CN103208039B - Method and device for evaluating software project risks - Google Patents
Method and device for evaluating software project risks Download PDFInfo
- Publication number
- CN103208039B CN103208039B CN201210009752.5A CN201210009752A CN103208039B CN 103208039 B CN103208039 B CN 103208039B CN 201210009752 A CN201210009752 A CN 201210009752A CN 103208039 B CN103208039 B CN 103208039B
- Authority
- CN
- China
- Prior art keywords
- risk
- project
- evaluation
- mentioned
- 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.)
- Expired - Fee Related
Links
Abstract
The invention aims at providing a method for evaluating software project risks. According to the method, a software project risk evaluation model with high risk occurrence probability judgment accuracy can be established. The method includes collection: collecting historical data about project indexes and project risks of a plurality of samples; segmentation: segmenting a range mapped by the project indexes into a plurality of sub-ranges and getting a representative value aiming at each sub-range, wherein the range is an m-dimensional vector space; statistics: conducting statistics ratio of the number of project index mapping points occurring with one project risk in each sub-range and the number of all project index mapping points mapped in the sub-range and utilizing the ratio as the occurrence probability of the project risk; and calculation: conducting normalization processing according to the representative value and the risk probability obtained by statistics and utilizing the processed data as training data to manufacture a project risk evaluation model which is used for risk evaluation of a new project.
Description
Technical field
The present invention relates to a kind of software project risk evaluation methodology and software project risk evaluating apparatus, more particularly to pass through
Quantitative relationship between analysis project index and project risk is used on model to evaluate new projects setting up risk evaluation model
In various risk probability of happening, and the method and apparatus that model is optimized.
Background technology
Often occur various risks during various software project developments.Such as, in some project development process,
There is situation that main developer leaves office, frequently repeated code error or situations such as frequently demand is changed.More than this
It is a little to claim project risk to the occurrence of probability that project development produces impact.Project risk in development process can be to project
Impact is produced, causes some to negatively affect.Some project risks cause project schedule to postpone, and some project risks increase project and open
Cost is sent out, some project risks reduce software quality.Accordingly, it is desirable to be able to be directed to each project forecast each project risk
Probability of happening, to take measures to be prevented in early stage.
It is that the potential wind of project is recognized in this area to carry out evaluation to project risk probability of occurrence using risk evaluation model
The effective ways of danger.In fact, availability risk appraisement system includes multiple targets for example, risk probability of happening is evaluated, risk is tight
Weight degree evaluation, the prediction of risk time of occurrence, the evaluation of venture influence scope etc..The present invention processes emphatically commenting for risk probability of happening
Valency.
Existing Project Risk Evaluation has following several.
For example there is the method that simple evaluation is carried out to risk probability by the way of questionnaire.In non-patent literature 1：
(paper) " Software Risk Management:Principles and Practices”,Barry Boehm,IEEE
Software, teaches a kind of method with questionnaire survey to carry out risk assessment in 1991..In the method, set by expert
The special questionnaire of meter, for collecting some with the related project information of risk.By the questionnaire carried out to project administrator
Investigation, collection management person scores the view of item status.Based on the scoring collected, certain comprehensive analysis (such as AHP is carried out
Or DELPHI methods), draw the probability that risk occurs.After delivering from the achievement of author Barry Boehm, other scholar edges
Same route has carried out many follow-up studies and perfect, with questionnaire survey as elementary tactics.
But, the method has the larger uncertain and dependency to expert's questionnaire.Due to the process in questionnaire survey
In, project administrator is easily added to bias in scoring, has had influence on the accuracy of project appraisal, therefore can bring compared with
Big uncertainty.The expertise in addition, formulation of questionnaire places one's entire reliance upon.Changing after a kind of item types must require specially again
Family works out corresponding questionnaire, adjusts analysis method.This defines dependence to expert, have impact on the scope of application of questionnaire.Therefore, should
Method has that anthropic factor is big, accuracy is low and the unmanageable shortcoming of cost.
Additionally, also have using rule-based or model evaluation method, for example with the method analysis for earning value analysis (EVM)
The probability that risk occurs.
In patent documentation 2：A kind of this method is disclosed in (United States Patent (USP) 7669180B2)：It is that every group of risk factor are created
Risk assessment task has been each task definition decision flow diagram, is regular collection that decision flow diagram creates correlation, base
Automatically overall merit is carried out to risk in project data and regular collection.
Additionally, the risk model referred in patent documentation 2 sets up rule can be described with certain machine language, also, specially
Device involved by sharp document 2 provides the API of regular programming.User can be relatively freely edited suitable for oneself software project
Risk supervision rule.
A kind of rule-based risk evaluating method is described in patent documentation 2.These rules can with certain language or
Programming API carries out free editor so that the method has good autgmentability.But, the method has user to be needed voluntarily to compile
Volume or risk of selection detected rule problem.In fact, the manager for lacking experience less knows which kind of rule is applied to detection
The risk of their projects, does not know how to create effective rule yet.
Additionally, in non-patent literature 3：(paper) " Risk management method using data from EVM
in software development projects”,A.Hayashi,CIMCA 2008,IEEE Computer Society.
In describe a kind of risk evaluating method.Specifically, the technology is using methods such as EVM (earning value analysis), by project process
The various impacts being subject to are quantized into schedule delay natural law, then carry out risk assessment according to the schedule delay natural law for calculating.
Non-patent literature 3 is used as a kind of risk assessment technology based on EVM.Its target is to carry out quantization to project risk to comment
Valency, is particularly suited for the related risk assessment of software progress.But problem is present in non-patent literature 3, this method compared with
It is difficult to it is non-enter degree type index, such as the index in terms of project quality is analyzed, and draws risk assessment.That is, method
The scope of application is than relatively limited, also, the accuracy of the method for quantitatively evaluating disclosed in non-patent literature 3 is not also high.
The content of the invention
It is an object of the invention to provide a kind of higher software item of accuracy that can set up the judgement of risk probability of occurrence
The software project risk evaluation methodology of mesh risk evaluation model and software project risk evaluating apparatus.
The present invention is a kind of software project risk evaluation methodology, including：Collection step, collect multiple samples with regard to project
The historical data of index and project risk；Segmentation step, above-mentioned project indicator institute range of a mapping is divided into it is multiple, for every
Subvalue domain after individual segmentation replaces tabular value, if be m relative to the number of species of the project indicator of certain intermediate item risk, above-mentioned value
Domain is m gts, wherein, m is greater than 0 integer；Statistic procedure, according to the historical data collected, statistics is per height
The quantity of the project indicator mapping point occurred with certain intermediate item risk in codomain is all in the subvalue domain relative to being mapped to
The ratio of the quantity of project indicator mapping point, as the risk probability that such project risk occurs；Calculation procedure, according to above-mentioned generation
Above-mentioned risk probability after tabular value and statistics, is normalized, and as training data, makes Evaluation of Risk model；
And evaluation procedure, risk is carried out to new software project using the Evaluation of Risk model calculated in above-mentioned calculation procedure
Evaluate, obtain forecasting risk probability.
