CN109739750A - The determination method and apparatus of the quality prediction model of R&D team - Google Patents
The determination method and apparatus of the quality prediction model of R&D team Download PDFInfo
- Publication number
- CN109739750A CN109739750A CN201811542200.4A CN201811542200A CN109739750A CN 109739750 A CN109739750 A CN 109739750A CN 201811542200 A CN201811542200 A CN 201811542200A CN 109739750 A CN109739750 A CN 109739750A
- Authority
- CN
- China
- Prior art keywords
- characteristic information
- quality
- model
- quality prediction
- prediction model
- 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
Landscapes
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
This disclosure relates to the determination method of the quality prediction model of R&D team, comprising: the application fetches foundation characteristic information based on R&D team's research and development, the foundation characteristic information are used to characterize the quality of the application；Each foundation characteristic information is extended respectively by preset algorithm, with the characteristic information that is expanded；Sample set is constituted according to the application and corresponding foundation characteristic information and extension feature information；Based on the sample set, preset model is trained by machine learning algorithm, to obtain quality prediction model.In accordance with an embodiment of the present disclosure, it on the one hand for the parameter of each characteristic information is obtained by machine learning, opposite artificial settings is more accurate, on the other hand since characteristic information includes foundation characteristic information and extension feature information, the quality of R&D team can comprehensively be characterized, so as to obtain more accurate quality prediction model, so that the subsequent quality to R&D team is predicted.
Description
Technical field
This disclosure relates to the determination method of the quality prediction model of machine learning techniques field more particularly to R&D team,
Determining device, electronic equipment and the computer readable storage medium of the quality prediction model of R&D team.
Background technique
Now with the development of computer, types of applications is also emerged in large numbers therewith, but the quality applied alternates betwwen good and bad, and the good and the bad is not
Together.In order to evaluate the quality of application, the quality of application is calculated based on some characteristic informations of application in the related technology.And it applies
It is to be researched and developed by R&D team, in order to further consider the quality of R&D team, is ground in the related technology based on the R&D team
The quality of the application of hair is weighted summation and obtains.
However under this calculation, the weight for being weighted summation is the calculation selected based on artificial experience, example
As for some application, weight is calculated based on following formula: the traffic of the application/all applications of the team are logical
The sum of traffic, wherein the traffic of application refers to using generated all valuable volumes of services.
Due to above-mentioned formula be it is selected based on artificial experience, for calculating the weight applied, very limited accuracy, such as
Valuable volume of services is not present in some applications, then calculating according to above-mentioned formula, weight is exactly 0, then this applies nothing
By the operation of program itself have that multithread is smooth, also can not as R&D team's quality with reference to, so this calculation in the presence of compared with
More unreasonable places.
Summary of the invention
The disclosure provide the determination method of quality prediction model of R&D team, R&D team quality prediction model really
Determine device, electronic equipment and computer readable storage medium.
According to the disclosure in a first aspect, proposing a kind of determination method of the quality prediction model of R&D team, the side
Method includes:
Based on the application fetches foundation characteristic information of R&D team's research and development, the foundation characteristic information is for characterizing described answer
Quality；
Each foundation characteristic information is extended respectively by preset algorithm, with the characteristic information that is expanded；
Sample set is constituted according to the application and corresponding foundation characteristic information and extension feature information；
Based on the sample set, preset model is trained by machine learning algorithm, to obtain quality prediction model,
Wherein, the input quantity of the quality prediction model includes the foundation characteristic information and the extension feature information, and output quantity is
The forecast quality of R&D team.
Optionally, the sample set includes training set and test set, described to be based on the sample set, is calculated by machine learning
Method is trained preset model, includes: to obtain quality prediction model
Based on the training set, preset model is trained by machine learning algorithm, to obtain model to be measured；
Based on the test set, the model to be measured obtained every time is adjusted by supervised learning, to obtain the matter
Measure prediction model.
Optionally, described to be based on the sample set, preset model is trained by machine learning algorithm, to obtain matter
Measure prediction model further include:
According to multiple machine learning algorithms, is executed described in above-described embodiment respectively and obtains the process of quality prediction model,
To obtain multiple quality prediction models；
Based on the test set, the difference of each quality prediction model corresponding forecast quality and actual mass is calculated
It is different；
Retain the corresponding quality prediction model of the smallest difference, and deletes other quality prediction models.
Optionally, the method also includes:
The parameter that each characteristic information is corresponded to according to the quality prediction model determines at least one in the characteristic information
A important feature information；
The important feature information is monitored；
When the variable quantity of the important feature information is greater than preset value, prompt information is generated.
Optionally, the preset algorithm includes at least one of:
Calculating is most worth, and calculates mean value, calculates and calculates applying in research and development group for the foundation characteristic information for meeting preset condition
Ratio in the application of team's research and development calculates applying in the respective default item of all satisfactions for the foundation characteristic information for meeting preset condition
Ratio in the application of the foundation characteristic information of part, wherein every kind of foundation characteristic information respectively corresponds a preset condition.
