CN109726764A - A kind of model selection method, device, equipment and medium - Google Patents
A kind of model selection method, device, equipment and medium Download PDFInfo
- Publication number
- CN109726764A CN109726764A CN201811643389.6A CN201811643389A CN109726764A CN 109726764 A CN109726764 A CN 109726764A CN 201811643389 A CN201811643389 A CN 201811643389A CN 109726764 A CN109726764 A CN 109726764A
- Authority
- CN
- China
- Prior art keywords
- model
- preset model
- training
- preset
- data
- 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
Abstract
The invention discloses a kind of model selection method, device, equipment and medium, the model selection method, comprising: obtain the corresponding training data of at least one preset model for belonging to same preset model classification;For each preset model at least one described preset model, the preset model is trained according to the training data, determines the preset model model metrics parameter corresponding with the preset model for completing training;According to the corresponding model metrics parameter of each preset model, from the preset model for completing training, determine that the same preset model is classified corresponding object module.The model metrics parameter that each preset model while being trained to multiple preset models, can also be calculated by the embodiment of the present application, improves the experience effect of user.
Description
Technical field
This application involves data processing field more particularly to a kind of model selection method, device, equipment and media.
Background technique
With the rapid development of industry, all kinds of manufacturings, process flow, management philosophy, managerial experiences in industrial circle
It is more and more etc. various invisible digital assets, for the ease of the management to these digital assets, these digital assets are changed
At industrial algorithm model, therefore, the quantity of industrial algorithm model is also more and more.
In the prior art, user needs to select in the industrial algorithm model of substantial amounts in applicable industry algorithm model
Meet user demand industry algorithm model out, the industrial algorithm model for meeting user demand that final choice goes out may have it is multiple,
Each industrial algorithm model that user obtains all substantially conforms to the demand of user.
Summary of the invention
In view of this, the application's has been designed to provide a kind of model selection method, device, equipment and medium, solve
The problem of selecting the low efficiency of object module in multiple models in the prior art.
In a first aspect, the embodiment of the present application provides a kind of model selection method, comprising:
Obtain the corresponding training data of at least one preset model for belonging to same preset model classification;
For each preset model at least one described preset model, according to the training data to the preset model
It is trained, determines the preset model model metrics parameter corresponding with the preset model for completing training;
It is determined from the preset model for completing training according to the corresponding model metrics parameter of each preset model
The same preset model is classified corresponding object module.
Optionally, the acquisition belongs to the corresponding training data of at least one preset model of same preset model classification,
Include:
Obtain the metadata for belonging at least one preset model of same preset model classification;
Based on the metadata, the training data of at least one preset model of the same preset model classification is obtained.
Optionally, the corresponding model metrics parameter of the determination preset model, comprising:
It obtains the corresponding mark test data of at least one preset model for belonging to same preset model classification and does not mark
Test data;
For each preset model for completing training, the test data that do not mark is input to the default of completion training
Model determines the test result of the preset model of completion training;
Test result and mark test data based on each preset model for completing training determine each default mould of completion training
The model accuracy rate of type, using the model accuracy rate as the model metrics parameter.
Optionally, described according to the corresponding model metrics parameter of each preset model, training is completed from described
In preset model, the corresponding object module of the same preset model classification is determined, comprising:
The corresponding preset model for completing training of the model accuracy rate for characterizing maximum value is determined as the same default mould
The corresponding object module of type classification;Alternatively,
The model training time for the characterizing minimum value corresponding preset model for completing training is determined as described same default
The corresponding object module of category of model.
Optionally, the acquisition belongs to the corresponding training data of at least one preset model of same preset model classification,
Include:
If it is determined that the data volume of the training data is greater than setting data-quantity threshold, then obtained according to setting data acquiring mode
Take the training data.
Second aspect, the embodiment of the present application provide a kind of model selection device, comprising:
Data module is obtained, for obtaining the corresponding training of at least one preset model for belonging to same preset model classification
Data;
Computing module, for each preset model at least one preset model for described in, according to the trained number
It is trained according to the preset model, determines the preset model model metrics parameter corresponding with the preset model for completing training;
Determining module, for completing the pre- of training from described according to the corresponding model metrics parameter of each preset model
If in model, determining the corresponding object module of the same preset model classification.
Optionally, the acquisition module is specifically used for:
Obtain the metadata for belonging at least one preset model of same preset model classification;
Based on the metadata, the training data of at least one preset model of the same preset model classification is obtained.
