CN107579844A - It is a kind of that failure method for digging is dynamically associated based on service path and frequency matrix - Google Patents
It is a kind of that failure method for digging is dynamically associated based on service path and frequency matrix Download PDFInfo
- Publication number
- CN107579844A CN107579844A CN201710710411.3A CN201710710411A CN107579844A CN 107579844 A CN107579844 A CN 107579844A CN 201710710411 A CN201710710411 A CN 201710710411A CN 107579844 A CN107579844 A CN 107579844A
- Authority
- CN
- China
- Prior art keywords
- mtd
- mrow
- msub
- mtr
- mtable
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Abstract
The present invention is a kind of to dynamically associate failure method for digging based on service path and frequency matrix:First, structure dynamically associates the data model of failure;2nd, the piecemeal Mining Strategy based on very big service path is designed;3rd, the scanning strategy based on frequency matrix is designed;4th, BFM algorithms global design.The present invention dynamically associates fault data model by structure and analysis judges feature, specify that excavation object, gives the optimization space of digging efficiency;Linked character of the onboard networks failure based on service path is analyzed, provides the partition strategy for avoiding noise combination, algorithm just filters out most of noise information before search iteration;Each piecemeal is decomposed into characteristic vector and frequency matrix, associate(d) matrix computing feature and frequent item set property, the cyclic policy that external memory scanning is greatly reduced is provided, greatly reduces algorithm cycle-index and comparing calculation;Step 4, in summary strategy devise BFM algorithms.The inventive method, the excavation speed of frequent item set can be effectively improved.
Description
Technical field
The invention belongs to onboard networks fault diagnosis technology field, more particularly to one kind to be based on service path and frequency matrix
Dynamically associate failure method for digging.
Background technology
Various Aerial Electronic Equipments on aircraft are connected by end system with interchanger, are formed by several centered on interchanger
Spoke subnet composition onboard networks system.Nearly half a century, undergoing scattered, joint, comprehensive, four ranks of high integrity
Duan Hou, the synthesization trend of onboard networks, cause equipment room relation complicated, influence each other more and more closely, showed between failure
Complicated incidence relation:One failure often triggers other multiple failures, these failure dynamic dependencies, constantly association triggering.Work
Cheng Zhong, in face of simultaneous multiple failures, in the case where its incidence relation is unknown, engineering staff can only investigate one by one, row
Therefore efficiency is low.
A large amount of fault datas are saved in onboard networks operation, contain related information between failure, but because of Mishap Database
Huge, it is unpractical manually to analyze searching one by one.And association rule mining main purpose is exactly from large data objectses, lead to
Cross historical information, excavate implying between individual or be not easy the incidence relation being found.Utilize association rule mining, Ke Yiyou
Effect finds out the incidence relation of onboard networks failure from Mishap Database.
Apriori algorithm is the classical Boolean Association Rules method for digging by propositions such as R.Agrawal.Algorithm is based on level
Iteration thought, according to minimum support, by connection and beta pruning, successively search iteration, until not meeting minimum support again
Item collection occur, Mining Frequent Itemsets Based, be then based on frequent item set generation Strong association rule.Algorithm is simple and practical, with a high credibility,
Strong applicability.However, the algorithm needs to be repeatedly scanned with transaction database (external memory) and connection, beta pruning circulation in excavating, follow every time
Ring is required for indifference combination producing candidate, then contrasts elimination noise information.Therefore, transaction database size, noise ratio
Digging efficiency is directly influenced again, and combination is also directly determined with eliminating amount of calculation size by transaction database.
Onboard networks system testing business is more, and transaction database noise proportion is larger, and failure, which produces, will not form class
Phenomenon is broken out like the concentration of communication network " alarm windstorm ", if Apriori algorithm is applied into onboard networks trouble correlation analytic,
The problems such as denoising is not thorough, sweep speed is slow will be caused.At present, traditional association rule digging method in actual applications more is based on
Apriori algorithm is improved, such as the methods of hash, Redundant Transaction Compression idea, sampling, dynamic item collection, though can solve efficiency to a certain degree
The problem of low, but during applied to similar onboard networks Mishap Database, can not solve onboard networks relevant fault number of transactions compared with
The more and unconspicuous problem of correlation, it is difficult to bring significantly improving for digging efficiency, engineering demand can not be met.Therefore, currently
Method for digging can not be applied to the association rule mining that onboard networks dynamically associate failure very well.
The content of the invention
The invention aims to realize to dynamically associate onboard networks the fast and effective standard of correlation rule between failure
True excavation, there is provided a kind of that failure method for digging is dynamically associated based on service path and frequency matrix, specially a kind of base
Piecemeal Mining Strategy in service path and the quick scanning strategy based on frequency matrix, this method support corresponding engineering software
Exploitation design, so as to which ancillary works personnel quickly carry out fault location in actual applications, complete fault diagnosis.
To achieve the above object, as shown in figure 1, technical scheme provided by the invention, which is one kind, is based on service path and frequency
Matrix dynamically associates failure method for digging, it is characterized in that methods described includes:
Step 1: structure dynamically associates the data model of failure;
This step dynamically associates failure by definition, and analysis dynamically associates fault verification feature, to dynamic in engineering practice
State relevant fault data are pre-processed, and structure onboard networks dynamically associate the data model of failure, so as to clear and definite this method
Object is excavated, this outer analysis simultaneously demonstrates the feasibility for dynamically associating fault verification feature reduction noise combination, so as to method stream
The statement of journey.
Step 101, clearly dynamically associate fault definition
In the present invention, the definition that onboard networks dynamically associate failure is:Using avionics business as flow in onboard networks operation
Each application between Dynamically logic connection relation be present, cause under certain proof stress, certain failure occur when, on current business road
On footpath association triggers new failure, and dynamically associating with service path, and failure shows dynamic dependency and constantly propagated
Feature.Failure with such dynamic associations is referred to as dynamically associating failure.
Wherein, service path, dynamic, relevance are defined respectively as:
Service path:Under certain proof stress, logic connecting relation is respectively applied in avionics operation flow.As what comes into a driver's strengthens
Control business, under the proof stress of particular device power-up, its service path is:What comes into a driver's enhancing → RDC → programmable integrated process →
Display processing unit.Fig. 2 is certain onboard networks moment service path schematic diagram of loading section avionics business.
Dynamic:Under same proof stress, failure generation is closely related with service path, and the dynamic of service path causes
The dynamic of failure.For example, Fig. 3 show certain research institute's physical fault case, certain test business can be realized by two paths,
Path 1:ADC1 → DPU1, path 2:ADC1→RDC1→DPU1.Experiment finds that ADC1 equally power-up and data is sent, test
During business passage path 1, the failure of vision signal shake misalignment can occur for DPU1, and during passage path 2, after RDC1,
This failure does not occur for DPU1.Obviously, same proof stress condition, some failures only can be just presented by specific transactions path
Out.
