CN109919314A - Processing result prediction technique, device and server based on prediction model - Google Patents
Processing result prediction technique, device and server based on prediction model Download PDFInfo
- Publication number
- CN109919314A CN109919314A CN201910121267.9A CN201910121267A CN109919314A CN 109919314 A CN109919314 A CN 109919314A CN 201910121267 A CN201910121267 A CN 201910121267A CN 109919314 A CN109919314 A CN 109919314A
- Authority
- CN
- China
- Prior art keywords
- processing result
- target network
- data
- network
- prediction
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Computer And Data Communications (AREA)
Abstract
The embodiment of the invention discloses a kind of processing result prediction technique, device and server based on prediction model, this method comprises: obtaining target network in the processing result set of preceding T t raining period, the processing result set includes that the target network is directed to the t1 processing result data that training data is handled and carries out the t2 processing result data that processing result is predicted for the target network using processing result prediction model, described T, t1 are positive integer, and the t2 is nonnegative integer；The processing result data in the processing result set is handled using the processing result prediction model, prediction obtains the target network in the processing result data of the T+1 t raining period；The processing result prediction model is that the network setup information based on the target network constructs.Using the embodiment of the present invention, the processing result of network can be predicted, to effectively improve the efficiency of the processing result of determining network.
Description
Technical field
The present invention relates to field of computer technology more particularly to a kind of processing result prediction technique based on prediction model,
Device and server.
Background technique
In planned network model, either engineer's network model or computer Automated Design network model (or
Person says that the network architecture is searched for), it usually needs a large amount of sample data is inputted in network model after being handled, can just obtain net
The processing result of network model, processing result for example can serve to indicate that the classification accuracy etc. of network model.But above-mentioned determining net
It is big that the mode of network model treatment result handles data volume, needs a large amount of computing resource and processing time, causes to determine network mould
The low efficiency of type processing result.Therefore, the efficiency for how effectively improving determining network model processing result is that have to be solved ask
Topic.
Summary of the invention
The embodiment of the present invention provides a kind of processing result prediction technique, device and server based on prediction model, can be with
The processing result of network is predicted, to effectively improve the efficiency of the processing result of determining network.
In a first aspect, the embodiment of the invention provides a kind of processing result prediction technique based on prediction model, comprising:
Target network is obtained in the processing result set of preceding T t raining period, the processing result set includes the mesh
Mark network is directed to the t1 processing result data and be directed to using processing result prediction model that training data is handled
The target network carries out the t2 processing result data that processing result is predicted, described T, t1 are positive integer, and the t2 is
Nonnegative integer；
The processing result data in the processing result set is handled using the processing result prediction model, in advance
The target network is measured in the processing result data of the T+1 t raining period；
Wherein, the processing result prediction model is that the network setup information based on the target network constructs.
In one embodiment, the method also includes:
The network setup information of the target network is obtained, the network setup information includes the structure of the target network
Parameter and hyper parameter；
Target signature parameter is determined according to the network setup information, and constructs to obtain institute according to the target signature parameter
State processing result prediction model.
In one embodiment, it is described according to the target signature parameter construct to obtain the processing result prediction model it
Afterwards, the method also includes:
The processing result prediction model is trained using Support vector regression LIBSVM software, after being trained
Processing result prediction model；
Wherein, the corresponding support vector machines type of the LIBSVM software is Support Vector Machines for Regression v-SVR, institute
Stating the corresponding kernel function of LIBSVM software is radial basis function RBF.
In one embodiment, the processing result data is used to indicate the classification accuracy of the target network, described
The processing result data in the processing result set is handled using the processing result prediction model, prediction obtains institute
Target network is stated after the processing result data of the T+1 t raining period, the method also includes:
Target network classification indicated by the processing result data of the T+1 t raining period that detection prediction obtains
Whether accuracy rate is less than default value；
If target network classification accuracy indicated by the processing result data of the T+1 t raining period is less than institute
Default value is stated, then obtains sample data, and be trained to the target network using the sample data.