Additionally, the present invention can also be a kind of software project risk evaluating apparatus, including：Collector unit, collects multiple samples
This with regard to the project indicator and the historical data of project risk；Cutting unit, by the range of a mapping segmentation of above-mentioned project indicator institute
For multiple, replace tabular values for the subvalue domain after each segmentation, if the species number of the project indicator relative to certain intermediate item risk
Measure for m when, above-mentioned codomain is m gts, wherein, m is greater than 0 integer；Statistic unit, according to the history number collected
According to counting the quantity of the project indicator mapping point occurred with certain intermediate item risk in each subvalue domain relative to being mapped to the son
The ratio of the quantity of all items index mapping point in codomain, as the risk probability that such project risk occurs；Calculate single
Unit, according to the above-mentioned risk probability after above-mentioned representative value and statistics, is normalized, and as training data, makes project
Risk evaluation model；And evaluation unit, the Evaluation of Risk model calculated using above-mentioned computing unit is to new soft
Part project carries out risk assessment, obtains forecasting risk probability.
Software project risk evaluation methodology of the invention and software project risk evaluating apparatus, can be independent of specially
In the case of family's experience, the pests occurrence rule of various risks in historical data is found out, and be quantified as risk evaluation model, new
The probability that various risks occur is estimated based on the project indicator and evaluation model in project, so as to improve the accurate of risk assessment
Property.
Additionally, the method that present invention employs machine learning, finds historical risk occurrence law, equivalent to " rule automatically
Then " automatization is created, and by collecting various types of project risks and index, expand the scope of application of device.
According to the invention it is thus possible to set up the higher Risk Evaluation Model of Software Project of accuracy to help project pipe
Reason person has found the various risks in software project.Using the present invention modeling method and device can effectively utilizes historical data,
The rule of project risk appearance is found out, risk present in new projects is found, the dependence to expertise is reduced, risk is improved and is sent out
Existing accuracy rate, and it is applied widely.
Description of the drawings
Fig. 1 is the flow chart of the software project risk evaluation methodology that first embodiment is related to；
Fig. 2 is the block diagram of the software project risk evaluating apparatus that first embodiment is related to；
Fig. 3 is the flow chart of the software project risk evaluation methodology that second embodiment is related to；
Fig. 4 is the block diagram of the software project risk evaluating apparatus that second embodiment is related to；
Fig. 5 is the illustration of project risk record and history item index；
Fig. 6 is the flow chart for preparing risk model training data；
Fig. 7 is the illustration for illustrating risk model training data；
Fig. 8 is the illustration for illustrating to be obtained by vector space segmentation training data；
Fig. 9 is the flow chart that risk model is trained using SVM algorithm；
Figure 10 is the flow chart that project risk probability is calculated using risk model；
Figure 11 is the flow chart for illustrating the degree of association between analysis project risk and the project indicator；
Figure 12 is the instance graph of the degree of association between analysis project risk and the project indicator；
Figure 13 is the flow chart for illustrating to be trained using checking data risk evaluation model；
Figure 14 is to illustrate to adjust the flow chart that vector space splits distance；
Figure 15 is the flow chart for adjusting relevance threshold.
Specific embodiment
Hereinafter, referring to the drawings, preferred implementation for the present invention is illustrated.
(first embodiment)
Fig. 2 is the block diagram of the software project risk evaluating apparatus that first embodiment is related to.As shown in Fig. 2 the present invention
Software project risk evaluating apparatus include collector unit 200, cutting unit 201, statistic unit 203 and computing unit 204.
Wherein, collector unit 200 by the history data collection with regard to the project indicator and project risk of multiple samples to number
According to being preserved in storehouse 201.For example, the input that the accepted user of collector unit 200 is carried out by input equipment, or directly from outer
Part device imports historical data, comes historical data and ongoing new projects' data that automatic/hand collects software project, and
In storing data into data base 201.
As a preservation example in data base 201, it is possible to use table storage project risk is recorded and history item
Index.Fig. 5 is the illustration of project risk record and history item index, and Current software management is schematically shown in Figure 5
In some projects risk and the project indicator.As shown in figure 5, save project risk in form " project risk record " remembering
Record.The row record entry title of form the 1st, the 2nd row record institute occurrence risk type, the 3rd arranges and have recorded afterwards certain in certain project
Plant the daily frequency of risk.Wherein, " paying and postponing (Delivery occurred on 2011/7/5th in project 1 (Project 1)
Delay) " once, project 2 equally occurred in that " pay and postpone " on 2011/7/6th, and risk is once for risk.
Form " history item index " have recorded the various project indicators of project.The row record entry title of form the 1st, the 2nd
The collected project indicator of row record, the 3rd arranges and have recorded afterwards the daily value of certain index in certain project.As shown in figure 5, form
2nd, 3 rows have recorded " defect concentration " and " demand change scale " two indexs of project 1, the 4th, 5 rows have recorded project 2 this
The situation of change of two indexs.
Here, because the historical data of similar projects more has reference value to the risk analyses of new projects therefore false
If collected data belong to similar projects (such as belonging to embedded development project), or according to item types
(such as e-commerce project, mobile phone application item, middleware project) is classified.
Additionally, the species such as project risk and the project indicator pointed in Fig. 5 is only example, it is certainly also existing including other
Some project risks and the project indicator.
If the number of species of the project indicator that collector unit 200 is collected, relative to certain intermediate item risk is that (m is big to m
In 0 integer) when, that is, when having the m kind project indicators, the codomain of the m kind project indicators constitutes m gts.Cutting unit 201
The codomain as the m gts that the just above-mentioned project indicator is mapped is divided into multiple subvalue domains, for each segmentation
Subvalue domain afterwards replaces tabular value, for follow-up process.
Statistic unit 203 is counted in each subvalue domain and occurred with certain intermediate item risk according to the historical data collected
Project indicator mapping point quantity relative to be mapped in the subvalue domain all items index mapping point (in certain project own
The m dimensional vectors that the project indicator is constituted) quantity ratio, as the risk probability that such project risk occurs, relevant mapping
Details be described below.
Computing unit 204 is used to make Evaluation of Risk model, and specifically, computing unit 204 is according to above-mentioned representative
Above-mentioned risk probability after value and statistics, is normalized, and as training data, makes Evaluation of Risk model, root
According to the above-mentioned risk probability after above-mentioned representative value and statistics, it is normalized, as training data, makes project risk and comment
Valency model.Wherein, above-mentioned normalized can be carried out using existing normalized algorithm, also, obtain training number
According to after, you can Evaluation of Risk model is set up using existing forecast model method for building up, therefore, here is omitted detailed
Explanation.
The software project risk evaluation methodology that the first embodiment of the present invention is related to is illustrated following with Fig. 1.Fig. 1 is
The flow chart of the software project risk evaluation methodology that one embodiment is related to.As shown in figure 1, first, collector unit 200 collects many
Individual sample with regard to the project indicator and the historical data (step 100) of project risk.