Relevance:Multiple failures may be caused by same failure in same test business, i.e., same proof stress and same
There is relevance between the failure of service path.For example, Fig. 4 is a typical case:Failure 1 shows as bus data monitoring and set
There is packet loss in the standby AFDX interchangers forwarding data that monitor;Failure 2 shows as without Distal promoter equipment alarm believing on display screen
Number display, " service path " and " proof stress " information of two failures is identical.Two faulty equipments be respectively AFDX interchangers with
Display screen, its failure procedure are:AFDX exchange data packet losses, cause the data-transmission interruptions on service path, and display is set
It is standby not receive valid data, so as to which presentation of information can not be carried out.It can be seen that in Mishap Database, have same from the example
Proof stress condition and the identical or intersecting failure of service path, it is most likely that be association triggering.
Step 102, determination dynamically associate the judgement feature of failure
Analyzed according to step 101, onboard networks failure has dynamically associates feature based on service path, its correlation rule
It is closely related with fault attribute " service path " and " proof stress ", therefore the present invention provides following judgement feature:
Feature 1:Service path is non-intersect, two different failures of proof stress, it is impossible to be that direct correlation triggers.
Feature 2:Two failures of indirect association triggering, the failure pair that can be triggered by limited individual direct correlation, find it
Incidence relation.
Step 103, fault data pretreatment and model construction
Investigate certain research institute's physical fault list information, failure mainly include phenomenon of the failure, faulty equipment, service path, should
The attributes such as power condition, according to method for digging demand, fault message model F constructed by the present invention is as follows:
Wherein, S={ S1,S2,...Si... } and it is fault attribute " experiment ID ", X={ X1,X2,...Xi... } and it is failure
Attribute " proof stress ", I=A, B ... } it is phenomenon of the failure,For fault attribute " service path ",For failure i service path, wherein nj,nk∈ n, n=1,2 ... } it is malfunctioning node equipment.Each parameter phase
The different value for answering sequence to be corresponding field in raw data base is formed by dictionary ordering map.
Example 1, if certain network topology, as shown in figure 5,1~7 is device node, research Mishap Database is found, the network operation
In 6 kinds of failures such as A~F often occur in 1,2,3,5,6,7 equipment.
According to fault message model, 12 failure loggings are as shown in table 1 after quantization,
Table 1 is pretreated Mishap Database
In step 104, analysis mining method, the feasibility for judging feature reduction noise combination is utilized
The judgement feature provided according to step 102, it can judge whether to there may be between two failures in advance to dynamically associate
Relation, so as to avoid the combination of noise in the connection step of frequent-item.
Example 2. is used as an affairs, shape based on the Mishap Database application Apriori algorithm shown in table 1 once to test
Into transaction set as shown in table 2, by demonstrating by L1Generate L2Process, it was demonstrated that, can be significantly using the feature 1 in step 102
Improve the feasibility of digging efficiency;If minimum support Supmin=2;
Table 2 is trade type database-transaction set
(1) trade type database-transaction set of scan table 2, L is generated1
Table 3 is 1 Frequent Set list
(2) carry out First Contact Connections step to calculate, generation candidate C2
Table 4 is 2 Candidate Set lists
Observe C2, according to judgement feature 1, contrast table 2, it is known that failure combines in table 4:AD、AE、AF、BD、BE、BF、CD、
CE, CF, its service path have no to intersect, and proof stress is entirely different, it is impossible to are relevant faults, therefore are unlikely to be frequent episode
Collection;These combinations are entirely the noise information of connection step generation;
In search iteration, each step of algorithm all can a large amount of similar noise information of combination producing, then beta pruning disappears one by one
Except wherein most, this greatly reduces digging efficiency and the degree of accuracy;Analyzed according to judgement feature 1 and example 2, this partial noise is
It is avoidable.
Therefore, in dynamically associating failure and excavating, using the judgement feature of service path, it is feasible to reduce noise combination
's;
Step 2: piecemeal Mining Strategy of the design based on very big service path;
This step proposes very big service path concept, designed on this basis according to the judgement feature for dynamically associating failure
The piecemeal Mining Strategy of frequent item set.
Step 201, propose very big service path concept
Very big service path:Travel through in fault database after institute faulty proof stress and service path attribute, if business road
Footpath R1~RnWith inclusion relation, and proof stress is Xi, wherein RjComprising remaining (n-1) bar service path, then (Xi,Rj) be
One very big service path.
Define service path vector R2Include R1Computing be R1@R2:if(X1=X2)&Len(R1)≤Len(R2),Then R1@R2.Wherein, Part (Ri, k) and represent vectorial RiThe continuous fragment that length is k forms
Gather (k≤Len (Ri)), Part (Ri,k)jRepresent in the set jth item (j=1,2 ..., Len (Ri)-k+1)。
If R1=(1,2,3,4), Len (R1)=4, Part (R1,3)2=(2,3,4), Part (R1, 3)=(1,2,3),
(2,3,4) }, if R2=(2,3), R3=(3,4,5), and X1=X2=X3, then R2@R1Set up, R3@R1It is invalid.
The Mishap Database that example is 3. 1 can generate following 4 very big service paths after traversal:
{(X1,(1,2,3)),(X2,(2,3)),(X3,(5,6,7)),(X4,(5,6))}
Step 202, clearly very big service path characteristic
Very big service path characteristic:The failure of two direct correlation triggerings, maximum probability betide same greatly business road
Under footpath, and as experiment carries out being intended to 100% with fault database increase, this probability.
Specificity analysis:In fact, under same stress condition, the failure that two service paths mutually include is most likely direct pass
Connection;Under same stress condition, two service paths have the failure of certain coincidence relation, the i.e. intersecting event in service path part
Barrier, it may be possible to direct correlation, but it is integrated with testing, after traveling through fault database, very big service path is more perfect, this
Possibility can be less and less.
Characteristic proves:As shown in fig. 6, the failure direct correlation of AB two on test path 1, as system testing is integrated, industry
Business Path extension is path 2, then the failure C on path 2 and AB is likely to relevant.Because two paths equally should
It is also turned under the conditions of power, and the precedent that two fault correlations occur is present, therefore can not be by two service paths with inclusion relation
Piecemeal.
Two paths with overlapping relation are also possible to fault correlation relation be present, the failure AB in path 1 and the event in path 3
There is certain incidence relation in barrier BD, simply two test business are currently without integrated, i.e., does not have that " Distal promoter is to showing in fault database
Show out of order precedent on the test service path of processor 2 ", therefore the current generation does not travel through out the very big industry comprising ABD also
Business path.But with the integrated more and more higher of experiment, very big service path is gradually perfect, the situation of final " partially overlapping "
Can be fewer and fewer.