In one embodiment, it is described using the processing result prediction model to the processing in the processing result set
Result data is handled, and prediction obtains the target network after the processing result data of the T+1 t raining period, described
Method further include:
Output includes the prompt information of processing result data of the target network in the T+1 t raining period, described to mention
Show information for prompt whether the target network is trained using sample data；
If detecting for the confirmation training instruction of prompt information input, the sample data is obtained, and utilize
The sample data is trained the target network.
In one embodiment, the corresponding processing result data of each t raining period of the target network, the T
Equal to the t1 and it is described t2's and.
In one embodiment, the target network is neural network.
Second aspect, the embodiment of the invention provides a kind of the processing result prediction meanss based on prediction model, the processing
Prediction of result device includes for executing the described in any item processing result prediction sides based on prediction model of above-mentioned first aspect
The unit of method.
The third aspect, the embodiment of the invention provides a kind of server, including processor, communication interface and memory, institutes
Processor, the communication interface and the memory to be stated to be connected with each other, wherein the memory is used to store computer program,
The computer program includes program instruction, and the processor is configured for calling described program instruction, executes above-mentioned first
The described in any item processing result prediction techniques based on prediction model of aspect.
Fourth aspect, the embodiment of the invention provides a kind of computer readable storage medium, the computer-readable storage
Media storage has computer program, and the computer program includes program instruction, and described program instructs when being executed by a processor
The processor is set to execute the described in any item processing result prediction techniques based on prediction model of above-mentioned first aspect.
The embodiment of the present invention is by obtaining processing result set of the target network in preceding T t raining period, processing result collection
Close includes that target network is directed to the t1 processing result data that training data is handled and is predicted using processing result
Model carries out the t2 processing result data that processing result is predicted for target network, then predicts mould using processing result
Type handles the processing result data in processing result set, and prediction obtains target network in the T+1 t raining period
Processing result data predicts so as to the processing result to network, to effectively improve the processing result of determining network
Efficiency.
Detailed description of the invention
Technical solution in order to illustrate the embodiments of the present invention more clearly, below will be to needed in embodiment description
Attached drawing is briefly described, it should be apparent that, drawings in the following description are some embodiments of the invention, general for this field
For logical technical staff, without creative efforts, it is also possible to obtain other drawings based on these drawings.
Fig. 1 is a kind of process signal of processing result prediction technique based on prediction model provided in an embodiment of the present invention
Figure；
Fig. 2 is a kind of structural representation of processing result prediction meanss based on prediction model provided in an embodiment of the present invention
Figure；
Fig. 3 is a kind of structural schematic diagram of server provided in an embodiment of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description.
Referring to Figure 1, Fig. 1 is a kind of processing result prediction technique based on prediction model provided in an embodiment of the present invention
Flow diagram.Specifically, as shown in Figure 1, the processing result prediction technique may comprise steps of:
S101, server obtain target network in the processing result set of preceding T t raining period.
In the embodiment of the present invention, target network includes: that target network is directed in the processing result set of preceding T t raining period
T1 processing result data that training data is handled and using processing result prediction model for target network into
The t2 processing result data that row processing result is predicted.Wherein, T and t1 is positive integer, that is, is greater than 0 integer；T2 is
Nonnegative integer, i.e. positive integer and 0.When t2 is 0, target network is only included in the processing result set and is carried out for training data
T1 obtained processing result data is handled, does not carry out processing result for target network including the use of processing result prediction model
Predict obtained processing result data.
In one embodiment, the corresponding processing result data of each t raining period of target network, then T is equal to t1
With t2's and.T1 t raining period of the target network in preceding T t raining period is handled to obtain t1 for training data
Processing result data.It can use yt1Target network is indicated in the processing result data of the t1 t raining period, for having trained
The target network in t1 period can recorde lower y (t1)=y1,y2,y3,…,yt1.Server by utilizing processing result prediction model
Processing result prediction is carried out for target network, prediction obtains the corresponding t2 processing result data of t2 t raining period.It can be with
Utilize xt2Indicate the target network predicted using processing result prediction model in the processing result number of the t2 t raining period
According to can recorde lower x (t2)=x1,x2,x3,…,xt2.Wherein, which is later than this t1 training
The corresponding time in period.The t2 processing result data is that processing result prediction model is based on the t1 processing result data prediction
It obtains.