According to the second aspect of the disclosure, a kind of determining device of the quality prediction model of R&D team, the dress are proposed
It sets and includes:
Characteristic extracting module, the application fetches foundation characteristic information for being researched and developed based on R&D team, the foundation characteristic
Information is used to characterize the quality of the application；
Feature expansion module, for being extended respectively by preset algorithm to each foundation characteristic information, with
To extension feature information；
Sample set constitutes module, for according to the application and corresponding foundation characteristic information and extension feature information structure
At sample set；
Machine learning module, for being trained to preset model by machine learning algorithm based on the sample set, with
Obtain quality prediction model, wherein the input quantity of the quality prediction model includes the foundation characteristic information and the extension
Characteristic information, output quantity are the forecast quality of R&D team.
Optionally, the sample set includes training set and test set, and the machine learning module includes:
Machine learning submodule, for being trained to preset model by machine learning algorithm based on the training set,
To obtain model to be measured；
Model adjusting submodule, for be based on the test set, by supervised learning to the model to be measured obtained every time into
Row adjustment, to obtain the quality prediction model.
Optionally, the machine learning submodule and model adjusting submodule are also used to be calculated according to multiple machine learning
Method is based respectively on the multiple quality prediction models of the training set；
The machine learning module further include:
Difference computational submodule calculates the corresponding prediction of each quality prediction model for being based on the test set
The difference of quality and actual mass；
Model filter submodule, for retaining the corresponding quality prediction model of the smallest difference, and it is pre- to delete other quality
Survey model.
Optionally, described device further include:
Important determining module, for corresponding to the parameter of each characteristic information according to the quality prediction model, in the spy
At least one important feature information is determined in reference breath；
Feature monitoring modular, for being monitored to the important feature information；
Cue module when being greater than preset value for the variable quantity in the important feature information, generates prompt information.
Optionally, the preset algorithm includes at least one of:
Calculating is most worth, and calculates mean value, calculates and calculates applying in research and development group for the foundation characteristic information for meeting preset condition
Ratio in the application of team's research and development calculates applying in the respective default item of all satisfactions for the foundation characteristic information for meeting preset condition
Ratio in the application of the foundation characteristic information of part, wherein every kind of foundation characteristic information respectively corresponds a preset condition.
According to the third aspect of the disclosure, a kind of electronic equipment is proposed, comprising:
Processor；
Memory for storage processor executable instruction；
Wherein, the processor is configured to executing the step in any of the above-described embodiment the method.
According to the fourth aspect of the disclosure, proposes a kind of computer readable storage medium, is stored thereon with computer program,
The program realizes the step in any of the above-described embodiment the method when being executed by processor.
Based on embodiment of the disclosure, due to the limited amount of foundation characteristic information, by being carried out to foundation characteristic information
Extension, the extension feature information of the available quality that can more characterize application, to pass through foundation characteristic information and extension
Characteristic information characterizes the quality of application more fully hereinafter, and the quality applied can then characterize the quality of R&D team, therefore can
To characterize the quality of R&D team more fully hereinafter by foundation characteristic information and extension feature information.
And foundation characteristic information can only characterize the quality of some application mostly, and cannot characterize all using whole matter
Amount, and the quality of R&D team is all corresponded to using whole quality, and after being extended to foundation characteristic information polarity, such as to certain
A little foundation characteristic information summations, it is ensured that all using whole quality, and then realize the quality of characterization R&D team.
In turn, the sample set based on this building obtains quality prediction model come machine learning, is on the one hand directed to each feature
The parameter of information is obtained by machine learning, and opposite artificial settings is more accurate, on the other hand since characteristic information includes
Foundation characteristic information and extension feature information can comprehensively characterize the quality of R&D team, more accurate so as to obtain
Quality prediction model, so that the subsequent quality to R&D team is predicted.
It should be understood that above general description and following detailed description be only it is exemplary and explanatory, not
The disclosure can be limited.
Detailed description of the invention
The drawings herein are incorporated into the specification and forms part of this specification, and shows the implementation for meeting the disclosure
Example, and together with specification for explaining the principles of this disclosure.
Fig. 1 is a kind of showing for the determination method of the quality prediction model of the R&D team shown in accordance with an embodiment of the present disclosure
Meaning flow chart.
Fig. 2 is that the one kind shown in accordance with an embodiment of the present disclosure is based on the sample set, by machine learning algorithm to pre-
If model is trained, to obtain the schematic flow diagram of quality prediction model.
Fig. 3 is that the another kind shown in accordance with an embodiment of the present disclosure is based on the sample set, passes through machine learning algorithm pair
Preset model is trained, to obtain the schematic flow diagram of quality prediction model.
Fig. 4 is the determination method of the quality prediction model of another R&D team shown in accordance with an embodiment of the present disclosure
Schematic flow diagram.