If data are sufficiently large, path 1 and path 3 are not integrated into a service path yet, then failure ABD associations occur general
Rate is minimum.Because after system completes nearly all normal work to do, two paths are not integrated into very big service path yet, can recognize
It is small probability event to be also turned on for two paths in conventional operation.
Analysis can prove the characteristic of very big service path above.
Step 203, design piecemeal Mining Strategy
The very big service path characteristic provided according to step 202, proposes following piecemeal Mining Strategy:
(1) after generating transaction set, ergodic data storehouse:
Obtain very big service path set:
{(X1,R1),(X2,R2),...,(Xk,Rk)}
(2) very big service path set is utilized, database is divided into k blocks, as shown in table 5:
Table 5 is that database splits k blocks
(3) each piecemeal dredge operation isolation is carried out.Affairs frequency calculates, and is only added up in same transaction set piecemeal.Such as three
There is A in individual transaction set piecemeal, then respective meter frequency, the item in different transaction set piecemeals, connect walk when not combined crosswise.
Using the process of very big service path generation transaction set piecemeal as shown in example 4.
Example 4. sets certain network topology as shown in figure 8,1~7 is device node, and research Mishap Database is found, the network operation
In 7 kinds of failures such as A~G often occur in 1,2,3,4,5,7 equipment, totally 17 failure loggings.
The transaction set of algorithm generation is as shown in table 6:
Table 6 is the failure transaction set of case 4
Totally 8 affairs, algorithm connection need full transaction set combination to intersect to database when walking.In fact, need combined
There are three groups shown in table, be only possible to produce correlation rule inside only this three groups.Therefore, carried out using very big service path method
Piecemeal, the transaction set piecemeal of generation is such as
It is shown:
Table 7 is the transaction set database after piecemeal
When middle transaction set scans for iteration, deleted afterwards without first being combined failure different piecemeals, each piece is excavated mutually
Isolation, is carried out simultaneously.Although for example, include A in three piecemeals, during connection step operation, A frequency separate computations, and it is different
Item (such as B, C and D, E, and F, G etc.) between piecemeal does not combine when connecting and walking.Before and after contrasting piecemeal, connection step generation L2
Needed for the item collection C of candidate 22Number,Digging efficiency significantly improves.
Step 3: scanning strategy of the design based on frequency matrix;
After step 2 utilizes very big service path by database piecemeal, noise isolation is realized, excavation can be effectively improved
Efficiency.During scan round, to improve search iteration efficiency between each piecemeal, this step is on the basis of piecemeal, further frequently
Matrix and eigenvector method are spent, piecemeal is decomposed into characteristic vector and frequency matrix, utilizes matrix operation feature and frequent episode
Collect property, to reduce cycle-index and comparing calculation amount.
Step 301, structure frequency matrix and characteristic vector
Characteristic vector:If certain piecemeal transaction set TiInclude m mutually different affairs t1,t2,...,tm, by n mutual not phases
Same item I1,I2,...,InForm, wherein { ti},{IiSorted by dictionary, then claim n-dimensional vector (I1,I2,...,In) it is Ti's
Characteristic vector, it is designated asExample of the present invention represents failure item, therefore I with A, B, C etc.i=A, B, C....
Frequency matrix:If certain piecemeal transaction set TiInclude m mutually different affairs t1,t2,...,tm, its characteristic vector
ForThen existence anduniquess matrix Qi=(qjk)n×mSo that
Then claim QiFor TiFrequency matrix.Wherein qjk∈ N represent TiIn k-th of affairs tkJth item IjFrequency.
Therefore, a transaction set piecemeal can be analyzed to a characteristic vector and a frequency matrix.After piecemeal, changed in search
Dai Zhong, each piecemeal only need run-down characteristic vector, and carry out first moment battle array computing.
Example 5. can be by each piece of decomposition, final transaction set piecemeal based on the transaction set piecemeal shown in table 7, usage frequency matrix
It is as shown in table 8 with characteristic vector, frequency matrix:
Table 8 is each piecemeal frequency matrix and characteristic vector
Wherein:
Observation matrix is it can be found that often row addition is Count (C1), second and third row is Count (C2), first row is
Count(C3).Therefore frequency matrix both remains frequency information, and and can embodies the related information between items, in a scan need not
Ergodic data storehouse, contrast computing is carried out one by one, only need to be according to matrix operation, you can complete scanning.
Step 302, analysis frequency matrix properties
Frequency matrix has following property:
(1) the i-th row of frequency matrix sum is equal to the frequency of characteristic vector i-th:
(2) frequency matrix normalization is handled:
Wherein, frequency matrix jth row coefficient lambdajEqual to the frequency that jth arranges corresponding affairs:
Such as transaction set piecemeal T in example 51:
Count (ABC)=1, Count (AB)=1, Count (BC)=2
As can be seen that usage frequency matrix and its computing feature, can significantly simplify the comparing calculation in scan round.
Step 303, analysis frequent item set property
In frequent item set mining, following property be present:
(1) nonvoid subset of any frequent item set is also that frequently, also right and wrong are frequently for the supersets of nonmatching grids.
Inference 1A:If the frequency of all items is below minimum support in certain transaction set piecemeal, the transaction set piecemeal
In there is no frequent item set, can directly delete the transaction set piecemeal.
Inference 1B:If a certain purpose frequency in certain transaction set piecemeal be present is less than minimum support, the project is included
The equal right and wrong of any affairs frequently, directly can delete the project from transaction set piecemeal.
(2) if certain transaction length is less than k, the affairs do not support Lk+1。
Inference 2A:If the characteristic vector length of certain transaction set piecemeal is less than k+1, the maximum frequent set of the transaction set piecemeal
Integrate as Lk, the transaction set piecemeal can be directly deleted in circulation afterwards.
Inference 2B:If being less than k+1 in the presence of certain transaction length in certain transaction set piecemeal, L is generated in the transaction set piecemealkIt
The affairs can be directly deleted in circulation afterwards.
Step 304, design scanning strategy
The frequency matrix and frequent item set property provided based on step 301,302,303, this step are swept by designing five
Strategy is retouched, to greatly improve the sweep speed in excavating.