In one embodiment, processing result data can serve to indicate that target network in the performance results of t raining period,
The performance results can be classification accuracy.Target network carries out classification processing for training data in t raining period, is divided
Class processing result；It is then based on the classification accuracy that the classification processing result determines target network, and the classification that will be determined
Accuracy rate is determined as target network in the corresponding processing result data of t raining period.Server by utilizing processing result prediction model needle
Classification accuracy prediction is carried out to target network, and the classification accuracy that prediction obtains is determined as target network in t raining period
Corresponding processing result data.
In the embodiment of the present invention, processing result prediction model is that the network setup information based on target network constructs to obtain
's.Specifically, server obtains the network setup information of target network, which includes the structure ginseng of target network
Several and hyper parameter.In one embodiment, target network can be neural network, which includes the convolution of target network
Layer parameter and full connection layer parameter etc., can specifically include the weight parameter of each convolutional layer, the number of plies of convolutional layer, each convolution
The neuron number etc. of layer.The hyper parameter includes learning rate, Learning Step, the weight decaying of target network, input network
Whether chip size uses data to enhance (such as indicating with 0 or 1) etc..Then server is believed according to the network settings of target network
Breath determines target signature parameter, and is constructed to obtain the corresponding processing result prediction mould of target network according to the target signature parameter
Type.In one embodiment, server can directly determine the structural parameters and hyper parameter that the network setup information includes
For the corresponding target signature parameter of processing result prediction model.
Further, server constructs to obtain the corresponding processing result prediction mould of target network according to the target signature parameter
After type, the processing result prediction model can be trained using Support Vector Machines for Regression v-SVR model, be trained
Processing result prediction model afterwards.It can be rapidly completed using Support Vector Machines for Regression v-SVR model pre- for processing result
Survey the training of model.Specifically, server can use Support vector regression LIBSVM software and predict mould to the processing result
Type is trained, the processing result prediction model after being trained.Wherein it is possible to the specified corresponding supporting vector of LIBSVM software
Machine SVM type is Support Vector Machines for Regression v-SVR, and specifying the corresponding kernel function of LIBSVM software is radial basis function
(Radial Basis Function, RBF)；The corresponding categorical data of LIBSVM software then can be according to the network of target network
Setting information determination obtains.In one embodiment, it also can use bayes method to carry out the processing result prediction model
Training, the processing result prediction model after being trained.
S102, the server by utilizing processing result prediction model are to the processing result data in the processing result set
It is handled, prediction obtains the target network in the processing result data of the T+1 t raining period.
In the embodiment of the present invention, server by utilizing processing result prediction model is to the target network needle in processing result set
The t1 processing result data and utilization processing result prediction model handled training data is for target network
It carries out the t2 processing result data that processing result is predicted to be handled, prediction obtains target network in the T+1 training
The processing result data in period.
As an example it is assumed that numerical value indicated by t2 is 0, then it represents that only include target network needle in the processing result set
To the t1 processing result data that training data is handled, preceding T t raining period is using training data to target
The period that network is trained.Server can use processing result prediction model to target network in preceding T t raining period pair
The t1 processing result data answered is handled, and prediction obtains target network in the processing result data of the T+1 t raining period；
Then using processing result prediction model to target network t1 corresponding processing result data of preceding T t raining period, with
And the target network predicted using processing result prediction model is at the processing result data of the T+1 t raining period
Reason, prediction obtain target network in the processing result data of the T+2 t raining period.Predict target network in T+3, T+
4......T+n the processing result data of a t raining period then can and so on, details are not described herein again.N is positive integer.Using upper
Mode is stated, only needs target network to be handled in preceding T t raining period for training data, obtains t1 processing result number
According to；Processing result for target network in rear n t raining period can then be based on processing result prediction model and the t1
Processing result data is predicted to obtain.Not only can quick predict obtain target network in the processing knot of rear n t raining period
Fruit data can also greatly reduce the number being trained using training data to target network, so as to reduce at data
Reason amount achievees the purpose that save computing resource and time, effectively improves the efficiency of the processing result of determining network.