Then, cutting unit 201 above-mentioned project indicator institute range of a mapping is divided into it is multiple, for each segmentation after
Subvalue domain replace tabular value, if relative to certain intermediate item risk the project indicator number of species be m when, above-mentioned codomain be m tie up to
Quantity space (step 101).
Then, statistic unit 203 is counted in each subvalue domain with certain intermediate item risk according to the historical data collected
The quantity of the project indicator mapping point of generation is relative to the quantity of all items index mapping point being mapped in the subvalue domain
Ratio, as the risk probability (step 102) that such project risk occurs.
Finally, computing unit 204 is normalized according to the above-mentioned risk probability after above-mentioned representative value and statistics,
As training data, Evaluation of Risk model (step 103) is made.
Below citing describes each step in detail.
In general, risk evaluation model is actually the quantization function between " risk probability " and " relevant item index "
Relation.In order to train Evaluation of Risk model, substantial amounts of (risk probability, relevant item index) data are needed.Data base
In existing history item index, can directly find, but corresponding historical risk probability needs to calculate.In order to calculate
Historical risk probability of occurrence, is defined as " the multiple sample cases under the conditions of same or similar the calculating of historical risk probability here
In example, certain risk odds ".Here add " case under condition of similarity " be because finding in the historical data it is multiple complete
The case of exactly the same condition is highly difficult." same or similar condition " is actually referred to " relevant item index is identical or phase
As under the conditions of ".For ease of calculating, it is all relevant item index that all items index is set here, so as to only for project kind
Classified, you can carry out in the software project risk evaluating apparatus that the data of certain intermediate item in data base are brought into the present invention
Process.
Fig. 6 is the flow chart for preparing risk model training data.It is to step 101～103 in the flow chart in Fig. 1
Illustrate.
As shown in fig. 6, the risk probability of happening is being defined as into " (possess same or like under the conditions of same or similar
The project indicator), the probability that the risk occurs " (step 800) when, statistic unit 203 is directed to certain risk i, finds out data base
In the record that occurs of multiple project risks i, m dimensional vector X are set up using m class achievement datas related to risk i in these projects
{x1,…,xm}；Because each project there are multigroup (risk record, index) data pair, many groups of realities of vectorial X can be obtained
Example, and the corresponding risk record (step 801) of every group of example；Here " data to " actually refer to that a certain mesh day risk is sent out
Raw situation and the relevant item index on the same day, on September 1st, 1 " is paid and is postponed " risk and occurs 1 time in certain project,
The index of correlation on the same day is " defect concentration=9.2K/LOC, demand changing range=5 module ", and corresponding data are to being exactly
(risk record=1, indicator vector X=(9.2,5)) (step 801).
Cutting unit 201 regards m dimensional vectors as m-dimensional space, in each dimension, the codomain of the dimension is divided into into L sections,
Exactly by whole m-dimensional space be divided into Lm it is interval, every interval central point is designated as into { x1c,x2c,…,xmcAs each section
Representative value (step 802).
Then, statistic unit 203 is mapped to all sample instances of vectorial X in m-dimensional space；It is in each central point
{x1c,x2c,…,xmcInterval in, statistics is grown in the interval sample size, and can count the sample that there occurs risk
Quantity.Using (risk generation sample number)/(total number of samples) as the interval risk probability (step 803).It is above-mentioned by counting
Risk probability in each interval, can obtain one group of (risk probability, indicator vector) data set, such as (15%, X1),
(88%, X2) ..., (75%, Xi), X hereiIt is each interval central point (step 804).
Finally, computing unit 204 is normalized the vectorial X values in above-mentioned data set, by each value conversion of vectorial X
For the value (step 805) between [0,1].So, just obtained can be used for risk evaluation model training it is multigroup (risk probability,
The project indicator) record.
In figure 6, statistic unit 203 first carries out the DUAL PROBLEMS OF VECTOR MAPPING occurred with risk, and cutting unit 202 carries out again codomain
Segmentation, but codomain segmentation can also be first carried out, then mapped.That is, the execution sequence of each step of the present invention is simultaneously
It is not limited to institute's example illustrative embodiments in Fig. 6.In the case where there is no data inheritance relation, the execution sequence of each step is to appoint
Meaning.
Additionally, in the above example, using the vector corresponding to the central point of the section (subvalue domain) after segmentation as representative value,
It is normalized with reference to the risk probability in the subvalue domain.But, acquired representative value is not limited to center in the present invention
Point.It can also be the vector of other points.For example by the special pattern comprising all mapping points in subvalue domain (such as circular)
The vector of heart point is used as representative value.Or the density degree of mapping point distribution that can also be in subvalue domain is calculating representative
Value, in a word, the definition mode of representative value is not limited in the concrete example for being enumerated here, can be according to project environment or side
Emphasis to set subvalue domain in any vector value as representative value.
Fig. 7 is the illustration for illustrating risk model training data.Figure 7 illustrates obtained by each step of Fig. 6
The example of training data, every a line of form have recorded the value of certain vector interval central point project indicator, and the interval in figure
The probability of happening (risk probability, the project indicator) of interior certain risk.In the example of Fig. 7, the various project indicators have been carried out returning
One changes.
The calculating process of risk probability is illustrated with reference to Fig. 8.Fig. 8 is to illustrate to obtain training data by vector space segmentation
Illustration.For convenience of description, illustrate by taking simple two-dimensional space as an example.In the software project risk evaluation side of the present invention
In method, mainly there are following characteristics.
(1) by the m dimension project indicator related to project risk, m gts are regarded as.In upper legend, " defect is close
Degree (Bug density) " and " demand change scale (Require change scale) " are regarded as and " pay and postpone " risk
Two kinds of related project indicators, therefore the two dimensional vector space such as figure is just obtained, " defect concentration ", used as abscissa, " demand becomes
More scale " is used as vertical coordinate.
(2) all index samples in data base and relevant risk record are mapped in vector space.Risky generation
Index sample represented with filled circles, devoid of risk occur index sample represented with open circles.In fig. 8, in data base
" defect concentration " of project 1 on the 2011/7/5th and " demand change scale " index be mapped to the sample point in vector space (7,3),
Because the sample is labeled as filled circles with " pay and postpone " risk, this two of same project 2 on the 2011/7/6th refer to
Mark be mapped to sample point (3,4), and be also filled circles.Conversely, two indexs of 2011/7/3 day be mapped to sample point (3,
1), but because of the same day there is not " pay and postpone " risk, therefore the sample point is open circles.
(3) according to the span of sample in each dimension, vector space is divided into into many parts.In this example, by each stroke of transverse axis, the longitudinal axis
3 deciles, whole two-dimensional space are divided to be divided into 3 × 3 deciles, 9 intervals.With regard to the division principle of boundary value, boundary value is included
In forward relatively minizone.For example, in fig. 8, transverse axis be divided into [0,3], (3,6], (6,9] etc. 3 intervals, the longitudinal axis divide
For [0,2], (2,4], (4,6] etc. 3 intervals.