(1) scanning strategy 1~4 is designed:Frequency matrix and characteristic vector are compressed using minimum support, so as to compress affairs
Collect piecemeal,
Scanning strategy 1:According to inference 1A, if often row sum is respectively less than Sup to the frequency matrix of certain transaction set piecemealmin, then directly
Connect and delete the transaction set piecemeal:
Such as the transaction set piecemeal T of example 52Frequency matrix rows sum be respectively less than 2, therefore can directly delete the second piecemeal, press
Contracting database, improve scan efficiency:
Scanning strategy 2:According to inference 1B, if certain transaction set piecemeal has a frequency to be less than minimum support, i.e. frequency square
Certain row sum of battle array is less than Supmin, then the blocking characteristic is vectorial deletes this, and frequency matrix deletes the row, so as to compress the affairs
Collect piecemeal:
Such as transaction set piecemeal T in example 53Frequency matrix the third line sum be less than 2, therefore compress the piecemeal:Characteristic vector
G is deleted, frequency matrix deletes the third line, T3Compressible is T3', so as to improve scan efficiency:
Scanning strategy 3:According to inference 2A, if the characteristic vector length of certain transaction set piecemeal is less than k+1, in generation Lk
Afterwards, the transaction set piecemeal is directly deleted:
Generate Ck+1:
Scanning strategy 4:According to inference 2B, if certain row sum after certain transaction set piecemeal frequency matrix decomposition is less than k+1,
In generation LkAfterwards, the row are directly deleted:
Generate Ck+1:
(2) scanning strategy 5 is designed:Traditional algorithm is calculating Count (Ck) when need scan round transaction set, introduce frequency square
After battle array, using Matrix Properties, contrast scans work is skipped, directly obtains result with product.
Defined function jo (x):
Scanning strategy 5:The frequency matrix properties provided according to step 302, calculating the Count of certain transaction set piecemeal
(si),si∈CkWhen, by siIt is frequency vector by eigendecomposition, the product of frequency vector and frequency matrix can obtain Count
(si), contrasted without circulating:
Such as the transaction set piecemeal T of example 51, calculate Count (AC), AC ∈ C2:
Using scanning strategy 5, product calculation is directly carried out in scan round, without scanning the frequency of Frequent Set one by one,
Digging efficiency can be effectively improved.
Step 4: BFM algorithm global designs
Step 1 determines excavation object, and step 2 gives the piecemeal Mining Strategy of very big service path, and step 3 is given
Go out the scanning strategy based on frequency matrix, on this basis, this step global design BFM (Blocking Frequency
Matrix) method for digging.
Step 401, algorithm flow description
Algorithm ergodic data storehouse first, very big service path is generated, and utilize very big service path by database piecemeal.Root
Defined according to frequency matrix, the failure item collection of each piecemeal is decomposed into characteristic vector and frequency matrix.Then scanning strategy 1,2 is utilized
Simplification before scanning for, the piecemeal of the condition of being unsatisfactory for and, is deleted by scanning strategy 1, by sweeping according to the row of matrix in block form
Strategy 2 is retouched by matrix in block form depression of order.
1 Frequent Set of each piecemeal is firstly generated, starts the cycle over search iteration.Directly skipped using scanning strategy 3 discontented
The Circulant Block of sufficient condition, the further depression of order of frequency matrix is optimized using scanning strategy 4.Piecemeal after optimization is attached step
Generation candidate is walked with beta pruning, candidate frequency is calculated by scanning strategy 5, and then generate Frequent Set.Circulated in piecemeal
Terminate when current frequent item set is empty set, circulate in after the completion of all piecemeals and terminate between piecemeal, finally return that all frequent episodes
Collection.Fig. 9 show BFM algorithm flows.
Step 402, algorithm pseudo code
Input:Transaction database D, minimum support Supmin
Output:All frequent item set L
1.
2.XRi=find_i_maxpath (D);I-th very big service path of // generation
3.Ti=find_i_block (XRi,D);I-th of transaction set piecemeal of // generation
4.Iti=find_i_featurevector (Ti);The characteristic vector of i-th of transaction set piecemeal of // generation
5.Qi=find_i_frequentmatrix (Iti,Ti);The frequency matrix of i-th of transaction set piecemeal of // generation
6.[ni,mi]=size (Qi);The size of i-th of frequency matrix of // calculating
7.}
8.N=NUM (XRi);// calculate very big service path number
9.FOR (k=1;k≤N;k++){
10.{delete Qk,Itk;} // scanning strategy 1, utilizes QkRow and letter
Change { Tk}
11.ELSE{
12.
13.
14. } // scanning strategy 2, utilizes QkRow and simplified Tk
15.}
16.}
17.FOR EACH Qk(i,j){
18.IF(Qk(i, j)=0) { Ek(i, j)=0;}
19.ELSE{Ek(i, j)=1;λk(j)=Qk(i,j);} // by QkIt is normalized to matrix EkAnd corresponding factor lambdak
20.}
21.FOR (k=1;k≤N;K++) { // beginning piecemeal k search iteration
22.// utilize Qk1 Frequent Set of row and generation
23.// calculate EkRow and
24.The search iteration of { // start l item Frequent Sets
25.// scanning strategy 3, utilizes EkArrange and skip piecemeal k scanning
26.ELSE{
27.FOR EACH Mk(j){
28.// scanning strategy 4, utilizes EkRow and simplified Tk
29.}
30.}
31.Ck,l=apriori_gen (Lk,(l-1),Supmin);// walked by connecting step and beta pruning, generate l item Candidate Sets
32.FOR EACH si∈Ck,l{
33.gi=find_vector (si,Itk);// by affairs siIt is converted into vector
34.// scanning strategy 5, s is solved using matrix operationiFrequency
35.}
36.Lk,l={ si∈Ck,l|Count(si)≥Supmin};// return to l item Frequent Sets
37.}
38.// return piecemeal k all frequent item sets
39.}
40.All frequent item sets in // returned data storehouse
The present invention gives a kind of association rule mining method based on service path and frequency matrix, to efficient, accurate
Really excavate and show onboard networks and dynamically associate incidence relation between failure.
After analysis Apriori algorithm efficiency is mainly limited to external memory scanning and noise combination,
Step 1, fault data model is dynamically associated by structure and analysis judges feature, excavation object is specify that, provides
The optimization space of digging efficiency;
Step 2, linked character of the onboard networks failure based on service path is analyzed, provides the piecemeal for avoiding noise combination
Strategy, algorithm just filter out most of noise information before search iteration.
Step 3, each piecemeal is decomposed into characteristic vector and frequency matrix, associate(d) matrix computing feature and frequent item set
Matter, the cyclic policy that external memory scanning is greatly reduced is provided, greatly reduces algorithm cycle-index and comparing calculation;
Step 4, in summary strategy devise BFM algorithms.
Association rule mining method more than constructed by four steps, the excavation speed of frequent item set can be effectively improved
Rate.
Brief description of the drawings
Fig. 1 is to dynamically associate failure method for digging schematic flow sheet based on service path and frequency matrix.
Fig. 2 is onboard networks service path schematic diagram.
Fig. 3 is the failure dynamic schematic diagram based on service path.
Fig. 4 is the fault correlation schematic diagram based on service path.
Fig. 5 is the network topological diagram of example 1.
Fig. 6 is the service path schematic diagram for having inclusion relation.
Fig. 7 is the service path schematic diagram for having overlapping relation.
Fig. 8 is the network topological diagram of example 4.
Fig. 9 is BFM algorithm flow charts.
Figure 10 is execution time contrast schematic diagram under different supports.
Figure 11 is frequent item set number contrast schematic diagram under different supports.