In one embodiment, processing result data is used to indicate the classification accuracy of target network, processing result prediction
Expression formula indicated by model is as follows: S=f (w, hp, S_mean, S_sd, S ', S ").Wherein, S indicates the target network of prediction
In the classification accuracy of the T+1 t raining period, w, hp, S_mean, S_sd, S ', S " indicates input item；W represents target network
The weight information of network is each layer of target network of convolution kernel weight；Hp represents the hyper parameter of target network；S_mean is target
Network specifically can be in preceding T t raining period in the mean value of the classification accuracy of preceding T t raining period according to the training time
Sequencing sequence, comes the mean value of the classification accuracy of last preset quantity (such as 3) a t raining period；S_sd is preceding T
The standard deviation of the processing result data of a t raining period；S ' is to arrange in preceding T t raining period according to training time sequencing
Sequence comes the mean value of the changing value of the classification accuracy of last preset quantity t raining period；S " includes preceding T t raining period
In sort according to training time sequencing, come that t raining period last is opposite to come penultimate t raining period institute
The classification accuracy of promotion；And it comes penultimate t raining period and comes what t raining period third from the bottom was promoted relatively
Classification accuracy.
In one embodiment, processing result data is used to indicate the classification accuracy of target network, at server by utilizing
Reason prediction of result model handles the processing result data in processing result set, and prediction obtains target network in T+1
After the processing result data of a t raining period, detection predicts obtained target network in the processing knot of the T+1 t raining period
Whether classification accuracy indicated by fruit data is less than default value, which is, for example, 90%；If detecting target network
Network classification accuracy indicated by the processing result data of the T+1 t raining period is less than the default value, then obtains sample
Data, and target network is trained using the sample data, it is adjusted with the parameter to target network, to improve target
The classification accuracy of network.Wherein, which can be identical as above-mentioned training data, can also be with above-mentioned training data not
Together.
In another embodiment, server by utilizing processing result prediction model is to the processing result in processing result set
Data are handled, prediction obtain target network after the processing result data of the T+1 t raining period, by with server
The terminal output for establishing communication connection includes prompt information of the target network in the processing result data of the T+1 t raining period,
Whether the prompt information is trained target network using sample data for warning terminal user.If terminal detects terminal
User is directed to the confirmation training instruction of prompt information input, then sends the confirmation training instruction to server.Server receives
And respond the confirmation training instruction and obtain sample data, and target network is trained using the sample data, to target
The parameter of network is adjusted, to improve the accuracy rate of the processing result of target network.
The embodiment of the present invention is by obtaining processing result set of the target network in preceding T t raining period, processing result collection
Close includes that target network is directed to the t1 processing result data that training data is handled and is predicted using processing result
Model carries out the t2 processing result data that processing result is predicted for target network, then predicts mould using processing result
Type handles the processing result data in processing result set, and prediction obtains target network in the T+1 t raining period
Processing result data predicts so as to the processing result to network, to effectively improve the processing result of determining network
Efficiency.
Fig. 2 is referred to, Fig. 2 is a kind of processing result prediction meanss based on prediction model provided in an embodiment of the present invention
Structural schematic diagram.The processing result prediction meanss of the embodiment of the present invention include the list for executing above-mentioned processing result prediction technique
Member.Specifically, the processing result prediction meanss 200 of the embodiment of the present invention may include: acquiring unit 201, processing unit 202,
Construction unit 203 and training unit 204.Wherein:
The acquiring unit 201, for obtaining target network at the processing result set of preceding T t raining period, the place
Reason results set includes the t1 processing result data and utilization that the target network is handled for training data
The t2 processing result data that processing result prediction model is predicted for target network progress processing result, the T,
T1 is positive integer, and the t2 is nonnegative integer；
The processing unit 202, for utilizing the processing result prediction model to the place in the processing result set
Reason result data is handled, and prediction obtains the target network in the processing result data of the T+1 t raining period；
Wherein, the processing result prediction model is that the network setup information based on the target network constructs.
In one embodiment, the acquiring unit 201, is also used to obtain the network setup information of the target network,
The network setup information includes the structural parameters and hyper parameter of the target network；
The construction unit 203, for determining target signature parameter according to the network setup information, and according to the mesh
Mark characteristic parameter constructs to obtain the processing result prediction model.