Fig. 5 is terminal where the determining device of the quality prediction model of the R&D team shown in accordance with an embodiment of the present disclosure
Or a kind of hardware structure diagram of server
Fig. 6 is a kind of showing for the determining device of the quality prediction model of the R&D team shown in accordance with an embodiment of the present disclosure
Meaning block diagram.
Fig. 7 is a kind of schematic block diagram of the machine learning module shown in accordance with an embodiment of the present disclosure.
Fig. 8 is the schematic block diagram of another machine learning module shown in accordance with an embodiment of the present disclosure.
Fig. 9 is the determining device of the quality prediction model of another R&D team shown in accordance with an embodiment of the present disclosure
Schematic block diagram.
Specific embodiment
Example embodiments are described in detail here, and the example is illustrated in the accompanying drawings.Following description is related to
When attached drawing, unless otherwise indicated, the same numbers in different drawings indicate the same or similar elements.Following exemplary embodiment
Described in embodiment do not represent all implementations consistent with this disclosure.On the contrary, they be only with it is such as appended
The example of the consistent device and method of some aspects be described in detail in claims, the disclosure.
Fig. 1 is a kind of showing for the determination method of the quality prediction model of the R&D team shown in accordance with an embodiment of the present disclosure
Meaning flow chart.Method shown in the present embodiment can be applied to the electronics such as terminal, such as mobile phone, tablet computer, wearable device
Equipment also can be applied to service.As shown in Figure 1, the method may include following steps:
Step S1, based on the application fetches foundation characteristic information of R&D team's research and development, the foundation characteristic information is used for table
Levy the quality of the application；
Step S2 is extended each foundation characteristic information by preset algorithm, respectively with the feature that is expanded
Information；
Step S3 constitutes sample set according to the application and corresponding foundation characteristic information and extension feature information；
Step S4 is based on the sample set, is trained by machine learning algorithm to preset model, pre- to obtain quality
Survey model, wherein the input quantity of the quality prediction model includes the foundation characteristic information and the extension feature information, defeated
Output is the forecast quality of R&D team.
In one embodiment, foundation characteristic information, which can according to need, extracts, wherein can extract a basis
Characteristic information can also extract a plurality of foundation characteristic information.It is the example of several foundation characteristic information below:
Slowly the quantity inquired, slow inquiry refer to SQL (Structured Query Language, structured query language)
The inquiry duration of sentence is greater than the case where the first preset duration, and the quantity inquired slowly refers to the number that such case occurs, wherein
First preset duration, which can according to need, to be configured, such as can be set to 1 second, may be set to be 3 seconds (below mainly with
Illustrated for 3 seconds).If the value is larger, illustrate that the SQL statement inquiry duration of all applications entirety of R&D team is longer,
The quality of R&D team is lower.
Slowly the number of being on the list inquired carries out the application when the frequency of occurrences inquired slowly is greater than the first default frequency range
Report, the number of being on the list inquired slowly is exactly the quantity of the application reported in this case, wherein the first predeterminated frequency can be according to need
It is configured, such as can be set to 10 times/week, 2 times/day etc..If the value is larger, illustrate to inquire the frequency of occurrences slowly higher
Using more, the quality of R&D team is lower.
Percentage line response, according to interface (such as service interface or Http interface of percentile) response when
Between descending (can also be ascending, carry out example mainly for descending situation below), it is corresponding when being used in default
Between multiple response operation in section be ranked up, wherein coming the response time of the response operation at preset percentage line.Wherein,
Preset percentage line, which can according to need, to be configured, such as can be set to 95, such as altogether operates 100 secondary responses according to sound
Descending sequence between seasonable, then percentage line response just refers to the response time for wherein coming the 95th response operation.
If the value is larger, illustrate this using Whole Response overlong time, the quality of R&D team is lower.
Percentage line responds number of being on the list, and is greater than the second preset duration whenever counting on percentage line response, to the application
It is reported, the response of percentage line is on the list the quantity of applications that number reports in this case, wherein the second preset duration can be with
It is configured as needed.
Report an error quantity, using the number of operation error.
The quantity that reports an error is on the list number, when the frequency of application operation error is greater than the second predeterminated frequency, is reported to the application,
The quantity that reports an error be on the list number be exactly the application reported quantity, wherein the second predeterminated frequency, which can according to need, to be configured.
The not available duration in predetermined period is applied in availability value, expression, which can according to need setting,
Such as one month, season etc..
Availability is on the list number, when the availability value of application is greater than third preset duration, is reported to the application, availability
Value number of being on the list is exactly the quantity for reporting application, wherein third preset duration, which can according to need, to be configured.
Further, it can be extended respectively by preset algorithm for each foundation characteristic information, wherein pre- imputation
Method, which can be, calculates most value (such as when maximum value, minimum), calculates mean value, calculates and calculates and meet the basic special of preset condition
The ratio etc. of reference breath applied in the application of R&D team's research and development.