Optionally, the computing module is specifically used for:
It obtains the corresponding mark test data of at least one preset model for belonging to same preset model classification and does not mark
Test data;
For each preset model for completing training, the test data that do not mark is input to the default of completion training
Model determines the test result of the preset model of completion training;
Test result and mark test data based on each preset model for completing training determine each default mould of completion training
The model accuracy rate of type, using the model accuracy rate as the model metrics parameter.
The third aspect, the embodiment of the present application provide a kind of computer equipment, including memory, processor and are stored in institute
The computer program that can be run on memory and on the processor is stated, the processor executes real when the computer program
The step of showing above-mentioned method.
Fourth aspect, the embodiment of the present application provide a kind of computer readable storage medium, the computer-readable storage
Computer program is stored on medium, the computer program executes above-mentioned method when being run by processor the step of.
Model selection method provided in an embodiment of the present invention is classified extremely by training data to same preset model is belonged to
A few preset model is trained, and determines the model metrics parameter of each preset model, according to model metrics parameter subordinate
Optimal models are selected at least one preset model of same preset model classification, optimal models are exactly the target that user needs
Model can also be calculated each default by the embodiment of the present application while being trained to multiple preset models
The model metrics parameter of model can select the optimal models for meeting user demand, so that user according to each model metrics parameter
The object module of obtained negligible amounts can allow the selection of user more targeted compared with prior art, can be same
The preset model for being best suitable for user's requirement is selected in multiple preset models of type, also improves the experience effect of user.
To enable the above objects, features, and advantages of the application to be clearer and more comprehensible, preferred embodiment is cited below particularly, and cooperate
Appended attached drawing, is described in detail below.
Detailed description of the invention
Technical solution in ord to more clearly illustrate embodiments of the present application, below will be to needed in the embodiment attached
Figure is briefly described, it should be understood that the following drawings illustrates only some embodiments of the application, therefore is not construed as pair
The restriction of range for those of ordinary skill in the art without creative efforts, can also be according to this
A little attached drawings obtain other relevant attached drawings.
Fig. 1 is a kind of flow diagram of model selection method provided by the embodiments of the present application;
Fig. 2 is a kind of flow diagram of computation model metric parameter method provided by the embodiments of the present application;
Fig. 3 is a kind of structural schematic diagram of model selection device provided by the embodiments of the present application;
Fig. 4 is a kind of structural schematic diagram of computer equipment provided by the embodiments of the present application.
Specific embodiment
To keep the purposes, technical schemes and advantages of the embodiment of the present application clearer, below in conjunction with the embodiment of the present application
Middle attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is only
It is some embodiments of the present application, instead of all the embodiments.The application being usually described and illustrated herein in the accompanying drawings is real
The component for applying example can be arranged and be designed with a variety of different configurations.Therefore, below to the application's provided in the accompanying drawings
The detailed description of embodiment is not intended to limit claimed scope of the present application, but is merely representative of the selected reality of the application
Apply example.Based on embodiments herein, those skilled in the art institute obtained without making creative work
There are other embodiments, shall fall in the protection scope of this application.
In view of army's simulated training system is that single-chip microcontroller is arranged by embedded mode to set in operation in the prior art
In standby, and then the operation data of warfare equipment is acquired, single-chip microcontroller is arranged by embedded mode and needs to change warfare equipment
Dress, the warfare equipment after repacking is in actual combat in application, can adversely affect to actual combat result.Based on this, the present invention is implemented
Example provides a kind of fixed structure, sensor and simulation system, is described below by embodiment.
User is as shown in Figure 1, the embodiment of the present application provides a kind of model selection method, comprising:
101, obtain the corresponding training data of at least one preset model for belonging to same preset model classification;
Here, preset model is the model pre-set, for example, the model pre-set can be convolutional Neural net
Network model, bayesian algorithm model, regularization algorithm model etc., preset model, which is classified, is generally based on the applied field of preset model
Scape determines that, for example, cutter life prediction, the exceeded analysis of equipment pollution etc., training data is the data for training preset model.
For example, default classification is cutter life prediction, corresponding cutter life prediction instruction is obtained by cutter life prediction
Practice data;Default classification is the exceeded analysis of equipment pollution, and it is exceeded to obtain corresponding equipment pollution by the exceeded analysis of equipment pollution
Analyzing and training data.
Specifically, all default moulds of same preset model classification are selected in all preset models according to user demand
Type is associated with same training data for each preset model of selection.
For example, the demand of user is prediction model, the prediction model for belonging to prediction classification has 3, including prediction model A,
Prediction model B, prediction model C, according to prediction classification obtain training data, by the training data respectively with prediction model A, prediction
Model B, prediction model C association.
In a step 101, the metadata for belonging at least one preset model of same preset model classification is obtained;
Based on the metadata, the training data of at least one preset model of the same preset model classification is obtained.