In one embodiment, the training unit 204, for utilizing Support vector regression LIBSVM software to described
Processing result prediction model is trained, the processing result prediction model after being trained；
Wherein, the corresponding support vector machines type of the LIBSVM software is Support Vector Machines for Regression v-SVR, institute
Stating the corresponding kernel function of LIBSVM software is radial basis function RBF.
In one embodiment, the processing result data is used to indicate the classification accuracy of the target network, described
Processing unit 202 is also used to detect the obtained target network of prediction in the processing result data institute of the T+1 t raining period
Whether the classification accuracy of instruction is less than default value；
If target network classification accuracy indicated by the processing result data of the T+1 t raining period is less than institute
Default value is stated, then triggers the training unit 204 and obtains sample data, and using the sample data to the target network
It is trained.
In one embodiment, the processing unit 202, being also used to export includes that the target network is instructed at the T+1
Practice the prompt information of the processing result data in period, whether the prompt information utilizes sample data to the target for prompting
Network is trained；
If detecting the confirmation training instruction for prompt information input, triggers the training unit 204 and obtain
The sample data, and the target network is trained using the sample data.
In one embodiment, the corresponding processing result data of each t raining period of the target network, the T
Equal to the t1 and it is described t2's and.
In one embodiment, the target network is neural network.
Specifically, the processing result prediction meanss 200 can realize the place in above-mentioned embodiment illustrated in fig. 1 by said units
Step some or all of in reason prediction of result method.It should be understood that the embodiment of the present invention is the device reality of corresponding method embodiment
Example is applied, the description to embodiment of the method is also applied for the embodiment of the present invention.
The embodiment of the present invention is by obtaining processing result set of the target network in preceding T t raining period, processing result collection
Close includes that target network is directed to the t1 processing result data that training data is handled and is predicted using processing result
Model carries out the t2 processing result data that processing result is predicted for target network, then predicts mould using processing result
Type handles the processing result data in processing result set, and prediction obtains target network in the T+1 t raining period
Processing result data predicts so as to the processing result to network, to effectively improve the processing result of determining network
Efficiency.
Fig. 3 is referred to, Fig. 3 is a kind of structural schematic diagram of server provided in an embodiment of the present invention.The server is used for
Execute the above-mentioned processing result prediction technique based on prediction model.As shown in figure 3, the server 300 in the present embodiment can wrap
It includes: one or more processors 301 and memory 302.Optionally, which may also include one or more communication interfaces
303.Above-mentioned processor 301, communication interface 303 and memory 302 can be connected by bus 304, or can pass through its other party
Formula connects, and is illustrated in Fig. 3 with bus mode.
Wherein, the processor 301 can be central processing unit (Central Processing Unit, CPU), should
Processor can also be other general processors, digital signal processor (Digital Signal Processor, DSP), specially
With integrated circuit (Application Specific Integrated Circuit, ASIC), ready-made programmable gate array
(Field-Programmable Gate Array, FPGA) either other programmable logic device, discrete gate or transistor are patrolled
Collect device, discrete hardware components etc..General processor can be microprocessor or the processor be also possible to it is any conventional
Processor etc..
The communication interface 303 can be used for receiving and sending messages or the interaction of signaling and the reception of signal and transmitting, communication connect
Mouth 303 may include receiver and transmitter, for being communicated with other equipment.The memory 302 can mainly include storage
Program area and storage data area, wherein storing program area can storage program needed for storage program area, at least one function
(such as text store function, position store function etc.)；Storage data area, which can be stored, uses created number according to server
It according to (such as image data, lteral data) etc., and may include application memory program etc..In addition, memory 302 may include height
Fast random access memory, can also include nonvolatile memory, a for example, at least disk memory, flush memory device,
Or other volatile solid-state parts.
The memory 302 is also used to store program instruction.The processor 301 can call above-mentioned memory 302 to deposit
The program instruction of storage realizes the processing result prediction technique as shown in the embodiment of the invention based on prediction model.