By taking the quantity that reports an error as an example, calculating is most worth, such as maximum value, then being exactly the report for calculating all applications of R&D team
The maximum value of wrong quantity；Calculating mean value is exactly to calculate the mean value of the quantity that reports an error of all applications of R&D team；It calculates and is exactly
Calculate the sum of the quantity that reports an error of all applications of R&D team；The applying for foundation characteristic information that calculating meets preset condition is being ground
The ratio in the application of team's research and development is sent out, in the case where the frequency that preset condition is error is greater than the second predeterminated frequency, the ratio
Example refers to that the quantity that reports an error is on the list ratios of the number in all applications that R&D team is researched and developed.
Under represent part basis characteristic information and extension feature information is for reference:
| Characteristic information | Meaning |
| Has3sSlowSql | 3s inquires number of being on the list slowly |
| HasHigh95Line | 95 lines respond number of being on the list |
| HasCatError | Cat is on the list number extremely |
| IsQualityNotUpToStandard | Availability number whether up to standard of being on the list |
| Availability | Availability and |
| AvailabilityAvg | Availability mean value |
| AvailabilityMax | Availability maximum value |
| AvailabilityMin | Availability minimum value |
| Cat95Line | 95 line values and |
| Cat95LineAvg | 95 line value mean values |
| Cat95LineMax | 95 line value maximum values |
| Cat95LineMin | 95 line value minimum values |
| Has3sSlowSqlCount | 3s inquire slowly quantity and |
| Has3sSlowSqlCountAvg | 3s inquires number average value slowly |
| Has3sSlowSqlCountMax | 3s inquires quantity maximum value slowly |
| Has3sSlowSqlCountMin | 3s inquires quantity minimum value slowly |
| CatErrorCount | Cat report an error quantity and |
| CatErrorCountAvg | Cat reports an error number average value |
| CatErrorCountMax | Cat reports an error quantity maximum value |
| CatErrorCountMin | Cat reports an error quantity minimum value |
| Has3sSlowSqlRatio | 3s inquires number of being on the list slowly, accounts for the ratio of all applications |
| HasHigh95LineRatio | 95 lines respond number of being on the list, Zhan Suoyou application bar ratio |
| HasCatErrorRatio | Cat is on the list number extremely, accounts for the ratio of all applications |
| IsQualityNotUpToStandardRatio | Availability number up to standard of being on the list, accounts for the ratio of all applications |
| allErrorCount | It is all types of be on the list number statistics and |
| errorRatio | It is on the list item number accounting |
| Has3sSlowSqlErrorRatio | 3s inquires number of being on the list slowly, accounts for all types and is on the list several ratios |
| HasHigh95LineErrorRatio | 95 lines are on the list number, are accounted for all types and are on the list several ratios |
| HasCatErrorErrorRatio | CatError is on the list number, accounts for all types and is on the list several ratios |
| IsQualityNotUpToStandardErrorRatio | Availability number whether up to standard of being on the list accounts for all types and is on the list several ratios |
Wherein, 3s inquires the case where i.e. above-mentioned quantity inquired slowly of quantity is for 3 seconds slowly；The Cat quantity that reports an error is i.e. above-mentioned
The quantity that reports an error in foundation characteristic information；95 lines response i.e. above-mentioned percentage line response is with percentage for 95% exemplary feelings
Condition.
3s inquires number of being on the list slowly, accounts for the ratio of all applications, refers in preset condition to be that the frequency of occurrences inquired slowly is greater than
When the first default frequency range, 3s inquires the ratio being on the list Shuo in all applications that R&D team is researched and developed slowly；The response of 95 lines is on the list
Number, Zhan Suoyou application bar ratio refer to when preset condition is to count on percentage line response greater than the second preset duration, 95
Line responds ratio of the number in all applications that R&D team is researched and developed of being on the list；Cat is on the list number extremely, accounts for the ratio of all applications
Example refers to when preset condition be to be greater than the second predeterminated frequency using the frequency of operation error, and the quantity that reports an error is on the list several researching and developing
The ratio in all applications that team is researched and developed；Availability number up to standard of being on the list, accounts for the ratio of all applications, refers in preset condition
When being greater than third preset duration for the availability value of application, availability all applications up to standard that count and researched and developed in R&D team of being on the list
In ratio.
All types, which are on the list, counts the quantity for the application for referring to all foundation characteristic information for meeting respective preset condition,
3s is inquired example slowly as described above, 95 lines respond, Cat is abnormal, for four kinds of foundation characteristic information of availability value, and all types are on the list
Number refers to that this 3s inquire number of being on the list slowly, and the response of 95 lines is on the list number, and Cat is on the list number extremely, availability is up to standard be on the list it is the sum of several.
Due to the limited amount of foundation characteristic information, by being extended to foundation characteristic information, available more multipotency
The extension feature information of the quality of enough characterization applications, to pass through foundation characteristic information and extension feature information table more fully hereinafter
Levy the quality of application, and the quality applied can then characterize the quality of R&D team, thus can by foundation characteristic information and
Extension feature information characterizes the quality of R&D team more fully hereinafter.