(4) in each interval after division, the sample point that statistics falls in the interval is total, and the sample for occurring in that risk
Originally count out (filled circles number)." risk sample point number/sample point sum " is calculated i.e. as the risk probability in the interval
Go out the interval central point (star-like symbol in figure).By (risk probability, central point vector) in the interval, (risk is general used as one group
Rate, the project indicator) data.Such as, there are 2 sample points in the 2nd interval of left column in Fig. 8, wherein 1 sample point is occurred in that
Risk, the interval risk probability is 1/2=50%.According to transverse and longitudinal coordinate, can calculate the interval central point for (1.5,
3) obtain, therefore just one group (risk probability, project indicator) and be recorded as [50%, (1.5,3)].Wherein the project indicator (1.5,3)
For 2 dimensional vectors.In the same manner, can obtain in other intervals other two groups of records in figure [0%, (1.5,1)] and [100%,
(7.5,3)]。
(5) collect (risk probability, the central point vector) result of calculation in all intervals, that is, obtain can be used for risk assessment mould
Multigroup training data (risk probability, the project indicator) record of type training.
The division of two dimensional vector space is taught in upper example.In fact, this vector space division methods can expand completely
Open up multidimensional (dimension>=3).Below with four dimensional vectors (i.e. 4 indexs are related to certain risk) as example illustrating hyperspace
The example of segmentation.If four dimensional vectors are (x1,x2,x3,x4), risk probability calculation step is as follows in vector space segmentation and interval：
1) for vector per one-dimensional, from being divided into L sections (such as 5 sections) between its peak to peak.Such as, it is assumed that x1
Value between [0,50], then its codomain is divided into into x10,x11,…x145 sections altogether, x10The value of section is between [0,10], x11Section
Value between (10,20], and x14Section value between (40,50].
2) by each section of every dimension, each section with other 3-dimensionals carries out unduplicated combination, forms L × L × L × L combination.
Each combines namely an interval, such as (x10,x21,x33,x42).The interval center is central point of each dimension on this section,
Such as (x10c,x21c,x33c,x42c)。
3) in each interval after division, sample point of the searching data storehouse decline in the interval, and find these samples
The sample point of occurrence risk in this point.Such as data base declines in (x10,x21,x33,x42) sample point have 10, there is risk
Sample point have 4.In this example, it is considered as, in the close x of 4 intermediate item indexs10c,x21c,x33c,x42c(interval central point)
In the case of, risk probability is 4/10=40%.
The risk probability in each interval is counted, and records the interval centerpoint value, that is, obtain that substantial amounts of (risk is general
Rate, x1,x2,x3,x4) data combination.The set that 5 tuple is formed can be used for the training of risk evaluation model.Here,
The number of combination is exactly in fact the number of space demarcation interval, such as L4It is individual.Interval division is more, and each interval is less, interval
Interior sample point is less away from central point distance, and sample point condition is more approximate, and risk probability computational accuracy is higher, but amount of calculation is just
It is bigger.Conversely, interval division is fewer, single interval is bigger, and the sample point fallen in single interval is more, but risk probability precision
Decline, amount of calculation also declines.
The method for dividing statistical history risk probability based on vector space by more than, can obtain various risks
(historical risk probability, respective item index) data set.These data are by for the evaluation model training of various risks.Model is instructed
White silk can adopt support vector machine (SVM), neutral net, KNN (K-Nearest Neighbor algorithm：K nearest neighbor algorithm)
Etc. method, this specification illustrates the step of setting up forecast model using training data by taking SVM algorithm as an example.Training step is referring to attached
Fig. 9.First, computing unit 204 is that non-linear SVM selectes kernel function, for example：
Here σ is constant (step 1000).Then, the equation group of model training is provided for, for example：
Here c is constant, b and aiBe in equation it needs to be determined that parameter (step 1001).(risk is general by off-the-shelf
Rate, indicator vector) data set as equation in (yi,xi) substitute into, solve equation parameter b and aiValue (step 1002).Work as institute
After having parameter determination, risk evaluation model is created, such as：
Here, xiIt is related indicator vector { x1,x2,…,xm)(1003).So far, the probability assessment model of certain risk
Draw.The related data that various risks should be used is that every kind of risk trains respective probability assessment model.
Additionally, after Evaluation of Risk model is obtained, it is also possible to enter sector-style to new software project inside device
Evaluate danger.In this case, software project risk evaluating apparatus of the invention can also include evaluation unit, single by evaluating
Unit carries out risk assessment using the Evaluation of Risk model that computing unit 204 is calculated to new software project, obtains pre-
Survey risk probability.
Figure 10 is the flow chart that project risk probability is calculated using risk model.As shown in Figure 10, in new projects are collected
The m dimensional vector index related to risk i, by its normalization, and creates after m dimensional vectors X (step 1100), and evaluation unit is adopted
The risk evaluation model for training, the risk for substituting into the project indicator (m dimensional vector X) in new projects to calculate prediction occurs
Probability (step 1101).The size (step 1102) of risk possibility occurrence is judged according to result above.Meanwhile, can by with it is new
The relevant various data of project are saved into data base, used as the reference data of subsequent project risk assessment.
And, or, after new software project is completed, collector unit 200 collect the software project with
The project indicator information relevant with project risk, as historical data, remakes Evaluation of Risk model.So, it is new
The sample size of Evaluation of Risk model foundation increases, so as to result is more accurate.
According to the present invention, by this method that codomain is refined as into subvalue domain, can simply find out project risk and go out
Existing rule, finds risk present in new projects, reduces the dependence to expertise, improves the accuracy rate that risk finds, and
And it is applied widely.
(second embodiment)
In the first embodiment, if all items index is all that the project indicator related to risk is pre- to carry out risk
Survey the foundation of model.But, all it is relevant item index due to setting all items index, project kind is have ignored to the project indicator
Impact, and the high project indicator of the degree of association project indicator low with degree of association is mixed, increases the amount of calculation of model training,
And it is likely to decrease the accuracy of forecast model to a certain extent.Therefore, in this second embodiment, increase to collecting
Data carry out the process of early stage process, with degree of association as foundation, the species of the project indicator are screened, so as to reduce model
The amount of calculation of training, and improve the accuracy of forecast model.The second embodiment in the case where sampling range is larger especially
Effectively.
Apparatus main body in second embodiment constitute and predictive mode to set up mode similar with first embodiment,
Main difference is the data that collector unit 200 is collected to be carried out with Controlling UEP, only by indicator for coherence in the phase for specifying
The project indicator more than pass degree threshold value is reused for segmentation step as relevant item index, will multiple project indicators limit
For relevant item index, relevant item index to be only used for the generation of the training data of forecast model.
Fig. 4 is the block diagram of the software project risk evaluating apparatus that second embodiment is related to.The software project risk is evaluated
Device includes：Data collection module 400, Controlling UEP module 402, sample data analysis module 403, model regression analyses mould
Block 404 and risk probability evaluation module 405.