Wherein, processor 301 can be used for calling described program instruction execution following steps: obtain target network at preceding T
The processing result set of t raining period, the processing result set include that the target network handle for training data
T1 processing result data arriving and processing result prediction is carried out for the target network using processing result prediction model
T2 obtained processing result data, described T, t1 are positive integer, and the t2 is nonnegative integer；It is predicted using the processing result
Model handles the processing result data in the processing result set, and prediction obtains the target network at T+1
The processing result data of t raining period；Wherein, the processing result prediction model is the network settings based on the target network
What information architecture obtained.
In one embodiment, processor 301 can also call described program instruction execution following steps: obtain the target
The network setup information of network, the network setup information include the structural parameters and hyper parameter of the target network；According to institute
It states network setup information and determines target signature parameter, and constructed to obtain the processing result prediction according to the target signature parameter
Model.
In one embodiment, processor 301 is calling described in described program instruction execution according to target signature ginseng
It after number building obtains the processing result prediction model, is also used to execute following steps: utilizing Support vector regression
LIBSVM software is trained the processing result prediction model, the processing result prediction model after being trained；Wherein, institute
Stating the corresponding support vector machines type of LIBSVM software is Support Vector Machines for Regression v-SVR, and the LIBSVM software is corresponding
Kernel function be radial basis function RBF.
In one embodiment, the processing result data is used to indicate the classification accuracy of the target network, processing
Device 301 is calling described in described program instruction execution using the processing result prediction model in the processing result set
Processing result data is handled, and prediction obtains the target network after the processing result data of the T+1 t raining period,
It is also used to execute following steps: processing result data of the target network that detection prediction obtains in the T+1 t raining period
Whether indicated classification accuracy is less than default value；If the target network is in the processing result of the T+1 t raining period
Classification accuracy indicated by data is less than the default value, then obtains sample data, and using the sample data to institute
Target network is stated to be trained.
In one embodiment, processor 301 call described program instruction execution described in using the processing result it is pre-
It surveys model to handle the processing result data in the processing result set, prediction obtains the target network in T+1
It after the processing result data of a t raining period, is also used to execute following steps: including institute by the communication interface 303 output
Target network is stated in the prompt information of the processing result data of the T+1 t raining period, whether the prompt information is used for prompt
The target network is trained using sample data；If detecting, the confirmation training for prompt information input refers to
It enables, then obtains the sample data, and be trained to the target network using the sample data.
In one embodiment, the corresponding processing result data of each t raining period of the target network, the T
Equal to the t1 and it is described t2's and.
In one embodiment, the target network is neural network.
In the specific implementation, above-mentioned method reality shown in FIG. 1 can be performed in processor 301 described in the embodiment of the present invention etc.
Implementation described in example is applied, the implementation of each unit described in Fig. 2 of the embodiment of the present invention also can be performed, herein not
It repeats.
The embodiment of the present invention is by obtaining processing result set of the target network in preceding T t raining period, processing result collection
Close includes that target network is directed to the t1 processing result data that training data is handled and is predicted using processing result
Model carries out the t2 processing result data that processing result is predicted for target network, then predicts mould using processing result
Type handles the processing result data in processing result set, and prediction obtains target network in the T+1 t raining period
Processing result data predicts so as to the processing result to network, to effectively improve the processing result of determining network
Efficiency.
The embodiment of the invention also provides a kind of computer readable storage medium, the computer-readable recording medium storage
There is computer program, can be realized when the computer program is executed by processor and handle knot described in embodiment corresponding to Fig. 1
Step some or all of in fruit prediction technique can also realize the processing result prediction meanss of embodiment illustrated in fig. 2 of the present invention
Function can also realize the function of the server of embodiment illustrated in fig. 3 of the present invention, not repeat herein.
The embodiment of the invention also provides a kind of computer program products comprising instruction, when it runs on computers
When, so that step some or all of in the above-mentioned processing result prediction technique of computer execution.
The storage medium can be the inside of processing result prediction meanss or server described in previous embodiment and deposit
Storage unit, such as the hard disk or memory of processing result prediction meanss or server.The storage medium is also possible to the place
What reason prediction of result device was perhaps equipped in the External memory equipment such as processing result prediction or server of server
Plug-in type hard disk, intelligent memory card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card dodge
Deposit card (Flash Card) etc..