Here, metadata is the information for describing data attribute, the including but not limited to storage address of data, database
User account, user password etc., training data, test data etc. are stored in database, test data is pre- for detecting
If the whether accurate data of model, test data includes not marking test data and mark test data.
After getting and belonging to the metadata of same preset model classification with each preset model, by metadata and each default mould
Type is associated, and finds corresponding database according to the database address in metadata, using in metadata user account and
User password obtains user requested data from database, and user requested data includes training data.
102, it is default to this according to the training data for each preset model at least one described preset model
Model is trained, and determines the preset model model metrics parameter corresponding with the preset model for completing training;
Here, model metrics parameter is the data for measuring the preset model for completing training, and model metrics parameter includes
Accuracy rate and calculating time, accuracy rate are that all test datas that do not mark in test data are input to the default of completion training
The percentage of the correct result quantity that is calculated in model and all fruiting quantities, calculate the time be by test data not
Mark test data is input to the time needed for calculating after the completion in the preset model for completing training.
As shown in Fig. 2, the embodiment of the present application provides a kind of computation model metric parameter method, it is in a step 102, described
Determine the corresponding model metrics parameter of the preset model, comprising:
201, obtain the corresponding mark test data of at least one preset model for belonging to the classification of same preset model and not
Mark test data;
202, for each preset model for completing training, the test data that do not mark is input to completion training
Preset model determines the test result of the preset model of completion training;
203, test result and mark test data based on each preset model for completing training determine that each completion training is pre-
If the model accuracy rate of model, using the model accuracy rate as the model metrics parameter.
Specifically, the training data that will acquire is input in each preset model to the classification of same preset model, it is right
The preset model of input training data is trained, and after the completion of training, obtains test data, will not mark test data input
Into the preset model for completing training, is calculated by preset model to not marking test data, obtain calculated result, will
To calculated result with mark test data be compared, calculated result and mark test data be greater than given threshold when, label
The calculated result of the preset model be it is correct, when calculated result and mark test data are less than given threshold, mark the default mould
The calculated result of type is mistake, calculates the percentage of correct calculated result quantity Yu all calculated result quantity, the percentage
Completion meter can also be recorded when calculating has input the preset model for not marking test data for the accuracy rate of the preset model
Calculate the calculating time used, wherein given threshold can be calculated result and mark the similarity of test data, for example, 90%.
Continue above-mentioned steps 101 in obtain training data example, the training data that will acquire respectively to prediction model A,
Prediction model B, prediction model C are trained, and obtain the prediction model A for completing training1, prediction model B1, prediction model C1, obtain
It has taken 100 groups of mark test datas and has not marked test data, the test data that do not mark in this 100 groups of data has been inputted respectively
To prediction model A1, prediction model B1, prediction model C1In, for prediction model A1, pass through prediction model A1To not marking test
Data are calculated, and 100 groups of calculated results are obtained, this 100 groups of calculated results are compared with 100 groups of labeled data, are obtained
It is correctly then prediction model A to 80 groups of calculated results1Accuracy rate be 80%, similarly, by with calculate prediction model A1It is quasi-
The identical method of example of true rate calculates prediction model B1, prediction model C1Accuracy rate, prediction model B1Accuracy rate be
70%, prediction model C1Accuracy rate be 85%.
103, according to the corresponding model metrics parameter of each preset model, from the preset model for completing training,
Determine the corresponding object module of the same preset model classification.
Here, object module is the model for meeting user demand.
In step 103, the corresponding preset model for completing training of the model accuracy rate for characterizing maximum value is determined as institute
State the corresponding object module of same preset model classification;Alternatively,
The model training time for the characterizing minimum value corresponding preset model for completing training is determined as described same default
The corresponding object module of category of model.
When the model needed for user is accuracy rate higher object module, according to each accuracy rate being calculated, belonging to
Preset model corresponding with maximum accuracy rate is selected in all preset models of same preset model classification, then the default mould
Type is exactly object module needed for user.
The example for continuing the calculating preset model accuracy rate in step 102, according to prediction model A1Accuracy rate be 80%,
Prediction model B1Accuracy rate be 70%, prediction model C1Accuracy rate be 85%, accuracy rate in three preset models is maximum
Be prediction model C1, then prediction model C1It is object module needed for user.
When the model needed for user is calculating time short object module, the pre- of test data is not marked according to having input
If the calculating time of model, selected and the most short calculating time pair in all preset models for belonging to same preset model classification
The preset model answered, then the preset model is exactly object module needed for user.