And foundation characteristic information can only characterize the quality of some application mostly, and cannot characterize all using whole matter
Amount, and the quality of R&D team is all corresponded to using whole quality, and after being extended to foundation characteristic information polarity, such as to certain
A little foundation characteristic information summations, it is ensured that all using whole quality, and then realize the quality of characterization R&D team.
In turn, the sample set based on this building obtains quality prediction model come machine learning, is on the one hand directed to each feature
The parameter (such as weight) of information is obtained by machine learning, and opposite artificial settings is more accurate, on the other hand due to spy
Reference breath includes foundation characteristic information and extension feature information, can comprehensively characterize the quality of R&D team, so as to
To more accurate quality prediction model, so that the subsequent quality to R&D team is predicted.
It should be noted that machine learning algorithm used by the present embodiment can according to need and be selected, such as can
Can also be returned using linear using GBDT (Gradient Boosting Decision Tree, gradient promote decision tree)
Return, wherein the preset model trained is regression model.
Fig. 2 is that the one kind shown in accordance with an embodiment of the present disclosure is based on the sample set, by machine learning algorithm to pre-
If model is trained, to obtain the schematic flow diagram of quality prediction model.As shown in Fig. 2, the basis of embodiment shown in Fig. 1
On, the sample set includes training set and test set, and it is described to be based on the sample set, by machine learning algorithm to preset model
It is trained, includes: to obtain quality prediction model
Step S41 is based on the training set, is trained by machine learning algorithm to preset model, to be measured to obtain
Model；
Step S42 is based on the test set, is adjusted by supervised learning to the model to be measured obtained every time, with
To the quality prediction model.
In one embodiment, training set and test set can be divided into for sample set, wherein training set and test set
It can be by being obtained to the specimen sample in sample set.
Due to being directly trained by training set to preset model, feature in the quality prediction model that may cause
The parameter of information is inaccurate, and according to this implementation, supervised learning can be passed through based on test set to the mould to be measured obtained every time
Type is adjusted, and specifically can be the parameter of adjustment characteristic information, so that finally obtained quality prediction model can be quasi-
Really realize the prediction of corresponding R&D team's quality.
Fig. 3 is that the another kind shown in accordance with an embodiment of the present disclosure is based on the sample set, passes through machine learning algorithm pair
Preset model is trained, to obtain the schematic flow diagram of quality prediction model.As shown in figure 3, the base of embodiment shown in Fig. 2
It is described to be based on the sample set on plinth, preset model is trained by machine learning algorithm, to obtain quality prediction model
Include:
Step S43 is executed in embodiment illustrated in fig. 2 obtain quality prediction model respectively according to multiple machine learning algorithms
Process, to obtain multiple quality prediction models；
Namely according to each machine learning algorithm in multiple machine learning algorithms, following steps are executed respectively:
Step S41 is based on the training set, is trained by machine learning algorithm to preset model, to be measured to obtain
Model；
Step S42 is based on the test set, is adjusted by supervised learning to the model to be measured obtained every time, with
To the quality prediction model.
So as to respectively obtain a quality prediction model for every kind of machine learning algorithm.
Step S44 is based on the test set, calculates each corresponding forecast quality of the quality prediction model and practical matter
The difference of amount；
Step S45 retains the corresponding quality prediction model of the smallest difference, and deletes other quality prediction models.
Since every kind of machine learning algorithm remembers calculating process difference, the accuracy of finally obtained quality prediction model
Difference, according to this embodiment, it can be based on test set, calculate the corresponding forecast quality of each quality prediction model and
The difference of actual mass, wherein difference is bigger, indicates that quality prediction model is also not allowed, can delete calmly, and retain minimum
The corresponding quality prediction model of difference for subsequent prediction use, so as to guarantee prediction accuracy.
Fig. 4 is the determination method of the quality prediction model of another R&D team shown in accordance with an embodiment of the present disclosure
Schematic flow diagram.As shown in figure 4, on the basis of embodiment shown in Fig. 1, the method also includes:
Step S5 corresponds to the parameter of each characteristic information according to the quality prediction model, in the characteristic information really
At least one fixed important feature information；
Step S6 is monitored the important feature information；
Step S7 generates prompt information when the variable quantity of the important feature information is greater than preset value.
In one embodiment, the parameter of each characteristic information is corresponded to according to the quality prediction model, can determine mould
Those characteristic informations are more important in type.
Such as quality prediction model is the model obtained based on linear regression training, then the parameter of characteristic information is power
Value, can determine the importance of feature, wherein the bigger characteristic information of weight is more important according to weight.
Such as quality prediction model is the model obtained based on GBDT training, since decision tree is when carrying out node split,
Characteristic information can be traversed and attempt division, final choice makes division front and back square error reduce most features and divided,
In, the reduction amount of square error is exactly the parameter of characteristic information, i.e. the characteristic information that the reduction amount of square error is bigger is more important.