Embodiment
Below in conjunction with accompanying drawing and preferred embodiment, the invention will be further described.It is emphasized that the description below
It is merely exemplary, the scope being not intended to be limiting of the invention and application.
Embodiment 1
To verify BFM efficiency of algorithm, certain research institute aircraft 2009-2011 Mishap Databases are chosen as experiment number
According to saving as the calculating that 8GB, CPU are Intel core i5-3337u 1.87GHz, operating system is 64 windows10 inside
On machine by Matlab programming realizations Apriori algorithm and BFM algorithms.
If min confidence is 0.3, table 9 is partial simulation result example:
Table 9 is correlation rule result
Wherein, A31 refers to " flying pipe abnormal signal ";A67 refers to " AFDX loads Signal Fail ", and corresponding service path is " winged pipe
System → AFDX1 → long-range 1 → inertial navigation ".Contrast troubleshooting record in failure procedure illustrate, excavate " A31 → A67 " is effective
It is and believable.
As shown in Figure 10, BFM efficiency of algorithm is higher than Apriori algorithm, and as support improves, advantage is first increasingly
Substantially, faded away after.Because as minimum support increases, gradually exceed the frequency of most projects in database, and
Scanning for the first time judges to have filtered out most piecemeals and item collection after BFM algorithm piecemeals, compared to Apriori algorithm, efficiency
Advantage can become apparent from;And as minimum support exceeds all items frequency, two algorithm an iterations judge all
Frequent item set, it is iterated after first having carried out piecemeal computing due to BFM, directly starts iteration compared to Apriori algorithm, algorithm is excellent
Gesture fades away.In engineering, the excavation support of onboard networks relevant fault is required typically within 20%, now BFM algorithms
Efficiency will much be higher by Apriori algorithm.
As shown in figure 11, frequent item set number caused by BFM algorithms will be less than Apriori algorithm, and as support carries
Height, frequent item set number gap are less and less.Because as minimum support increases, the cycle-index of search iteration is more next
Fewer, caused noise combination is also fewer and fewer in circulation, the advantage increasingly unobvious of BFM algorithms " denoising ".
To sum up, BFM algorithms can support onboard networks relevant fault to excavate, and have higher efficiency compared to traditional algorithm.
Claims (1)
1. a kind of dynamically associate failure method for digging based on service path and frequency matrix, it is characterised by:This method is included such as
Lower step:
Step 1: structure dynamically associates the data model of failure;
Step 101, clearly dynamically associate fault definition
The definition that onboard networks dynamically associate failure is:Exist in onboard networks operation using avionics business between each application of flow
Dynamically logic connection relation, cause under certain proof stress, when certain failure occurs, association triggering is new on current business path
Failure, and dynamically associating with service path, the characteristics of failure shows dynamic dependency and constantly propagated;
Service path, dynamic, relevance are defined respectively as:
Service path:Under certain proof stress, logic connecting relation is respectively applied in avionics operation flow;
Dynamic:Under same proof stress, failure generation is closely related with service path, and the dynamic of service path causes failure
Dynamic;
Relevance:Multiple failures may be caused by same failure in same test business, i.e., same proof stress and same business
There is relevance between the failure in path;
Step 102, determination dynamically associate the judgement feature of failure
Analyzed according to step 101, onboard networks failure has dynamically associates feature based on service path, its correlation rule and event
It is closely related to hinder attribute " service path " and " proof stress ", therefore provides following judgement feature:
Feature 1:Service path is non-intersect, two different failures of proof stress, it is impossible to be that direct correlation triggers;
Feature 2:Two failures of indirect association triggering, the failure pair that can be triggered by limited individual direct correlation, find its association
Relation;
Step 103, fault data pretreatment and model construction
Failure mainly includes the attributes such as phenomenon of the failure, faulty equipment, service path, stress condition, according to method for digging demand, structure
It is as follows to build fault message model F:
F={ S, X, N, Δ }
Wherein, S={ S1,S2,...Si... } and it is fault attribute " experiment ID ", X={ X1,X2,...Xi... } and it is fault attribute
" proof stress ", I={ A, B ... } are " phenomenon of the failure ", N={ N1,N2,...,Ni... } and it is fault attribute " service path ",
Ni=(nj,nk...) and be failure i service path, wherein nj,nk∈ n, n=1,2 ... } it is malfunctioning node equipment;Each parameter
Corresponding sequence is that the different value of corresponding field in raw data base is formed by dictionary ordering map;
In step 104, analysis mining method, the feasibility for judging feature reduction noise combination is utilized
The judgement feature provided according to step 102, can judge whether there may be dynamic associations between two failures in advance,
So as to avoid the combination of noise in the connection step of frequent-item;
Once to test as an affairs, trade type database-transaction set is formed, by demonstrating by L1Generate L2Process, card
The bright feature 1 using in step 102, the feasibility of digging efficiency can be greatly improved;If minimum support Supmin=2;
(1) trade type database-transaction set is scanned, generates 1 Frequent Set list L1
(2) carry out First Contact Connections step to calculate, generate 2 candidate C2
According to feature 1 is judged, failure combination is obtained, its service path has no to intersect, and proof stress is entirely different, it is impossible to be to close
Join failure, therefore be unlikely to be frequent item set;These combinations are entirely the noise information of connection step generation;
In search iteration, each step of algorithm all can a large amount of similar noise information of combination producing, then beta pruning eliminates it one by one
Middle major part, this greatly reduces digging efficiency and the degree of accuracy, and this partial noise is avoidable;
Therefore, in dynamically associating failure and excavating, using the judgement feature of service path, it is feasible to reduce noise combination;
Step 2: piecemeal Mining Strategy of the design based on very big service path;
Step 201, propose very big service path concept
Very big service path:Travel through in fault database after institute faulty proof stress and service path attribute, if service path R1~
RnWith inclusion relation, and proof stress is Xi, wherein RjComprising remaining (n-1) bar service path, then (Xi,Rj) it is a pole
Big service path;
Define service path vector R2Include R1Computing be R1@R2:if(X1=X2)&Len(R1)≤Len(R2),Then R1@R2;Wherein, Part (Ri, k) and represent vectorial RiThe continuous fragment that length is k forms
Gather (k≤Len (Ri)), Part (Ri,k)jRepresent in the set jth item (j=1,2 ..., Len (Ri)-k+1);
If R1=(1,2,3,4), Len (R1)=4, Part (R1,3)2=(2,3,4), Part (R1, 3)=(1,2,3), (2,3,