The example for continuing the calculating preset model accuracy rate in step 102, is calculating prediction model A1, prediction model B1, it is pre-
Survey MODEL C1While accuracy rate, prediction model A is recorded1, prediction model B1, prediction model C1The calculating time, prediction model A1
The calculating time be 10s, prediction model B1The calculating time be 15s, prediction model C1The calculating time be 20s, three predict
Model fall into a trap evaluation time it is shortest be prediction model A1, then prediction model A1It is object module needed for user.
There is the accuracy rate of at least two preset models identical when belonging in all preset models that same preset model is classified
When, according to the calculating time of at least two preset models, selected at least two preset models and the most short calculating time pair
The preset model answered, then the preset model is exactly object module needed for user.
For example, three prediction models for completing training include prediction model A2, prediction model B2, prediction model C2, predict mould
Type A2Accuracy rate be 85%, prediction model B2Accuracy rate be 70%, prediction model C2Accuracy rate be 85%, prediction model
A2The calculating time be 10s, prediction model B2The calculating time be 15s, prediction model C2The calculating time be 20s, it is pre- at three
Surveying the maximum prediction model of accuracy rate in model is prediction model A2With prediction model C2, according to prediction model A2The calculating time
With prediction model C2The calculating time, two prediction models fall into a trap evaluation time it is shortest be prediction model A2, then prediction model A2
The object module of user.
When the calculating time phase for having at least two preset models in all preset models for belonging to same preset model classification
Meanwhile according to the accuracy rate of at least two preset models, selected at least two preset models corresponding with maximum accuracy rate
Preset model, then the preset model is exactly object module needed for user.
For example, three prediction models for completing training include prediction model A3, prediction model B3, prediction model C3, predict mould
Type A3Accuracy rate be 75%, prediction model B3Accuracy rate be 70%, prediction model C3Accuracy rate be 85%, prediction model
A3The calculating time be 10s, prediction model B3The calculating time be 15s, prediction model C3The calculating time be 10s, it is pre- at three
It surveys model and falls into a trap the shortest prediction model of evaluation time for prediction model A3With prediction model C3, according to prediction model A3Accuracy rate
With prediction model C3Accuracy rate, it is prediction model C that accuracy rate is maximum in two prediction models3, then prediction model C3User
Object module.
In a step 101, the acquisition belongs to the corresponding training of at least one preset model of same preset model classification
Data, comprising:
If it is determined that the data volume of the training data is greater than setting data-quantity threshold, then obtained according to setting data acquiring mode
Take the training data.
Here, setting data-quantity threshold is the value for flag data amount, for example, 1TB, 1GB etc., set data recipient
Formula is according to the mode of the rules Acquires data set, for example, message queue etc., the application not limits this.
When the data volume of the training data got by the database address in metadata is too big, server cannot be propped up
The transmission of the training data is held, prevent preset model is from being normally trained, in order to make training data timely
Transmission and preset model can be trained according to the training data, need by the training data according to certain rules into
Row transmission, is trained preset model timely according to the training data of transmission, and then improves the instruction of preset model
Practice efficiency.
For example, training data is got by the database address in metadata, if detecting the data of training data
Amount is greater than 1TB, then the data volume of training data is excessive, cannot support to be greater than 1TB data volume with the data transmission capabilities of server
Data transmission, cause preset model that cannot be trained according to the training data, need the training data according to message team
The mode of column is transmitted, for example, training data is split according to the data volume of 64M, the data volume of training data may
It is not the integral multiple of 64M, then includes the data volume resentful to 64M, more than or less than 64M's after training data segmentation
Data volume carries out data transmission.
It is different from the parameter format in preset model in the training data got by the database address in metadata
When, it is also necessary to the format for the training data that will acquire is converted into the format of parameter in preset model, for example, the training data obtained
It is document format, by various conversion, which is converted into string format.
Algorithm provided by the embodiments of the present application executes method, is classified extremely by training data to same preset model is belonged to
A few preset model is trained, and determines the model metrics parameter of each preset model, according to model metrics parameter subordinate
Optimal models are selected at least one preset model of same preset model classification, optimal models are exactly the target that user needs
Model can also be calculated each default by the embodiment of the present application while being trained to multiple preset models
The model metrics parameter of model can select the optimal models for meeting user demand according to each model metrics parameter, also improve
The experience effect of user.
Present invention also provides a kind of specific algorithms to execute method usage scenario, according to user demand it is found that user's need
Want two kinds of object modules, comprising: disaggregated model and prediction model, wherein user gets 5 disaggregated models and 8 prediction moulds
Type executes the process for the object module for being best suitable for user demand that method obtains by above-mentioned algorithm in 5 disaggregated models,
The longer calculating time is needed, in order to save the time of user, can calculate and classified by opening multithreading realization
When object module in model, while the object module of prediction model is calculated and being obtained, calculate and obtains the target of disaggregated model
The task of model and the task for the object module for calculating and obtaining prediction model can carry out simultaneously, save the time of user,
It improves user and obtains the efficiency of object module, to improve the experience effect of user.