At least one identified important feature information, can be most important characteristic information, can be with former important several
A characteristic information,
It is affected since important feature information changes to the quality of R&D team, by believing the important feature
Breath is monitored, and can determine when important feature information changes in time, and in the variable quantity of the important feature information
When greater than preset value namely when the variable quantity of important feature information is larger, research and development group is reminded in time by generating prompt information
Team, so as to R&D team in time determine important feature information situation of change and react, avoid sole mass by compared with
It is big to influence.
Optionally, the preset algorithm includes at least one of:
Calculating is most worth, and calculates mean value, calculates and calculates applying in research and development group for the foundation characteristic information for meeting preset condition
Ratio in the application of team's research and development calculates applying in the respective default item of all satisfactions for the foundation characteristic information for meeting preset condition
Ratio in the application of the foundation characteristic information of part, wherein every kind of foundation characteristic information respectively corresponds a preset condition.
Corresponding with the embodiment of the method for the quality prediction model of foregoing task R&D team, present invention also provides grind
Send out the embodiment of the device of the quality prediction model of team.
The embodiment of the determining device of the quality prediction model of the application R&D team can be applied in terminal or server
On.Installation practice can also be realized by software realization by way of hardware or software and hardware combining.With software reality
It is to be deposited by the processor of terminal where it or server by non-volatile as the device on a logical meaning for existing
Corresponding computer program instructions are read into memory what operation was formed in reservoir.For hardware view, as shown in figure 5, being
The one of terminal where the determining device of the quality prediction model of the R&D team shown in accordance with an embodiment of the present disclosure or server
Kind hardware structure diagram, other than processor shown in fig. 5, memory, network interface and nonvolatile memory, embodiment
Terminal or server where middle device can also include other hardware generally according to the terminal or the actual functional capability of server,
This is repeated no more.
Fig. 6 is a kind of showing for the determining device of the quality prediction model of the R&D team shown in accordance with an embodiment of the present disclosure
Meaning block diagram.Method shown in the present embodiment can be applied to the electronics such as terminal, such as mobile phone, tablet computer, wearable device and set
It is standby, it also can be applied to service.As shown in fig. 6, the determining device of the quality prediction model of the R&D team includes following knot
Structure:
Characteristic extracting module 1, the application fetches foundation characteristic information for being researched and developed based on R&D team, the foundation characteristic
Information is used to characterize the quality of the application；
Feature expansion module 2, for being extended respectively by preset algorithm to each foundation characteristic information, with
To extension feature information；
Sample set constitutes module 3, for according to the application and corresponding foundation characteristic information and extension feature information
Constitute sample set；
Machine learning module 4, for being trained to preset model by machine learning algorithm based on the sample set,
To obtain quality prediction model, wherein the input quantity of the quality prediction model includes the foundation characteristic information and the expansion
Characteristic information is opened up, output quantity is the forecast quality of R&D team.
Fig. 7 is a kind of schematic block diagram of the machine learning module shown in accordance with an embodiment of the present disclosure.As described in Figure 7, exist
On the basis of embodiment illustrated in fig. 6, the sample set includes training set and test set, and the machine learning module 4 includes:
Machine learning submodule 41 instructs preset model by machine learning algorithm for being based on the training set
Practice, to obtain model to be measured；
Model adjusting submodule 42, for being based on the test set, by supervised learning to the model to be measured obtained every time
It is adjusted, to obtain the quality prediction model.
Fig. 8 is the schematic block diagram of another machine learning module shown in accordance with an embodiment of the present disclosure.As described in Figure 8,
On the basis of the embodiment shown in fig. 7, the machine learning submodule 41 and model adjusting submodule 42 are also used to according to more
A machine learning algorithm is based respectively on the multiple quality prediction models of the training set；
The machine learning module 4 further include:
It is corresponding pre- to calculate each quality prediction model for being based on the test set for difference computational submodule 43
The difference of mass metering and actual mass；
Model filter submodule 44 for retaining the corresponding quality prediction model of the smallest difference, and deletes other quality
Prediction model.
Fig. 9 is the determining device of the quality prediction model of another R&D team shown in accordance with an embodiment of the present disclosure
Schematic block diagram.As shown in figure 9, described device further include:
Important determining module 5, for corresponding to the parameter of each characteristic information according to the quality prediction model, in the spy
At least one important feature information is determined in reference breath；
Feature monitoring modular 6, for being monitored to the important feature information；
Cue module 7 when being greater than preset value for the variable quantity in the important feature information, generates prompt information.
Optionally, the preset algorithm includes at least one of:
Calculating is most worth, and calculates mean value, calculates and calculates applying in research and development group for the foundation characteristic information for meeting preset condition
Ratio in the application of team's research and development calculates applying in the respective default item of all satisfactions for the foundation characteristic information for meeting preset condition
Ratio in the application of the foundation characteristic information of part, wherein every kind of foundation characteristic information respectively corresponds a preset condition.
About the device in above-described embodiment, wherein modules execute the concrete mode of operation in related this method
Embodiment in be described in detail, no detailed explanation will be given here.