4) }, if R2=(2,3), R3=(3,4,5), and X1=X2=X3, then R2@R1Set up, R3@R1It is invalid;
Mishap Database presented hereinbefore can generate following 4 very big service paths after traversal:
{(X1,(1,2,3)),(X2,(2,3)),(X3,(5,6,7)),(X4,(5,6))}
Step 202, clearly very big service path characteristic
Very big service path characteristic:The failure of two direct correlation triggerings, maximum probability are betided under same greatly service path,
And as experiment carries out being intended to 100% with fault database increase, this probability;
Step 203, design piecemeal Mining Strategy
It is proposed following piecemeal Mining Strategy:
(1) after generating transaction set, ergodic data storehouse:
<mrow>
<mi>i</mi>
<mi>f</mi>
<mo>:</mo>
<mo>&ForAll;</mo>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<mi>n</mi>
<mo>,</mo>
<msub>
<mi>R</mi>
<mi>i</mi>
</msub>
<mo>@</mo>
<msub>
<mi>R</mi>
<mi>n</mi>
</msub>
<mo>,</mo>
<mi>t</mi>
<mi>h</mi>
<mi>e</mi>
<mi>n</mi>
<mo>:</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>X</mi>
<mi>n</mi>
</msub>
<mo>,</mo>
<msub>
<mi>R</mi>
<mi>n</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
Obtain very big service path set:
{(X1,R1),(X2,R2),...,(Xk,Rk)}
(2) very big service path set is utilized, database is divided into k blocks,
(3) each piecemeal dredge operation isolation is carried out;Affairs frequency calculates, and is only added up in same transaction set piecemeal;
Step 3: scanning strategy of the design based on frequency matrix;
Step 301, structure frequency matrix and characteristic vector
Characteristic vector:If certain piecemeal transaction set TiInclude m mutually different affairs t1,t2,...,tm, it is mutually different by n
Item I1,I2,...,InForm, wherein { ti},{IiSorted by dictionary, then claim n-dimensional vector (I1,I2,...,In) it is TiFeature
Vector, it is designated as Iti;Failure item, therefore I are represented with A, B, C etc.i=A, B, C...;
Frequency matrix:If certain piecemeal transaction set TiInclude m mutually different affairs t1,t2,...,tm, its characteristic vector is Iti
=(I1,I2,...,In), then existence anduniquess matrix Qi=(qjk)n×mSo that
<mrow>
<msub>
<mi>It</mi>
<mi>i</mi>
</msub>
<mo>&times;</mo>
<msub>
<mi>Q</mi>
<mi>i</mi>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>I</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>I</mi>
<mn>2</mn>
</msub>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<msub>
<mi>I</mi>
<mi>n</mi>
</msub>
<mo>)</mo>
</mrow>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>q</mi>
<mn>11</mn>
</msub>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mn>12</mn>
</msub>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mn>1</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>q</mi>
<mn>21</mn>
</msub>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mn>22</mn>
</msub>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mn>2</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mi>n</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mi>n</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mi>n</mi>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>&DoubleLeftRightArrow;</mo>
<msub>
<mi>T</mi>
<mi>i</mi>
</msub>
</mrow>
Then claim QiFor TiFrequency matrix;Wherein qjk∈ N represent TiIn k-th of affairs tkJth item IjFrequency;
Therefore, a transaction set piecemeal can be analyzed to a characteristic vector and a frequency matrix;After piecemeal, in search iteration
In, each piecemeal only needs run-down characteristic vector, and carries out first moment battle array computing;
Each piece of decomposition can be obtained transaction set, characteristic vector, frequency by transaction set piecemeal shown below, usage frequency matrix
Matrix;That is, transaction set T1Characteristic vector be (A, B, C), frequency matrix is
Transaction set T2Characteristic vector be (A, D, E), frequency matrix is
Transaction set T3Characteristic vector be (A, F, G), frequency matrix is
Wherein:
Observation matrix is it can be found that often row addition is Count (C1), second and third row is Count (C2), first row is Count
(C3);Therefore frequency matrix both remains frequency information, and and can embodies the related information between items, need not travel through number in a scan
According to storehouse, contrast computing is carried out one by one, only need to be according to matrix operation, you can complete scanning;
Step 302, analysis frequency matrix properties
Frequency matrix has following property:
(1) the i-th row of frequency matrix sum is equal to the frequency of characteristic vector i-th:
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>m</mi>
</munderover>
<msub>
<mi>q</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mi>C</mi>
<mi>o</mi>
<mi>u</mi>
<mi>n</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>I</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
</mrow>
(2) frequency matrix normalization is handled:
<mrow>
<msub>
<mi>Q</mi>
<mi>i</mi>
</msub>
<mo>=</mo>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>q</mi>
<mn>11</mn>
</msub>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mn>12</mn>
</msub>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mn>1</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>q</mi>
<mn>21</mn>
</msub>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mn>22</mn>
</msub>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mn>2</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mi>n</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mi>n</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mrow>
<mi>n</mi>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>&lambda;</mi>
<mn>1</mn>
</msub>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mn>11</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mn>21</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mi>n</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
<msub>
<mi>&lambda;</mi>
<mn>2</mn>
</msub>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mn>12</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mn>22</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mi>n</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
<mo>,</mo>
<msub>
<mi>&lambda;</mi>
<mi>m</mi>
</msub>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mn>1</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mn>2</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mi>n</mi>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>)</mo>
</mrow>
</mrow>
<mrow>
<msub>
<mi>q</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<msub>
<mi>&lambda;</mi>
<mi>j</mi>
</msub>
<msub>
<mi>e</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>,</mo>
<msub>
<mi>e</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
<msub>
<mi>&lambda;</mi>
<mi>j</mi>
</msub>
<mo>&Element;</mo>
<msub>
<mi>N</mi>
<mo>+</mo>
</msub>
</mrow>
Wherein, frequency matrix jth row coefficient lambdajEqual to the frequency that jth arranges corresponding affairs:
<mrow>
<msub>
<mi>&lambda;</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<mi>C</mi>
<mi>o</mi>
<mi>u</mi>
<mi>n</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>t</mi>
<mi>j</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>,</mo>
<msub>
<mi>t</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>I</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>I</mi>
<mn>2</mn>
</msub>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<msub>
<mi>I</mi>
<mi>n</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>&times;</mo>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mn>1</mn>