During calculating for each preset model training data, if the data volume of training data is larger
When, the training of preset model can be realized by distributed system, for example, prediction model D needs the number of the training data of training
It is larger according to measuring, the training mission of prediction model D is distributed in multiple computers (e.g., 3) in computer S, firstly, by pre-
It surveys model D to be distributed in 3 computers, there is computer S1, computer S2, computer S3 then to divide training data respectively
At three divided datas, respectively with data J1, data J2, data J3 characterization, the data J1 training data characterized is distributed into computer
S1 is trained the data J1 training data characterized by prediction model D, the data J2 training data characterized is distributed to meter
Calculation machine S2 is trained the data J2 training data characterized by prediction model D, and the training data that data J3 is characterized distributes
Computer S3 is given, the data J3 training data characterized is trained by prediction model D, computer S1, computer S2, is calculated
Training result is transmitted to computer S by machine S3, and computer S is according to the training result meter of computer S1, computer S2, computer S3
Prediction model needed for calculation obtains user.The biggish training data of data volume is trained by distributed system, is not needed
User is always trained all training datas with the same prediction model, training data is divided into more parts, for each
The training data of part is all trained with the same prediction model, is shortened prediction model to the training time of training data, is mentioned
High user obtains the efficiency of object module, and then improves the experience effect of user.
As shown in figure 3, the embodiment of the present application provides a kind of model selection device, comprising:
Data module 301 is obtained, to belong at least one preset model that same preset model is classified corresponding for obtaining
Training data;
Computing module 302, for each preset model at least one preset model for described in, according to the training
Data are trained the preset model, determine the preset model for completing training model metrics ginseng corresponding with the preset model
Number;
Determining module 303 completes training from described for according to the corresponding model metrics parameter of each preset model
In preset model, the corresponding object module of the same preset model classification is determined.
Optionally, the acquisition module 301 is specifically used for:
Obtain the metadata for belonging at least one preset model of same preset model classification;
Based on the metadata, the training data of at least one preset model of the same preset model classification is obtained.
Optionally, the computing module 302 is specifically used for:
It obtains the corresponding mark test data of at least one preset model for belonging to same preset model classification and does not mark
Test data;
For each preset model for completing training, the test data that do not mark is input to the default of completion training
Model determines the test result of the preset model of completion training;
Test result and mark test data based on each preset model for completing training determine each default mould of completion training
The model accuracy rate of type, using the model accuracy rate as the model metrics parameter.
Optionally, the determining module 303 is specifically used for:
The corresponding preset model for completing training of the model accuracy rate for characterizing maximum value is determined as the same default mould
The corresponding object module of type classification;Alternatively,
The preset model that the model for characterizing minimum value calculates time corresponding completion training is determined as described same default
The corresponding object module of category of model.
Optionally, the model selection device further include: transmission module 304, the transmission module 304 are specifically used for:
If it is determined that the data volume of the training data is greater than setting data-quantity threshold, then obtained according to setting data acquiring mode
Take the training data.
Corresponding to the model selection method in Fig. 1, the embodiment of the present application also provides a kind of computer equipments 400, such as Fig. 4
Shown, which includes memory 401, processor 402 and is stored on the memory 401 and can transport on the processor 402
Capable computer program, wherein above-mentioned processor 402 realizes the step of above-mentioned model selection method when executing above-mentioned computer program
Suddenly.
Specifically, above-mentioned memory 401 and processor 402 can be general memory and processor, do not do have here
Body limits, and when the computer program of 402 run memory 401 of processor storage, is able to carry out above-mentioned model selection method, uses
It, can be by the embodiment of the present application in the low efficiency for solving the problems, such as to select object module in multiple models in the prior art
While being trained to multiple preset models, the model metrics parameter of each preset model can also be calculated, it can basis
Each model metrics parameter selects the optimal models for meeting user demand, so that the object module for the negligible amounts that user obtains, with
The prior art can be selected in multiple preset models of same type and most be accorded with compared to the selection of user can be allowed more targeted
The preset model for sharing family requirement, improves the experience effect of user.
Corresponding to the model selection method in Fig. 1, the embodiment of the present application also provides a kind of computer readable storage medium,
It is stored with computer program on the computer readable storage medium, which executes above-mentioned model when being run by processor
The step of selection method.