Wherein, data collection module 400 is equivalent to collector unit, collects the historical data of software project and ongoing
New projects' data, and store data in data base 401.
Controlling UEP module 402 is dividing function, statistical function and the Controlling UEP function with a dimension indicator
Module, certain the intermediate item risk time occurred according to the central value in each subvalue domain of certain intermediate item index and with the subvalue domain
Number, calculates the indicator for coherence between certain intermediate item risk and certain intermediate item index, only by indicator for coherence in regulation
The project indicator more than relevance threshold is split as relevant item index, the codomain for multi-C vector space.
Sample data analysis module 403 is equivalent to the module that the function of statistic unit and cutting unit becomes one.
For according to the correlation analysiss result of Controlling UEP module 402, the project related to risk being extracted from data base 401
Index, further in accordance with the workflow management of first embodiment (historical risk probability, relevant item index) data set is gone out.Wherein,
The flow process of described Fig. 6 is exactly to be completed by sample data analysis module 403 in one embodiment, therefore omits detailed description.
Again, for purposes of illustration only, statistic unit and cutting unit are included in into Controlling UEP module 402 and sample number
Illustrate according to analysis module 403, but can also only with independent statistic unit, cutting unit and sample data point
Analysis unit, the degree of association to be carried out present embodiment by the transmission of data is processed.In such case, statistic unit and segmentation list
Unit first sends the occurrence number of certain the intermediate item index occurred with certain intermediate item risk in come out, each subvalue domain
Controlling UEP is carried out to sample data analytic unit, then receive the result of Controlling UEP, merely with relevant item index weight
Newly carry out codomain segmentation and statistical disposition.The process of wherein codomain segmentation and statistical disposition is identical with first embodiment, therefore
Detailed.
Model regression analyses module 404 equivalent to computing unit, based on the continuous item that sample data analysis module 403 is obtained
The data set that the risk probability (representative value in subvalue domain) in mesh index and subvalue domain is constituted, as training data, trains various
The risk evaluation model of risk.Training flow process is identical with the Fig. 9 illustrated in first embodiment, therefore omits detailed description.
Risk probability evaluation module 405, based on these models and the index of correlation of new projects, is calculated equivalent to evaluation unit
Go out the probability of happening of various risks in new projects.
Fig. 3 is the flow chart of the software project risk evaluation methodology that second embodiment is related to.First, data collection module
The polytype project indicator of 400 collections, and the project risk of appearance is recorded, the project indicator collected, risk are saved in
In data base 401 (step 300).
Then, the degree of association between the analysis project index of Controlling UEP module 402 and project risk, sample data analysis
Module 403 and model regression analyses module 404 create the evaluation model of project risk probability, and the model being capable of quantization means project
Relation (step 301) between risk and index, specific Controlling UEP process is described in detail below.
It is determined that after risk evaluation model, risk probability evaluation module 405 collects the project indicator of new projects, adopts these
The project indicator and risk probability model calculate the probability of happening (step 302) of various risks.
So far, using the software project risk evaluating apparatus of the present invention, project administrator can be according to the various wind for drawing
Dangerous probability of happening carries out the judgement and prevention of risk.
The project indicator of new projects and practical risk occurrence record can also be saved in by number by data collection module 400
According to storehouse, as needed, using the data after renewal risk probability model (step 303) is updated.This is equivalent to the number to new projects
According to being used, the risk assessment in Future Project is reapplied.
Hereinafter the Controlling UEP to being related in second embodiment is described in detail.
Before the risk evaluation model of user-specific is trained, first have to find out the factor of influence in model, that is,
Find out with the related project indicator of risk.This index of identifying project is said with the process of risk degree of association in accompanying drawing 11
It is bright.
Controlling UEP module 402 performs the flow process illustrated in Figure 11.As shown in figure 11, for index and the wind of identifying project
The degree of association of danger, needs to carry out the project indicator and risk by analysis.First, a pair of risks R are selectediWith project indicator Mj(step
It is rapid 600), its degree of association is analyzed.By project indicator MjCodomain be divided into L it is interval, by each interval midpoint note
For Vm, each interval bound is designated as into Vd,Vu(step 601).Statistical indicator MjPositioned at each interval (Vd,Vu) when risk send out
Raw number of times (step 602).According to above-mentioned statistical result, L groups (risk number of times, M are drawnj) data adopt to (step 603)
Scheduling algorithm (can also eliminate the repeatability impact of similar index) using net correlation parser, and calculation risk goes out
Occurrence number and project indicator MjBetween relevance degree Wj(step 604).Here the result for being calculated using Pearson models is position
Value between [- 1,1].Then judge | Wj| whether more than threshold value Cr, CrIt is threshold value constant (step 605).If greater than
Threshold value Cr, illustrate risk RiWith index MjBetween have significant association (step 606)；Conversely, then illustrating risk RiWith index MjIt
Between association can ignore (step 607).Continue to analyze the degree of association between sundry item risk and index, until having analyzed
Some risks and indicator combination (step 608).By this analysis process, can find out with every kind of project risk be associated it is several
The project indicator is planted, these indexs will become the factor of influence of respective risk evaluation model.
Here, a pair of risks RiWith project indicator MjProcessed, carried out point for a certain project indicator on one-dimensional
Cut, the process can also be completed by cutting unit, or the partial function of cutting unit and statistic unit is incorporated to into degree of association point
Carry out in analysis unit.But the process being formally trained in the method and first embodiment of segmentation and statistics during data are extracted
Similar, it is one-dimensional that different is Controlling UEP, and in formal training data extraction flow process is the vector space of m dimensions.
Also, the density degree of the codomain segmentation carried out to carry out Controlling UEP can be with formal training data
The density degree for extracting the codomain segmentation carried out in flow process is identical, it is also possible to different.
Figure 12 is the instance graph of the degree of association between analysis project risk and the project indicator.As shown in figure 12, select a pair
Risk i and project indicator MjAfterwards, by project indicator MjCodomain be divided into L it is interval, each interval midpoint is designated as into Vm, will
Each interval bound is designated as Vd,Vu。MjItself be the curve of a change, its codomain i.e. maximum and minima it
Between scope.Statistical indicator MjPositioned at each interval (Vd, Vu) when risk occur number of times, the interval intermediate value be Vm.Each
Interval can obtain the two values, therefore, L " risk i frequency and project indicator M can be obtained altogetherj" data
It is right.For example " pay and postpone " data pair that risk occurrence number and " defect concentration " index are formed.Next, using Pearson
The degree of association of model calculation risk i and Metrics j.Here, E is risk i occurrence number or index MjExpected value.According to phase
Pass degree threshold value determines risk i and MjBetween degree of association it is whether notable enough.
According to the Controlling UEP of first embodiment, improve the project indicator on the whole with the degree of association of risk, therefore
Can be software project risk evaluating apparatus evaluation result it is more accurate.