<mi>j</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mn>2</mn>
<mi>j</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mi>n</mi>
<mi>j</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Transaction set piecemeal T1:
<mrow>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>2</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>2</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>=</mo>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
<mn>1</mn>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>,</mo>
<mn>2</mn>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>)</mo>
</mrow>
</mrow>
Count (ABC)=1, Count (AB)=1, Count (BC)=2
As can be seen that usage frequency matrix and its computing feature, can significantly simplify the comparing calculation in scan round;
Step 303, analysis frequent item set property
In frequent item set mining, following property be present:
(1) nonvoid subset of any frequent item set is also that frequently, also right and wrong are frequently for the supersets of nonmatching grids;
Inference 1A:If the frequency of all items is below minimum support in certain transaction set piecemeal, do not have in the transaction set piecemeal
There is frequent item set, can directly delete the transaction set piecemeal;
Inference 1B:If a certain purpose frequency in certain transaction set piecemeal be present is less than minimum support, any of the project is included
The equal right and wrong of affairs frequently, directly can delete the project from transaction set piecemeal;
(2) if certain transaction length is less than k, the affairs do not support Lk+1;
Inference 2A:If the characteristic vector length of certain transaction set piecemeal is less than k+1, the maximum frequent itemsets of the transaction set piecemeal are
Lk, the transaction set piecemeal can be directly deleted in circulation afterwards;
Inference 2B:If being less than k+1 in the presence of certain transaction length in certain transaction set piecemeal, L is generated in the transaction set piecemealkAfterwards
The affairs can be directly deleted in circulation;
Step 304, design scanning strategy
The frequency matrix and frequent item set property provided based on step 301,302,303, this step is by designing five scanning plans
Omit, to greatly improve the sweep speed in excavating;
(1) scanning strategy 1~4 is designed:Frequency matrix and characteristic vector are compressed using minimum support, so as to compress transaction set point
Block,
Scanning strategy 1:According to inference 1A, if often row sum is respectively less than Sup to the frequency matrix of certain transaction set piecemealmin, then directly delete
Fall the transaction set piecemeal:
<mrow>
<mi>i</mi>
<mi>f</mi>
<mo>:</mo>
<munder>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
<mi>i</mi>
</munder>
<mrow>
<mo>(</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>m</mi>
</munderover>
<msub>
<mi>q</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mo><</mo>
<msub>
<mi>Sup</mi>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>,</mo>
<msub>
<mi>q</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>&Element;</mo>
<msub>
<mi>Q</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<mi>t</mi>
<mi>h</mi>
<mi>e</mi>
<mi>n</mi>
<mo>:</mo>
<mi>d</mi>
<mi>e</mi>
<mi>l</mi>
<mi>e</mi>
<mi>t</mi>
<mi>e</mi>
<mi> </mi>
<msub>
<mi>T</mi>
<mi>k</mi>
</msub>
</mrow>
Scanning strategy 2:According to inference 1B, if certain transaction set piecemeal has a frequency to be less than minimum support, i.e. frequency matrix
Row sum is less than Supmin, then the blocking characteristic is vectorial deletes this, and frequency matrix deletes the row, so as to compress the transaction set point
Block:
<mrow>
<mi>i</mi>
<mi>f</mi>
<mo>:</mo>
<mo>&Exists;</mo>
<mi>i</mi>
<mo>&le;</mo>
<mi>n</mi>
<mo>,</mo>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>m</mi>
</munderover>
<msub>
<mi>q</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo><</mo>
<msub>
<mi>Sup</mi>
<mi>min</mi>
</msub>
<mo>,</mo>
<mi>t</mi>
<mi>h</mi>
<mi>e</mi>
<mi>n</mi>
<mo>:</mo>
<mi>d</mi>
<mi>e</mi>
<mi>l</mi>
<mi>e</mi>
<mi>t</mi>
<mi>e</mi>
<mi> </mi>
<msub>
<mi>I</mi>
<mi>i</mi>
</msub>
<mo>,</mo>
<msub>
<mi>q</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<msub>
<mo>|</mo>
<mrow>
<mi>j</mi>
<mo>&le;</mo>
<mi>m</mi>
</mrow>
</msub>
</mrow>
Scanning strategy 3:According to inference 2A, if the characteristic vector length of certain transaction set piecemeal is less than k+1, in generation LkAfterwards, directly
Delete the transaction set piecemeal:
Generate Ck+1:
Scanning strategy 4:According to inference 2B, if certain row sum after certain transaction set piecemeal frequency matrix decomposition is less than k+1, in life
Into LkAfterwards, the row are directly deleted:
Generate Ck+1:
(2) scanning strategy 5 is designed:Traditional algorithm is calculating Count (Ck) when need scan round transaction set, after introducing frequency matrix,
Using Matrix Properties, contrast scans work is skipped, directly obtains result with product;
Defined function jo (x):
<mrow>
<mi>j</mi>
<mi>o</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mn>0</mn>
<mo>,</mo>
<mi>i</mi>
<mi>f</mi>
<mo>:</mo>
<mi>x</mi>
<mo>=</mo>
<mn>2</mn>
<mi>k</mi>
<mo>+</mo>
<mn>1</mn>
<mo>,</mo>
<mi>k</mi>
<mo>=</mo>
<mn>0</mn>
<mo>,</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>x</mi>
<mo>,</mo>
<mi>i</mi>
<mi>f</mi>
<mo>:</mo>
<mi>x</mi>
<mo>=</mo>
<mn>2</mn>
<mi>k</mi>
<mo>,</mo>
<mi>k</mi>
<mo>=</mo>
<mn>0</mn>
<mo>,</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mo>...</mo>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Scanning strategy 5:The frequency matrix properties provided according to step 302, calculating the Count (s of certain transaction set piecemeali),si∈
CkWhen, by siIt is frequency vector by eigendecomposition, the product of frequency vector and frequency matrix can obtain Count (si), without
Circulation contrast:
<mrow>
<mo>&ForAll;</mo>
<msub>
<mi>s</mi>
<mi>i</mi>
</msub>
<mo>&Element;</mo>
<msub>
<mi>C</mi>
<mi>k</mi>
</msub>
<mo>,</mo>
<msub>
<mi>s</mi>
<mi>i</mi>
</msub>
<mo>=</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>I</mi>
<mn>1</mn>
</msub>
<mo>,</mo>
<msub>
<mi>I</mi>
<mn>2</mn>
</msub>
<mo>,</mo>
<mo>...</mo>
<mo>,</mo>
<msub>
<mi>I</mi>
<mi>n</mi>
</msub>
<mo>)</mo>
</mrow>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>C</mi>
<mi>o</mi>
<mi>u</mi>
<mi>n</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<msub>
<mi>s</mi>
<mi>i</mi>
</msub>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mi>j</mi>
<mi>o</mi>
<mrow>
<mo>(</mo>
<mrow>
<msup>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mi>T</mi>
</msup>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mn>11</mn>
</msub>
</mtd>
<mtd>
<msub>
<mi>e</mi>
<mn>12</mn>
</msub>
</mtd>
<mtd>
<mo>...</mo>
</mtd>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mn>1</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mn>21</mn>
</msub>
</mtd>
<mtd>
<msub>
<mi>q</mi>
<mn>22</mn>
</msub>
</mtd>
<mtd>
<mo>...</mo>
</mtd>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mn>2</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mi>n</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mi>n</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
<mtd>
<mo>...</mo>
</mtd>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mi>n</mi>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
<mo>)</mo>
</mrow>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>&lambda;</mi>