Specifically, which can be general storage medium, such as mobile disk, hard disk, on the storage medium
Computer program when being run, above-mentioned model selection method is able to carry out, for solving in the prior art in multiple models
The problem of selecting the low efficiency of object module, by the embodiment of the present application can multiple preset models are trained it is same
When, the model metrics parameter of each preset model can also be calculated, can be selected according to each model metrics parameter and meet user
The optimal models of demand can allow the choosing of user so that the object module for the negligible amounts that user obtains compared with prior art
The preset model for being best suitable for user's requirement can more targetedly be selected in multiple preset models of same type by selecting, and be improved
The experience effect of user.
In embodiment provided herein, it should be understood that disclosed device and method, it can be by others side
Formula is realized.The apparatus embodiments described above are merely exemplary, for example, the division of the unit, only one kind are patrolled
Function division is collected, there may be another division manner in actual implementation, in another example, multiple units or components can combine or can
To be integrated into another system, or some features can be ignored or not executed.Another point, shown or discussed is mutual
Coupling, direct-coupling or communication connection can be INDIRECT COUPLING or communication link by some communication interfaces, device or unit
It connects, can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit
The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple
In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme
's.
In addition, each functional unit in embodiment provided by the present application can integrate in one processing unit, it can also
To be that each unit physically exists alone, can also be integrated in one unit with two or more units.
It, can be with if the function is realized in the form of SFU software functional unit and when sold or used as an independent product
It is stored in a computer readable storage medium.Based on this understanding, the technical solution of the application is substantially in other words
The part of the part that contributes to existing technology or the technical solution can be embodied in the form of software products, the meter
Calculation machine software product is stored in a storage medium, including some instructions are used so that a computer equipment (can be a
People's computer, server or network equipment etc.) execute each embodiment the method for the application all or part of the steps.
And storage medium above-mentioned includes: that USB flash disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), arbitrary access are deposited
The various media that can store program code such as reservoir (RAM, Random Access Memory), magnetic or disk.
It should also be noted that similar label and letter indicate similar terms in following attached drawing, therefore, once a certain Xiang Yi
It is defined in a attached drawing, does not then need that it is further defined and explained in subsequent attached drawing, in addition, term " the
One ", " second ", " third " etc. are only used for distinguishing description, are not understood to indicate or imply relative importance.
Finally, it should be noted that embodiment described above, the only specific embodiment of the application, to illustrate the application
Technical solution, rather than its limitations, the protection scope of the application is not limited thereto, although with reference to the foregoing embodiments to this Shen
It please be described in detail, those skilled in the art should understand that: anyone skilled in the art
Within the technical scope of the present application, it can still modify to technical solution documented by previous embodiment or can be light
It is readily conceivable that variation or equivalent replacement of some of the technical features;And these modifications, variation or replacement, do not make
The essence of corresponding technical solution is detached from the spirit and scope of the embodiment of the present application technical solution.The protection in the application should all be covered
Within the scope of.Therefore, the protection scope of the application shall be subject to the protection scope of the claim.
Claims (10)
1. a kind of model selection method characterized by comprising
Obtain the corresponding training data of at least one preset model for belonging to same preset model classification;
For each preset model at least one described preset model, the preset model is carried out according to the training data
Training determines the preset model model metrics parameter corresponding with the preset model for completing training;
According to the corresponding model metrics parameter of each preset model, from the preset model for completing training, determine described in
Same preset model is classified corresponding object module.
2. the method as described in claim 1, which is characterized in that the acquisition belongs at least one of same preset model classification
The corresponding training data of preset model, comprising:
Obtain the metadata for belonging at least one preset model of same preset model classification;
Based on the metadata, the training data of at least one preset model of the same preset model classification is obtained.
3. the method according to claim 1, wherein the corresponding model metrics ginseng of the determination preset model
Number, comprising:
It obtains the corresponding mark test data of at least one preset model for belonging to same preset model classification and does not mark test
Data;
For each preset model for completing training, by the default mould for not marking test data and being input to completion training
Type determines the test result of the preset model of completion training;
Test result and mark test data based on each preset model for completing training determine each completion training preset model
Model accuracy rate, using the model accuracy rate as the model metrics parameter.
4. method according to claim 1 or 3, which is characterized in that described corresponding according to each preset model
Model metrics parameter determines the corresponding target mould of the same preset model classification from the preset model for completing training
Type, comprising:
The corresponding preset model for completing training of the model accuracy rate for characterizing maximum value is determined as the same preset model point
The corresponding object module of class;Alternatively,
The model for characterizing minimum value is calculated into the time corresponding preset model for completing training and is determined as the same preset model
Classify corresponding object module.