In this application, term "and/or", only a kind of incidence relation for describing affiliated partner, indicates may exist
Three kinds of relationships, for example, A and/or B, can indicate: individualism A exists simultaneously A and B, these three situations of individualism B.Separately
Outside, character "/" herein typicallys represent the relationship that forward-backward correlation object is a kind of "or".
In the various embodiments of the application, magnitude of the sequence numbers of the above procedures are not meant to the elder generation of execution sequence
Afterwards, the execution sequence of each process should be determined by its function and internal logic, the implementation process structure without coping with the embodiment of the present invention
At any restriction.
The above, some embodiments only of the invention, but scope of protection of the present invention is not limited thereto, and it is any
Those familiar with the art in the technical scope disclosed by the present invention, can readily occur in various equivalent modifications or replace
It changes, these modifications or substitutions should be covered by the protection scope of the present invention.
Claims (10)
1. a kind of processing result prediction technique based on prediction model, which is characterized in that the described method includes:
Target network is obtained in the processing result set of preceding T t raining period, the processing result set includes the target network
Network is directed to the t1 processing result data that training data is handled and using processing result prediction model for described
Target network carries out the t2 processing result data that processing result is predicted, described T, t1 are positive integer, and the t2 is non-negative
Integer；
The processing result data in the processing result set is handled using the processing result prediction model, is measured in advance
To the target network the T+1 t raining period processing result data；
Wherein, the processing result prediction model is that the network setup information based on the target network constructs.
2. the method according to claim 1, wherein the method also includes:
The network setup information of the target network is obtained, the network setup information includes the structural parameters of the target network
And hyper parameter；
Target signature parameter is determined according to the network setup information, and constructs to obtain the place according to the target signature parameter
Manage prediction of result model.
3. according to the method described in claim 2, it is characterized in that, it is described construct to obtain according to the target signature parameter it is described
After processing result prediction model, the method also includes:
The processing result prediction model is trained using Support vector regression LIBSVM software, the place after being trained
Manage prediction of result model；
Wherein, the corresponding support vector machines type of the LIBSVM software is Support Vector Machines for Regression v-SVR, described
The corresponding kernel function of LIBSVM software is radial basis function RBF.
4. according to the method in any one of claims 1 to 3, which is characterized in that the processing result data is used to indicate
The classification accuracy of the target network, it is described using the processing result prediction model to the place in the processing result set
Reason result data is handled, and prediction obtains the target network after the processing result data of the T+1 t raining period, institute
State method further include:
Target network classification indicated by the processing result data of the T+1 t raining period that detection prediction obtains is accurate
Whether rate is less than default value；
If target network classification accuracy indicated by the processing result data of the T+1 t raining period is less than described pre-
If numerical value, then sample data is obtained, and be trained to the target network using the sample data.
5. according to the method in any one of claims 1 to 3, which is characterized in that described to be predicted using the processing result
Model handles the processing result data in the processing result set, and prediction obtains the target network at T+1
After the processing result data of t raining period, the method also includes:
Output includes the prompt information of processing result data of the target network in the T+1 t raining period, the prompt letter
Whether breath is trained the target network using sample data for prompting；
If detecting the confirmation training instruction for prompt information input, the sample data is obtained, and described in utilization
Sample data is trained the target network.
6. according to the method in any one of claims 1 to 3, which is characterized in that each training of the target network
Period correspond to a processing result data, the T be equal to the t1 and it is described t2's and.
7. according to the method in any one of claims 1 to 3, which is characterized in that the target network is neural network.
8. a kind of processing result prediction meanss based on prediction model, which is characterized in that including for execute as claim 1 to
The unit of method described in any one of 7 claims.
9. a kind of server, which is characterized in that including processor, communication interface and memory, the processor, the communication are connect
Mouth and the memory are connected with each other, wherein for the memory for storing computer program, the computer program includes journey
Sequence instruction, the processor are configured for calling described program instruction, execute as described in any one of claims 1 to 7
Processing result prediction technique based on prediction model.