(the first variation)
The basic structure and main flow of software project risk evaluation methodology according to the present invention and device are explained above,
But the present invention is not limited in above-mentioned embodiment, can be to carry out various modifications and optimization on the basis of main frame, this
It is included within the scope of the invention.It is described in detail below for several important variations.
First, need to create the probability assessment model of various risks in the present invention.In order to verify these risk evaluation models
Accuracy, the software project risk evaluating apparatus of the present invention can also include authentication unit, and the authentication unit is by historical data
In a part of data as checking data, substitute into Evaluation of Risk model and use forecasting risk probability to be verified, and count
The percentage ratio of the forecasting risk probability of the checking data and the difference of real history risk probability relative to real risk probability is calculated,
In the case that above-mentioned percentage ratio is more than regulation verification threshold, the Evaluation of Risk model is not used to carry out risk assessment.
Figure 13 is the flow chart for illustrating to be trained using checking data risk evaluation model.As shown in figure 13, first, will be initial
The historical data collected entered Controlling UEP and had the project risk and achievement data of significant correlation according to item
Mesh is divided into two groups, and one group is used for model training, and one group is used for model and verifies (step 700).
It can be that foundation is divided with project to divide above, such as using 7 one-tenth historical datas as training data, remaining 3 one-tenth
Historical data is used as checking data.The sample for model training is set up, sample includes historical risk probability and corresponding project
Achievement data (step 701).
Next, the method illustrated in first, second embodiment using more than extracts training data, and using SVM etc.
Regression algorithm trains the respective evaluation model (step 702) of every kind of risk.
Finally, these risk evaluation models, estimated risk evaluation model (step are updated to using the checking data for advancing for
It is rapid 703) testing whether the model accuracy for training is subjected to.
Specifically, the historical statistics in training data, calculates the real risk probability of checking data, then will instruction
The project indicator for practicing data substitutes into respectively risk evaluation model, obtains risk probability and predicts the outcome；To predict the outcome and true wind
Dangerous probability is compared.If both standard deviations are more than certain percentage threshold of real risk probability (such as more than true
Risk 20%), then it is assumed that the model accuracy is not enough, is abandoned；Conversely, it is new risk assessment mould then to receive the model
Type.
By above-mentioned verification step, the larger risk probability model of accuracy can be found and used, it is soft so as to improve
The precision of part Evaluation of Risk device.
(the second variation)
In above-mentioned historical risk probabilistic method, vector space is divided and plays key player.But how
It is a problem to prevent space division overstocked or dredge excessively.Therefore, cutting unit can also in advance carry out space stroke during codomain segmentation
Divide the judgement of density, it is determined that the partitioning scheme of optimum.
Figure 14 illustrates a kind of mechanism of dominant vector space division numbers, can divide overstocked or mistake with dominant vector space
Dredge.First cutting unit 202 selectes a kind of risk RiWith related m dimension indicators vector X (step 900), examination division is carried out, will be to
Often dimension X of amount XjCodomain is divided into 10 sections.Dividing line is interweaved, so as to form 10 in m gtsmIndividual interval (step
901)。
Then, the risky total sample number of institute is designated as into Nt, by the risk sample number in the most space of risk number of samples
Mesh is designated as Nm(step 902).Judge NmWhether (N is less thant/ 100), in NmLess than (Nt/ 100) when, be judged as interval excessive, divide
It is too intensive (step 903).Flow process enters step 904.
In step 904, the division hop count in every dimension is reduced every time 10%, judgement is re-started, until above-mentioned inequality
It is false.
If being negative in step 903, into step 905.Judge whether NmWhether (N is more thant/ 2) (step 905).
Certainly in the case of, it is judged as that interval is very few, divides too sparse, flow process enters step 906.Division hop count in every dimension is every
After secondary increase by 10%, judgement is re-started, until above-mentioned inequality is false.
If being negative in step 905, into step 907.More than adjust, can not be it is too intensive or
Too sparse vector space dividing mode.
The percentage ratio of above-mentioned progressive increase and decrease is not limited to 10%, be able to can be adjusted according to the calculating of equipment.Thus,
Can determine the optimal segmentation distance of vector space.
(the 3rd variation)
In the above-described 2nd embodiment, the degree of association between project risk and the project indicator is evaluated, determines phase
Guan Du carries out risk assessment higher than the project indicator of certain threshold value.But, exactly it is identical with change and the development of environment
Project risk and the project indicator, the degree of association between project risk and the project indicator is also during being continually changing.Therefore, phase
Pass degree analytic unit can also periodically carry out the amendment of relevance threshold, make relevance threshold be more nearly practical situation.
In this variation, Controlling UEP unit refers to the project abandoned with respect to certain intermediate item risk per the regular period
Mark is reused for the foundation of evaluation model to set up interim evaluation model, and the risk that will be obtained using interim evaluation model
Evaluation result compared with the risk evaluation results obtained using former evaluation model, in the risk assessment that interim evaluation model is obtained
As a result when the risk evaluation results for obtaining than former evaluation model are more nearly historical risk probability, the relevance threshold of regulation is adjusted
Degree of association corresponding to the whole project indicator abandoned for this.
Figure 15 is the flow chart for adjusting relevance threshold.As shown in figure 15, can be according to the precision pair of risk evaluation model
Index relevance threshold is modified.Whether central scope is to increase to judge threshold according to adjustment index of correlation post-evaluation precision
The necessity of value adjustment.
First, to risk i, the degree of association of all items index and the risk is calculated, is sorted by order of magnitude, to give tacit consent to
Threshold value t is divided into all items index two groups (step 1200) for boundary.
The selection of risk index of correlation is carried out according to the relevance threshold of acquiescence, based on index set (x1,…,xm), complete mark
The training (step 1201) of quasi-mode type.
Then, according to the descending order of degree of association, index x initially abandonedm+1Effective range is added to, is based on
Extended counter collection (x1,…,xm, xm+1), re -training model (step 1202).
Compare the precision (step 1203) of new model and master pattern.Precision comparison can be carried out based on checking data.It is first
First according to historical data, the real risk probability (for example trying to achieve by the step of Fig. 1 101,102) of checking data is calculated；So
Afterwards by the various project indicators of checking data, then new model and master pattern are substituted into respectively, risk probability prediction knot is drawn respectively
Really；Two kinds are predicted the outcome and is compared with real risk probability, which sees closer to true probability.If there is kinds of risks
Evaluation model or multiple projects checking data, in relatively new model and master pattern order of accuarcy, according to predicting the outcome
Carry out with the variance or standard deviation of true probability.
If the precision of new model is higher, relevance threshold is modified to into index xm+1Degree of association, and replaced with new model
Master pattern (step 1205) is changed, is further continued for attempting the indicator for coherence that the next one is abandoned being extended into index set (step
1206)；If new model precision is lower, illustrate to abandon index xm+1Be it is rational, relevance threshold and master pattern it is constant (step
It is rapid 1204).