5. the method according to claim 1, wherein the acquisition belongs at least the one of same preset model classification
The corresponding training data of a preset model, comprising:
If it is determined that the data volume of the training data is greater than setting data-quantity threshold, then institute is obtained according to setting data acquiring mode
State training data.
6. a kind of model selection device characterized by comprising
Data module is obtained, for obtaining the corresponding trained number of at least one preset model for belonging to same preset model classification
According to;
Computing module, for each preset model at least one preset model for described in, according to the training data pair
The preset model is trained, and determines the preset model model metrics parameter corresponding with the preset model for completing training;
Determining module, for according to the corresponding model metrics parameter of each preset model, from the default mould for completing training
In type, the corresponding object module of the same preset model classification is determined.
7. device as claimed in claim 6, which is characterized in that the acquisition module is specifically used for:
Obtain the metadata for belonging at least one preset model of same preset model classification;
Based on the metadata, the training data of at least one preset model of the same preset model classification is obtained.
8. device according to claim 6, which is characterized in that the computing module is specifically used for:
It obtains the corresponding mark test data of at least one preset model for belonging to same preset model classification and does not mark test
Data;
For each preset model for completing training, by the default mould for not marking test data and being input to completion training
Type determines the test result of the preset model of completion training;
Test result and mark test data based on each preset model for completing training determine each completion training preset model
Model accuracy rate, using the model accuracy rate as the model metrics parameter.
9. a kind of computer equipment, including memory, processor and it is stored on the memory and can be on the processor
The computer program of operation, which is characterized in that the processor realizes the claims 1-5 when executing the computer program
Any one of described in method the step of.
10. a kind of computer readable storage medium, computer program, feature are stored on the computer readable storage medium
The step of being, method described in any one of the claims 1-5 executed when the computer program is run by processor.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811643389.6A CN109726764A (en) | 2018-12-29 | 2018-12-29 | A kind of model selection method, device, equipment and medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811643389.6A CN109726764A (en) | 2018-12-29 | 2018-12-29 | A kind of model selection method, device, equipment and medium |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109726764A true CN109726764A (en) | 2019-05-07 |
Family
ID=66299442
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811643389.6A Pending CN109726764A (en) | 2018-12-29 | 2018-12-29 | A kind of model selection method, device, equipment and medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109726764A (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110333994A (en) * | 2019-07-08 | 2019-10-15 | 深圳乐信软件技术有限公司 | A kind of matched model of data set determines method, apparatus, equipment and storage medium |
CN110443310A (en) * | 2019-08-07 | 2019-11-12 | 浙江大华技术股份有限公司 | Compare update method, server and the computer storage medium of analysis system |
CN110457369A (en) * | 2019-08-07 | 2019-11-15 | 北京嘉和海森健康科技有限公司 | A kind of training method and relevant device of model |
CN112149708A (en) * | 2019-06-28 | 2020-12-29 | 富泰华工业(深圳)有限公司 | Data model selection optimization method and device, computer device and storage medium |
CN113284141A (en) * | 2021-07-09 | 2021-08-20 | 武汉精创电子技术有限公司 | Model determination method, device and equipment for defect detection |
CN115689626A (en) * | 2022-10-31 | 2023-02-03 | 荣耀终端有限公司 | User attribute determination method of terminal equipment and electronic equipment |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103268517A (en) * | 2013-04-23 | 2013-08-28 | 重庆科技学院 | Multivariate quality process out-of-control signal diagnostic method based on support vector machine and genetic optimization |
CN103412882A (en) * | 2013-07-18 | 2013-11-27 | 百度在线网络技术(北京)有限公司 | Method and device for distinguishing consumption intention |
CN107563999A (en) * | 2017-09-05 | 2018-01-09 | 华中科技大学 | A kind of chip defect recognition methods based on convolutional neural networks |
CN107566907A (en) * | 2017-09-20 | 2018-01-09 | 广东欧珀移动通信有限公司 | video clipping method, device, storage medium and terminal |
CN108446388A (en) * | 2018-03-22 | 2018-08-24 | 平安科技(深圳)有限公司 | Text data quality detecting method, device, equipment and computer readable storage medium |