According to the third aspect of the disclosure, a kind of electronic equipment is proposed, comprising:
Processor；
Memory for storage processor executable instruction；
Wherein, the processor is configured to executing the step in any of the above-described embodiment the method.
According to the fourth aspect of the disclosure, proposes a kind of computer readable storage medium, is stored thereon with computer program,
The program realizes the step in any of the above-described embodiment the method when being executed by processor.
Those skilled in the art will readily occur to its of the disclosure after considering specification and practicing disclosure disclosed herein
Its embodiment.This application is intended to cover any variations, uses, or adaptations of the disclosure, these modifications, purposes or
Person's adaptive change follows the general principles of this disclosure and including the undocumented common knowledge in the art of the disclosure
Or conventional techniques.The description and examples are only to be considered as illustrative, and the true scope and spirit of the disclosure are by following
Claim is pointed out.
It should be understood that the present disclosure is not limited to the precise structures that have been described above and shown in the drawings, and
And various modifications and changes may be made without departing from the scope thereof.The scope of the present disclosure is only limited by the accompanying claims.
Claims (10)
1. a kind of determination method of the quality prediction model of R&D team, which is characterized in that the described method includes:
Based on the application fetches foundation characteristic information of R&D team's research and development, the foundation characteristic information is for characterizing the application
Quality；
Each foundation characteristic information is extended respectively by preset algorithm, with the characteristic information that is expanded；
Sample set is constituted according to the application and corresponding foundation characteristic information and extension feature information；
Based on the sample set, preset model is trained by machine learning algorithm, to obtain quality prediction model,
In, the input quantity of the quality prediction model includes the foundation characteristic information and the extension feature information, and output quantity is to grind
Send out the forecast quality of team.
2. the method according to claim 1, wherein the sample set includes training set and test set, the base
In the sample set, preset model is trained by machine learning algorithm, includes: to obtain quality prediction model
Based on the training set, preset model is trained by machine learning algorithm, to obtain model to be measured；
Based on the test set, the model to be measured obtained every time is adjusted by supervised learning, it is pre- to obtain the quality
Survey model.
3. according to the method described in claim 2, it is characterized in that, it is described be based on the sample set, pass through machine learning algorithm
Preset model is trained, to obtain quality prediction model further include:
According to multiple machine learning algorithms, the process of quality prediction model is obtained described in perform claim requirement 2 respectively, to obtain
Multiple quality prediction models；
Based on the test set, the difference of each quality prediction model corresponding forecast quality and actual mass is calculated；
Retain the corresponding quality prediction model of the smallest difference, and deletes other quality prediction models.
4. the method according to claim 1, wherein the method also includes:
The parameter that each characteristic information is corresponded to according to the quality prediction model determines at least one weight in the characteristic information
Want characteristic information；
The important feature information is monitored；
When the variable quantity of the important feature information is greater than preset value, prompt information is generated.
5. method according to claim 1 to 4, which is characterized in that the preset algorithm include it is following at least
One of:
Calculating is most worth, and calculates mean value, calculate and, calculate and meet the applying for foundation characteristic information of preset condition and ground in R&D team
Ratio in the application of hair calculates applying for the foundation characteristic information for meeting preset condition and meets respective preset condition all
Ratio in the application of foundation characteristic information, wherein every kind of foundation characteristic information respectively corresponds a preset condition.
6. a kind of determining device of the quality prediction model of R&D team, which is characterized in that described device includes:
Characteristic extracting module, the application fetches foundation characteristic information for being researched and developed based on R&D team, the foundation characteristic information
For characterizing the quality of the application；
Feature expansion module, for being extended respectively by preset algorithm to each foundation characteristic information, to be expanded
Open up characteristic information；
Sample set constitutes module, for constituting sample according to the application and corresponding foundation characteristic information and extension feature information
This collection；
Machine learning module is trained preset model by machine learning algorithm, for being based on the sample set to obtain
Quality prediction model, wherein the input quantity of the quality prediction model includes the foundation characteristic information and the extension feature
Information, output quantity are the forecast quality of R&D team.
7. device according to claim 6, which is characterized in that the sample set includes training set and test set, the machine
Device study module includes:
Machine learning submodule is trained preset model by machine learning algorithm, for being based on the training set to obtain
To model to be measured；
Model adjusting submodule adjusts the model to be measured obtained every time by supervised learning for being based on the test set
It is whole, to obtain the quality prediction model.
8. device according to claim 7, which is characterized in that the machine learning submodule and model adjusting submodule
It is also used to be based respectively on the multiple quality prediction models of the training set according to multiple machine learning algorithms；
The machine learning module further include:
Difference computational submodule calculates the corresponding forecast quality of each quality prediction model for being based on the test set
With the difference of actual mass；
Model filter submodule for retaining the corresponding quality prediction model of the smallest difference, and deletes other prediction of quality moulds
Type.