It should be noted that increasing relevance threshold amendment step, may consequently contribute to improve risk assessment precision, but can increase
In the case that the amount of calculation of device, the particularly project indicator are more.This is because every time model training amount of calculation is all larger, at present
Threshold value correction strategy be premised on multiple re -training model.In practical situations both, it is best for computationally intensive consideration
Limit the index number related to risk.
In sum, the present invention is applicable in the evaluation of the software project risk in various fields, preferably helps project
Manager has found the various risks in software project.Being capable of effectively utilizes history number using the modeling method and device of the present invention
According to, the rule of project risk appearance is found out, risk present in new projects is found, the dependence to expertise is reduced, improve wind
The accuracy rate that danger finds, and it is applied widely.
Claims (16)
1. a kind of software project risk evaluation methodology, it is characterised in that
Including：
Collection step, collect multiple samples with regard to the project indicator and the historical data of project risk；
Segmentation step, above-mentioned project indicator institute range of a mapping is divided into multiple subvalue domain replacements for each after splitting
Tabular value, if relative to certain intermediate item risk the project indicator number of species be m when, above-mentioned codomain is m gts, its
In, m is greater than 0 integer；
Statistic procedure, according to the historical data collected, counts the project occurred with certain intermediate item risk in each subvalue domain
The quantity of index mapping point relative to the quantity of all items index mapping point being mapped in the subvalue domain ratio, as this
The risk probability that intermediate item risk occurs；
Calculation procedure, according to the above-mentioned risk probability after above-mentioned representative value and statistics, is normalized, used as training number
According to making Evaluation of Risk model；And
Evaluation procedure, risk is carried out using the Evaluation of Risk model calculated in above-mentioned calculation procedure to new software project
Evaluate, obtain forecasting risk probability.
2. software project risk evaluation methodology according to claim 1, it is characterised in that
In above-mentioned segmentation step, sample size in certain subvalue domain is will be mapped to compared with the amount threshold of regulation, root
The size in the subvalue domain of segmentation is adjusted according to comparative result.
3. software project risk evaluation methodology according to claim 1, it is characterised in that
Also include verification step, using a part of data in historical data as checking data, substitute into Evaluation of Risk model
Use forecasting risk probability to be verified, and calculate the checking forecasting risk probability with checking data historical risk probability it
Difference, in the case where above-mentioned percentage ratio is more than regulation verification threshold, does not use this relative to the percentage ratio of historical risk probability
Mesh risk evaluation model carries out risk assessment.
4. software project risk evaluation methodology according to claim 1, it is characterised in that
Also include feedback step, after new software project is completed, collect the software project with the project indicator and project
The relevant feedback of the information of risk gives above-mentioned collection step, as historical data, remakes Evaluation of Risk model.
5. software project risk evaluation methodology according to claim 1, it is characterised in that
In above-mentioned segmentation step, the representative value of the central value as the subvalue domain in each subvalue domain is taken.
6. software project risk evaluation methodology according to claim 1, it is characterised in that
Also include Controlling UEP step, according to adjoint certain category in come out in above-mentioned statistic procedure, each subvalue domain
The occurrence number of certain intermediate item index that wind syndrome of ocular connectors due to invasion of pathogenic wind nearly occurs, calculates the phase between certain the intermediate item risk and this certain intermediate item index
Pass degree index, the only project indicator using indicator for coherence more than the relevance threshold of regulation are used to divide as relevant item index
Cut step,
In above-mentioned segmentation step, only above-mentioned relevant item index institute's range of a mapping is split as m gts.
7. software project risk evaluation methodology according to claim 6, it is characterised in that
In above-mentioned Controlling UEP step, the project indicator abandoned with respect to certain intermediate item risk is used again per the regular period
Interim evaluation model is set up in the foundation of evaluation model, and compare historical data is substituted into respectively interim evaluation model and
Risk evaluation results obtained from former evaluation model, in the forecasting risk probability of the risk evaluation results as interim evaluation model
When the forecasting risk probability for being used for the risk evaluation results of former evaluation model is more nearly historical risk probability, by the phase of regulation
Pass degree adjusting thresholds are the degree of association corresponding to the project indicator that this is abandoned.
8. software project risk evaluation methodology according to claim 1, it is characterised in that
In above-mentioned calculation procedure, using support vector machine, neutral net or k nearest neighbor algorithm KNN, training data is modeled
Process to make Evaluation of Risk model.
9. a kind of software project risk evaluating apparatus, it is characterised in that
Including：
Collector unit, collect multiple samples with regard to the project indicator and the historical data of project risk；
Cutting unit, above-mentioned project indicator institute range of a mapping is divided into multiple subvalue domain replacements for each after splitting
Tabular value, if relative to certain intermediate item risk the project indicator number of species be m when, above-mentioned codomain is m gts, its
In, m is greater than 0 integer；
Statistic unit, according to the historical data collected, counts the project occurred with certain intermediate item risk in each subvalue domain
The quantity of index mapping point relative to the quantity of all items index mapping point being mapped in the subvalue domain ratio, as this
The risk probability that intermediate item risk occurs；
Computing unit, according to the above-mentioned risk probability after above-mentioned representative value and statistics, is normalized, used as training number
According to making Evaluation of Risk model；And
Evaluation unit, the Evaluation of Risk model calculated using above-mentioned computing unit carries out risk to new software project
Evaluate, obtain forecasting risk probability.
10. software project risk evaluating apparatus according to claim 9, it is characterised in that
Above-mentioned cutting unit will be mapped to sample size in certain subvalue domain compared with the amount threshold of regulation, according to comparing
As a result the size in the subvalue domain of segmentation is adjusted.
11. software project risk evaluating apparatus according to claim 9, it is characterised in that
Also include authentication unit, using a part of data in historical data as checking data, substitute into Evaluation of Risk model
Use forecasting risk probability to be verified, and calculate the checking forecasting risk probability with checking data historical risk probability it
Difference, in the case where above-mentioned percentage ratio is more than regulation verification threshold, does not use this relative to the percentage ratio of historical risk probability
Mesh risk evaluation model carries out risk assessment.
12. software project risk evaluating apparatus according to claim 9, it is characterised in that
After new software project is completed, above-mentioned collector unit collect the software project with the project indicator and project risk
Relevant information, as historical data, remakes Evaluation of Risk model.
13. software project risk evaluating apparatus according to claim 9, it is characterised in that
Above-mentioned cutting unit takes the representative value of the central value as the subvalue domain in each subvalue domain.
14. software project risk evaluating apparatus according to claim 9, it is characterised in that
Also include Controlling UEP unit, height that above-mentioned Controlling UEP unit comes out according to above-mentioned statistic unit, every
The occurrence number of certain the intermediate item index occurred with certain intermediate item risk in codomain, calculates certain intermediate item risk and certain class
Indicator for coherence between the project indicator, only the project indicator using indicator for coherence more than the relevance threshold of regulation is used as phase
Close the project indicator,
Above-mentioned cutting unit is only split above-mentioned relevant item index institute's range of a mapping as m gts.