CN109033921A (en) * | 2017-06-08 | 2018-12-18 | 北京君正集成电路股份有限公司 | A kind of training method and device of identification model |
CN109086825A (en) * | 2018-08-03 | 2018-12-25 | 北京邮电大学 | A kind of more disaggregated model fusion methods based on model adaptation selection |
-
2018
- 2018-12-29 CN CN201811643389.6A patent/CN109726764A/en active Pending
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103268517A (en) * | 2013-04-23 | 2013-08-28 | 重庆科技学院 | Multivariate quality process out-of-control signal diagnostic method based on support vector machine and genetic optimization |
CN103412882A (en) * | 2013-07-18 | 2013-11-27 | 百度在线网络技术(北京)有限公司 | Method and device for distinguishing consumption intention |
CN109033921A (en) * | 2017-06-08 | 2018-12-18 | 北京君正集成电路股份有限公司 | A kind of training method and device of identification model |
CN107563999A (en) * | 2017-09-05 | 2018-01-09 | 华中科技大学 | A kind of chip defect recognition methods based on convolutional neural networks |
CN107566907A (en) * | 2017-09-20 | 2018-01-09 | 广东欧珀移动通信有限公司 | video clipping method, device, storage medium and terminal |
CN108446388A (en) * | 2018-03-22 | 2018-08-24 | 平安科技(深圳)有限公司 | Text data quality detecting method, device, equipment and computer readable storage medium |
CN109086825A (en) * | 2018-08-03 | 2018-12-25 | 北京邮电大学 | A kind of more disaggregated model fusion methods based on model adaptation selection |
Non-Patent Citations (1)
Title |
---|
王喜宾 等: "《基于优化支持向量机的个性化推荐研究》", 30 June 2017, 重庆大学出版社 * |
Cited By (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112149708A (en) * | 2019-06-28 | 2020-12-29 | 富泰华工业(深圳)有限公司 | Data model selection optimization method and device, computer device and storage medium |
CN110333994A (en) * | 2019-07-08 | 2019-10-15 | 深圳乐信软件技术有限公司 | A kind of matched model of data set determines method, apparatus, equipment and storage medium |
CN110333994B (en) * | 2019-07-08 | 2023-06-06 | 深圳乐信软件技术有限公司 | Data set matching model determination method, device, equipment and storage medium |
CN110443310A (en) * | 2019-08-07 | 2019-11-12 | 浙江大华技术股份有限公司 | Compare update method, server and the computer storage medium of analysis system |
CN110457369A (en) * | 2019-08-07 | 2019-11-15 | 北京嘉和海森健康科技有限公司 | A kind of training method and relevant device of model |
CN110443310B (en) * | 2019-08-07 | 2022-08-09 | 浙江大华技术股份有限公司 | Updating method of comparison analysis system, server and computer storage medium |
CN113284141A (en) * | 2021-07-09 | 2021-08-20 | 武汉精创电子技术有限公司 | Model determination method, device and equipment for defect detection |
CN115689626A (en) * | 2022-10-31 | 2023-02-03 | 荣耀终端有限公司 | User attribute determination method of terminal equipment and electronic equipment |
CN115689626B (en) * | 2022-10-31 | 2024-03-01 | 荣耀终端有限公司 | User attribute determining method of terminal equipment and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109726764A (en) | A kind of model selection method, device, equipment and medium | |
TWI746814B (en) | Computer readable medium, car insurance risk prediction device and server | |
CN109902708B (en) | Recommendation model training method and related device | |
CN109670267B (en) | Data processing method and device | |
CN107613022B (en) | Content pushing method and device and computer equipment | |
CN108960719B (en) | Method and device for selecting products and computer readable storage medium | |
US20150039540A1 (en) | Method and apparatus for evaluating predictive model | |
CN108492194A (en) | Products Show method, apparatus and storage medium | |
CN108833458B (en) | Application recommendation method, device, medium and equipment | |
CN108062573A (en) | Model training method and device | |
Steiniger et al. | OpenJUMP HoRAE—A free GIS and toolbox for home‐range analysis | |
CN109388675A (en) | Data analysing method, device, computer equipment and storage medium | |
CN103646070A (en) | Data processing method and device for search engine | |
CN110909868A (en) | Node representation method and device based on graph neural network model | |
CN111460384A (en) | Policy evaluation method, device and equipment | |
CN109670653A (en) | A kind of method and device predicted based on industrial model predictive engine | |
CN109690571A (en) | Group echo system and method based on study | |
US20190220924A1 (en) | Method and device for determining key variable in model | |
CN110162609A (en) | For recommending the method and device asked questions to user | |
CN107633257B (en) | Data quality evaluation method and device, computer readable storage medium and terminal | |
CN112465141A (en) | Model compression method, model compression device, electronic device and medium | |
CN106156857B (en) | The method and apparatus of the data initialization of variation reasoning | |
CN115545103A (en) | Abnormal data identification method, label identification method and abnormal data identification device | |
CN109800138B (en) | CPU testing method, electronic device and storage medium | |
CN106776757A (en) | User completes the indicating means and device of Net silver operation |
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: 20190507 |
|
RJ01 | Rejection of invention patent application after publication |