10. a kind of computer readable storage medium, which is characterized in that the computer-readable recording medium storage has computer journey
Sequence, the computer program include program instruction, and described program instruction executes the processor such as
Processing result prediction technique described in any one of claims 1 to 7 based on prediction model.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910121267.9A CN109919314A (en) | 2019-02-15 | 2019-02-15 | Processing result prediction technique, device and server based on prediction model |
| PCT/CN2019/117992 WO2020164275A1 (en) | 2019-02-15 | 2019-11-13 | Processing result prediction method and apparatus based on prediction model, and server |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910121267.9A CN109919314A (en) | 2019-02-15 | 2019-02-15 | Processing result prediction technique, device and server based on prediction model |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109919314A true CN109919314A (en) | 2019-06-21 |
Family
ID=66961746
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910121267.9A Pending CN109919314A (en) | 2019-02-15 | 2019-02-15 | Processing result prediction technique, device and server based on prediction model |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN109919314A (en) |
| WO (1) | WO2020164275A1 (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020164275A1 (en) * | 2019-02-15 | 2020-08-20 | 平安科技（深圳）有限公司 | Processing result prediction method and apparatus based on prediction model, and server |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9794145B2 (en) * | 2014-07-23 | 2017-10-17 | Cisco Technology, Inc. | Scheduling predictive models for machine learning systems |
| CN107977754A (en) * | 2017-12-18 | 2018-05-01 | 深圳前海微众银行股份有限公司 | Data predication method, system and computer-readable recording medium |
| CN108596399A (en) * | 2018-05-04 | 2018-09-28 | 国家邮政局邮政业安全中心 | Method, apparatus, electronic equipment and the storage medium of express delivery amount prediction |
| CN109919314A (en) * | 2019-02-15 | 2019-06-21 | 平安科技（深圳）有限公司 | Processing result prediction technique, device and server based on prediction model |
-
2019
- 2019-02-15 CN CN201910121267.9A patent/CN109919314A/en active Pending
- 2019-11-13 WO PCT/CN2019/117992 patent/WO2020164275A1/en active Application Filing
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2020164275A1 (en) * | 2019-02-15 | 2020-08-20 | 平安科技（深圳）有限公司 | Processing result prediction method and apparatus based on prediction model, and server |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2020164275A1 (en) | 2020-08-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210294320A1 (en) | Data acquisition method and apparatus | |
| EP3736692A1 (en) | Using computational cost and instantaneous load analysis for intelligent deployment of neural networks on multiple hardware executors | |
| CN108520229A (en) | Image detecting method, device, electronic equipment and computer-readable medium | |
| CN106293914B (en) | A kind of method and terminal of task schedule | |
| CN110046706B (en) | Model generation method and device and server | |
| CN110275958A (en) | Site information recognition methods, device and electronic equipment | |
| CN107800546A (en) | The management method and device of a kind of broadcast message | |
| CN108984388A (en) | A kind of method and terminal device generating automatic test cases | |
| EP4113386A2 (en) | Method, apparatus and system for federated learning, electronic device, computer readable medium | |
| CN107145782A (en) | A kind of recognition methods, mobile terminal and the server of abnormal application program | |
| CN111125519B (en) | User behavior prediction method, device, electronic equipment and storage medium | |
| CN110300191A (en) | Service system and data processing method | |
| JP2020027616A (en) | Command execution method and device | |
| CN110097170A (en) | Information pushes object prediction model acquisition methods, terminal and storage medium | |
| CN109246027A (en) | A kind of method, apparatus and terminal device of network operation | |
| CN110162682A (en) | A kind of crawling method of network data, device, storage medium and terminal device | |
| CN109324962A (en) | A kind of test method and equipment of the application program based on cloud internet of things technology | |
| CN109740041A (en) | Web page crawl method, apparatus, storage medium and computer equipment | |
| CN108965404A (en) | Cloud net health services system and method | |
| CN107231647B (en) | Network detecting method, network detection device and intelligent terminal | |
| CN109167998A (en) | Detect method and device, the electronic equipment, storage medium of camera status | |
| CN109919314A (en) | Processing result prediction technique, device and server based on prediction model | |
| CN109961103A (en) | The training method of Feature Selection Model, the extracting method of characteristics of image and device | |
| CN110297764A (en) | Loophole test model training method and device | |
| CN110362748A (en) | User's recommended method, device, equipment, storage medium and system |
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 |