9. a kind of electronic equipment characterized by comprising
Processor；
Memory for storage processor executable instruction；
Wherein, the processor is configured to perform claim requires the step in any one of 1 to 5 the method.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
The step in any one of claims 1 to 5 the method is realized when execution.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811542200.4A CN109739750A (en) | 2018-12-17 | 2018-12-17 | The determination method and apparatus of the quality prediction model of R&D team |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811542200.4A CN109739750A (en) | 2018-12-17 | 2018-12-17 | The determination method and apparatus of the quality prediction model of R&D team |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109739750A true CN109739750A (en) | 2019-05-10 |
Family
ID=66359791
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811542200.4A Pending CN109739750A (en) | 2018-12-17 | 2018-12-17 | The determination method and apparatus of the quality prediction model of R&D team |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109739750A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111667107A (en) * | 2020-05-29 | 2020-09-15 | 中国工商银行股份有限公司 | Research and development management and control problem prediction method and device based on gradient random forest |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2015060237A (en) * | 2013-09-17 | 2015-03-30 | 日本電気株式会社 | Prediction model learning device, prediction model learning method, and computer program |
| CN105279692A (en) * | 2015-11-17 | 2016-01-27 | 中国建设银行股份有限公司 | Financial information technology system performance prediction method and apparatus |
| CN106294902A (en) * | 2015-05-28 | 2017-01-04 | 阿里巴巴集团控股有限公司 | Method, device and the electronic equipment of prediction mobile applications page performance |
| CN108846695A (en) * | 2018-06-07 | 2018-11-20 | 中国联合网络通信集团有限公司 | The prediction technique and device of terminal replacement cycle |
| CN108897834A (en) * | 2018-06-22 | 2018-11-27 | 招商信诺人寿保险有限公司 | Data processing and method for digging |
-
2018
- 2018-12-17 CN CN201811542200.4A patent/CN109739750A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2015060237A (en) * | 2013-09-17 | 2015-03-30 | 日本電気株式会社 | Prediction model learning device, prediction model learning method, and computer program |
| CN106294902A (en) * | 2015-05-28 | 2017-01-04 | 阿里巴巴集团控股有限公司 | Method, device and the electronic equipment of prediction mobile applications page performance |
| CN105279692A (en) * | 2015-11-17 | 2016-01-27 | 中国建设银行股份有限公司 | Financial information technology system performance prediction method and apparatus |
| CN108846695A (en) * | 2018-06-07 | 2018-11-20 | 中国联合网络通信集团有限公司 | The prediction technique and device of terminal replacement cycle |
| CN108897834A (en) * | 2018-06-22 | 2018-11-27 | 招商信诺人寿保险有限公司 | Data processing and method for digging |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111667107A (en) * | 2020-05-29 | 2020-09-15 | 中国工商银行股份有限公司 | Research and development management and control problem prediction method and device based on gradient random forest |
| CN111667107B (en) * | 2020-05-29 | 2024-05-14 | 中国工商银行股份有限公司 | Research and development management and control problem prediction method and device based on gradient random forest |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tabandeh et al. | Uncertainty propagation in risk and resilience analysis of hierarchical systems | |
| AU2017235914B2 (en) | Improving performance of communication network based on end to end performance observation and evaluation | |
| CN108388852B (en) | Regional crowd density prediction method and device based on deep learning | |
| US9380122B1 (en) | Multi-platform overlap estimation | |
| Knight et al. | Use of fuzzy logic for predicting design cost overruns on building projects | |
| AU2011238471B2 (en) | Measurements based on panel and census data | |
| WO2008102255A1 (en) | System and method for routing tasks to a user in a workforce | |
| CN110147925B (en) | Risk decision method, device, equipment and system | |
| CN107943809A (en) | Data quality monitoring method, device and big data calculating platform | |
| US8825726B2 (en) | Method of generating statistical opinion data | |
| WO2012162671A1 (en) | Combining measurements based on beacon data | |
| CN112702224B (en) | Method and device for analyzing quality difference of home broadband user | |
| Yamauchi | Social learning, neighborhood effects, and investment in human capital: Evidence from Green-Revolution India | |
| CN106708692A (en) | Method and device for building filtering alarm model, filtering alarm method and device and electronic equipment | |
| EP2612256A1 (en) | Performing what-if analysis | |
| CN109741818A (en) | Resource allocation management method and device are intervened in medical inferior health based on artificial intelligence | |
| CN112785089A (en) | Agent service configuration method and device, electronic equipment and storage medium | |
| Singh et al. | Application of software packages for monthly stream flow forecasting of Kangsabati River in India | |
| CN113723692A (en) | Data processing method, apparatus, device, medium, and program product | |
| CN109739750A (en) | The determination method and apparatus of the quality prediction model of R&D team | |
| US20160226733A1 (en) | Method and score management node for supporting service evaluation | |
| WO2021098652A1 (en) | Data processing method and device | |
| CN109995549B (en) | Method and device for evaluating flow value | |
| CN110443451B (en) | Event grading method and device, computer equipment and storage medium | |
| CN116909534A (en) | Operator flow generating method, operator flow generating device and storage medium |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190510 |