15. software project risk evaluating apparatus according to claim 14, it is characterised in that
The project indicator abandoned with respect to certain intermediate item risk is reused for commenting by above-mentioned Controlling UEP unit per the regular period
The foundation of valency model setting up interim evaluation model, and by the risk evaluation results obtained using interim evaluation model and profit
The risk evaluation results obtained with former evaluation model compare, and the risk evaluation results obtained in interim evaluation model are evaluated than former
When the risk evaluation results that model is obtained are more nearly historical risk probability, the relevance threshold of regulation is adjusted to into what this was abandoned
Degree of association corresponding to the project indicator.
16. software project risk evaluating apparatus according to claim 9, it is characterised in that
Above-mentioned computing unit utilizes support vector machine, neutral net or k nearest neighbor algorithm KNN, and process is modeled to training data,
To make Evaluation of Risk model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210009752.5A CN103208039B (en) | 2012-01-13 | 2012-01-13 | Method and device for evaluating software project risks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210009752.5A CN103208039B (en) | 2012-01-13 | 2012-01-13 | Method and device for evaluating software project risks |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103208039A CN103208039A (en) | 2013-07-17 |
| CN103208039B true CN103208039B (en) | 2017-05-03 |
Family
ID=48755256
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210009752.5A Expired - Fee Related CN103208039B (en) | 2012-01-13 | 2012-01-13 | Method and device for evaluating software project risks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103208039B (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105721410B (en) * | 2014-12-05 | 2020-05-01 | 北京奇虎科技有限公司 | Method and device for acquiring network security condition |
| CN107292463A (en) * | 2016-03-30 | 2017-10-24 | 阿里巴巴集团控股有限公司 | A kind of method and system that the project evaluation is carried out to application program |
| CN106600007A (en) * | 2016-11-29 | 2017-04-26 | 上海亿账通互联网科技有限公司 | Activity quality evaluation method and device |
| CN106874933A (en) * | 2016-12-30 | 2017-06-20 | 平安科技（深圳）有限公司 | Modeling method and device based on data mining |
| CN108449307B (en) * | 2017-02-16 | 2020-12-29 | 上海行邑信息科技有限公司 | Method for identifying risk equipment |
| CN107480435B (en) * | 2017-07-31 | 2020-12-08 | 广东精点数据科技股份有限公司 | Automatic search machine learning system and method applied to clinical data |
| CN109376929B (en) * | 2018-10-24 | 2021-04-23 | 北京星选科技有限公司 | Distribution parameter determination method, distribution parameter determination device, storage medium, and electronic apparatus |
| CN109816221B (en) * | 2019-01-07 | 2024-01-02 | 平安科技（深圳）有限公司 | Project risk decision method, apparatus, computer device and storage medium |
| CN109978330A (en) * | 2019-02-20 | 2019-07-05 | 华电电力科学研究院有限公司 | A kind of department's safety management level measurement system |
| CN110399302A (en) * | 2019-07-26 | 2019-11-01 | 中国工商银行股份有限公司 | Risk Identification Method, device, electronic equipment and the medium of Software Testing Project |
| CN111143212B (en) * | 2019-12-24 | 2023-06-23 | 中国航空工业集团公司西安飞机设计研究所 | Functional logic function library verification method under module integrated software architecture |
| CN112363928B (en) * | 2020-11-10 | 2023-08-22 | 网易（杭州）网络有限公司 | Processing method and device of test case, processor and electronic device |
| CN112379913B (en) * | 2020-11-20 | 2022-01-07 | 上海复深蓝软件股份有限公司 | Software optimization method, device, equipment and storage medium based on risk identification |
| CN113297333A (en) * | 2021-03-17 | 2021-08-24 | 无锡极数宝大数据科技有限公司 | Data processing method, device, server and storage medium |
| CN113296836B (en) * | 2021-06-08 | 2022-07-22 | 北京百度网讯科技有限公司 | Method for training model, test method, device, electronic equipment and storage medium |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7669180B2 (en) * | 2004-06-18 | 2010-02-23 | International Business Machines Corporation | Method and apparatus for automated risk assessment in software projects |
| AU2006213709A1 (en) * | 2005-02-11 | 2006-08-17 | Volt Information Sciences Inc. | Project work change in plan/scope administrative and business information synergy system and method |
| CN101339596A (en) * | 2008-08-26 | 2009-01-07 | 腾讯科技（深圳）有限公司 | Method and device for protecting computer software system |
| CN101826202A (en) * | 2009-08-26 | 2010-09-08 | 张艳红 | Patch space theory system, model and implementation after project management risk line breakage |
-
2012
- 2012-01-13 CN CN201210009752.5A patent/CN103208039B/en not_active Expired - Fee Related
Also Published As
| Publication number | Publication date |
|---|---|
| CN103208039A (en) | 2013-07-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103208039B (en) | Method and device for evaluating software project risks | |
| JP7090936B2 (en) | ESG-based corporate evaluation execution device and its operation method | |
| Hurst et al. | Size-specific tree mortality varies with neighbourhood crowding and disturbance in a montane Nothofagus forest | |
| Wang et al. | Compositional vector space models for improved bug localization | |
| CN101556553A (en) | Defect prediction method and system based on requirement change | |
| US9633403B2 (en) | Managing sustainable intellectual property portfolio of an enterprise | |
| US20120041979A1 (en) | Method for generating context hierarchy and system for generating context hierarchy | |
| CN108197737A (en) | A kind of method and system for establishing medical insurance hospitalization cost prediction model | |
| CN104820724B (en) | Text class educational resource knowledge point forecast model preparation method and application method | |
| CN106815652A (en) | A kind of distribution network reliability Forecasting Methodology based on big data correlation analysis | |
| CN110225055A (en) | A kind of network flow abnormal detecting method and system based on KNN semi-supervised learning model | |
| CN109471934B (en) | Financial risk clue mining method based on Internet | |
| Herraiz et al. | Impact of installation counts on perceived quality: A case study on debian | |
| CN111738610A (en) | Public opinion data-based enterprise loss risk early warning system and method | |
| Theron | The use of data mining for predicting injuries in professional football players | |
| CN114518988B (en) | Resource capacity system, control method thereof, and computer-readable storage medium | |
| CN115630732A (en) | City operation-oriented enterprise migration big data monitoring and early warning method and device | |
| CN114548494A (en) | Visual cost data prediction intelligent analysis system | |
| CN113420946A (en) | News media evaluation method | |
| CN109690585A (en) | Information processing unit, information processing method and message handling program | |
| Wei et al. | Comparative performance of species-richness estimators using data from a subtropical forest tree community | |
| Lin et al. | Decision Tree analysis of the relationship between defects and construction inspection grades | |
| Luo | Progress indication for machine learning model building: A feasibility demonstration | |
| Bargh et al. | Dynamic data quality management framework using issue tracking systems | |
| Manning et al. | Volume and Direction of the Atlantic Slave Trade, 1650-1870: Estimates by Markov Chain Carlo Analysis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20170503 Termination date: 20200113 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |