method achieved on basis of network crawler by means of machine learning |
CN106777132A (en) * | 2016-12-18 | 2017-05-31 | 深圳市辣妈帮科技有限公司 | Data processing method and device |
CN106919579A (en) * | 2015-12-24 | 2017-07-04 | 腾讯科技(深圳)有限公司 | A kind of information processing method and device, equipment |
-
2017
- 2017-11-09 CN CN201711100303.0A patent/CN110069617A/en active Pending
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101482888A (en) * | 2009-02-23 | 2009-07-15 | 阿里巴巴集团控股有限公司 | Website caller value computing method and system |
CN106294363A (en) * | 2015-05-15 | 2017-01-04 | 厦门美柚信息科技有限公司 | A kind of forum postings evaluation methodology, Apparatus and system |
CN105279556A (en) * | 2015-11-05 | 2016-01-27 | 国家卫星海洋应用中心 | Enteromorpha detection method and enteromorpha detection device |
CN106919579A (en) * | 2015-12-24 | 2017-07-04 | 腾讯科技(深圳)有限公司 | A kind of information processing method and device, equipment |
CN106682118A (en) * | 2016-12-08 | 2017-05-17 | 华中科技大学 | Social network site false fan detection method achieved on basis of network crawler by means of machine learning |
CN106777132A (en) * | 2016-12-18 | 2017-05-31 | 深圳市辣妈帮科技有限公司 | Data processing method and device |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111028584A (en) * | 2019-12-31 | 2020-04-17 | 北京网梯科技发展有限公司 | User space-based personalized teaching interaction method and device |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103853786B (en) | The optimization method and system of database parameter | |
CN108665159A (en) | A kind of methods of risk assessment, device, terminal device and storage medium | |
CN108154430A (en) | A kind of credit scoring construction method based on machine learning and big data technology | |
CN110175850B (en) | Transaction information processing method and device | |
CN109271374A (en) | A kind of database health scoring method and scoring system based on machine learning | |
CN110222722A (en) | Interactive image stylization processing method, calculates equipment and storage medium at system | |
CN105574680A (en) | Diagnosis method and system of enterprise competitiveness | |
CN108898476A (en) | A kind of loan customer credit-graded approach and device | |
CN106651097A (en) | Data collection method, data collection device and data collection server based on crowd sourcing | |
CN105786711A (en) | Data analysis method and device | |
CN110110663A (en) | A kind of age recognition methods and system based on face character | |
CN109033220A (en) | Automatically selecting method, system, equipment and the storage medium of labeled data | |
CN110516537A (en) | A kind of face age estimation method based on from step study | |
CN108734361A (en) | Share-car order processing method and apparatus | |
CN106778791A (en) | A kind of timber visual identity method based on multiple perceptron | |
CN112163671A (en) | New energy scene generation method and system | |
CN114358445A (en) | Business process residual time prediction model recommendation method and system | |
CN115544873A (en) | Training efficiency and personalized effect quantitative evaluation method for personalized federal learning | |
CN104657390B (en) | A kind of answer platform method and system | |
CN109272312A (en) | Method and apparatus for transaction risk detecting real-time | |
CN110069617A (en) | Management method, device, server and the computer readable storage medium of forum's note | |
CN113516192A (en) | Method, system, device and storage medium for identifying user electricity consumption transaction | |
CN109840308B (en) | Regional wind power probability forecasting method and system | |
CN108846746A (en) | A kind of carbon transaction behavior modeling method of combination discrete statistics and extreme learning machine | |
AU2020335019B2 (en) | Evaluation method based on mobile news client and system thereof |
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 | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190730 |
|
RJ01 | Rejection of invention patent application after publication |