<mn>1</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>&lambda;</mi>
<mn>2</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>&lambda;</mi>
<mi>n</mi>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<msub>
<mi>&lambda;</mi>
<mn>1</mn>
</msub>
<mi>j</mi>
<mi>o</mi>
<mrow>
<mo>(</mo>
<mrow>
<msup>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mi>T</mi>
</msup>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mn>11</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mn>21</mn>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mi>n</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
<mo>)</mo>
</mrow>
<mo>+</mo>
<mo>+</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<msub>
<mi>&lambda;</mi>
<mi>m</mi>
</msub>
<mi>j</mi>
<mi>o</mi>
<mrow>
<mo>(</mo>
<mrow>
<msup>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mn>1</mn>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mn>2</mn>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>g</mi>
<mrow>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
<mi>T</mi>
</msup>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mn>1</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mn>2</mn>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow></mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<msub>
<mi>e</mi>
<mrow>
<mi>n</mi>
<mi>m</mi>
</mrow>
</msub>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
Transaction set piecemeal T1, calculate Count (AC), AC ∈ C2:
<mrow>
<mi>A</mi>
<mi>C</mi>
<mo>&DoubleLeftRightArrow;</mo>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mo>,</mo>
<mi>B</mi>
<mo>,</mo>
<mi>C</mi>
<mo>)</mo>
</mrow>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mi>C</mi>
<mi>o</mi>
<mi>u</mi>
<mi>n</mi>
<mi>t</mi>
<mrow>
<mo>(</mo>
<mi>A</mi>
<mi>C</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mi>j</mi>
<mi>o</mi>
<mrow>
<mo>(</mo>
<msup>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mi>T</mi>
</msup>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
<mtd>
<mn>0</mn>
</mtd>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
<mo>)</mo>
</mrow>
<mfenced open = "(" close = ")">
<mtable>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>1</mn>
</mtd>
</mtr>
<mtr>
<mtd>
<mn>2</mn>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mrow>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>&times;</mo>
<mi>j</mi>
<mi>o</mi>
<mrow>
<mo>(</mo>
<mn>2</mn>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mn>1</mn>
<mo>&times;</mo>
<mi>j</mi>
<mi>o</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
<mo>,</mo>
<mn>2</mn>
<mo>&times;</mo>
<mi>j</mi>
<mi>o</mi>
<mrow>
<mo>(</mo>
<mn>1</mn>
<mo>)</mo>
</mrow>
</mrow>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mo>=</mo>
<mfrac>
<mn>1</mn>
<mn>2</mn>
</mfrac>
<mrow>
<mo>(</mo>
<mrow>
<mn>1</mn>
<mo>&times;</mo>
<mn>2</mn>
<mo>,</mo>
<mn>1</mn>
<mo>&times;</mo>
<mn>0</mn>
<mo>,</mo>
<mn>2</mn>
<mo>&times;</mo>
<mn>0</mn>
</mrow>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mn>1</mn>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
Using scanning strategy 5, product calculation is directly carried out in scan round, without scanning the frequency of Frequent Set one by one, can be had
Effect improves digging efficiency;
Step 4: BFM algorithm global designs
Step 401, algorithm flow description
Algorithm ergodic data storehouse first, very big service path is generated, and utilize very big service path by database piecemeal;According to frequency
Matrix definition is spent, the failure item collection of each piecemeal is decomposed into characteristic vector and frequency matrix;Then carried out using scanning strategy 1,2
Simplification before search, the piecemeal of the condition of being unsatisfactory for and, is deleted by scanning strategy 1, by scanning plan according to the row of matrix in block form
2 are omited by matrix in block form depression of order;
1 Frequent Set of each piecemeal is firstly generated, starts the cycle over search iteration;Directly skipped using scanning strategy 3 and be unsatisfactory for bar
The Circulant Block of part, the further depression of order of frequency matrix is optimized using scanning strategy 4;Piecemeal after optimization is attached step and cut
Branch step generation candidate, calculates candidate frequency, and then generate Frequent Set by scanning strategy 5;Circulate in and work as in piecemeal
Preceding frequent item set terminates when being empty set, circulates in after the completion of all piecemeals and terminates between piecemeal, finally returns that all frequent item sets;Figure
9 show BFM algorithm flows;
Step 402, algorithm pseudo code
Input:Transaction database D, minimum support Supmin
Output:All frequent item set L.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710710411.3A CN107579844A (en) | 2017-08-18 | 2017-08-18 | It is a kind of that failure method for digging is dynamically associated based on service path and frequency matrix |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710710411.3A CN107579844A (en) | 2017-08-18 | 2017-08-18 | It is a kind of that failure method for digging is dynamically associated based on service path and frequency matrix |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107579844A true CN107579844A (en) | 2018-01-12 |
Family
ID=61034717
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710710411.3A Pending CN107579844A (en) | 2017-08-18 | 2017-08-18 | It is a kind of that failure method for digging is dynamically associated based on service path and frequency matrix |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107579844A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108664548A (en) * | 2018-03-27 | 2018-10-16 | 中国科学院软件研究所 | A kind of network under degenerative conditions accesses behavioural characteristic team innovation method for digging and system |
CN111666519A (en) * | 2020-05-13 | 2020-09-15 | 中国科学院软件研究所 | Dynamic mining method and system for network access behavior feature group under enhanced condition |
CN112948187A (en) * | 2021-01-04 | 2021-06-11 | 杭州恒朴电子科技有限公司 | Multi-index fluctuation correlation analysis method for volume production link |
CN114513449A (en) * | 2022-02-23 | 2022-05-17 | 北京慧橙信息科技有限公司 | Method and system for optimizing routing in domain |
CN114513449B (en) * | 2022-02-23 | 2024-04-26 | 北京慧橙信息科技有限公司 | Intra-domain routing optimization method and system |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013038017A1 (en) * | 2011-09-16 | 2013-03-21 | Olaf Rudolph | Method for identifying one or more simultaneously occurring partial discharge sources |
CN106324429A (en) * | 2016-08-02 | 2017-01-11 | 天津大学 | RS-IA data mining-based power distribution network fault locating method |
-
2017
- 2017-08-18 CN CN201710710411.3A patent/CN107579844A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2013038017A1 (en) * | 2011-09-16 | 2013-03-21 | Olaf Rudolph | Method for identifying one or more simultaneously occurring partial discharge sources |
CN106324429A (en) * | 2016-08-02 | 2017-01-11 | 天津大学 | RS-IA data mining-based power distribution network fault locating method |
Non-Patent Citations (1)
Title |
---|
胡波等: "基于业务路径和频度矩阵的关联规则挖掘算法", 《计算机科学》 * |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108664548A (en) * | 2018-03-27 | 2018-10-16 | 中国科学院软件研究所 | A kind of network under degenerative conditions accesses behavioural characteristic team innovation method for digging and system |
CN108664548B (en) * | 2018-03-27 | 2021-08-03 | 中国科学院软件研究所 | Network access behavior characteristic group dynamic mining method and system under degradation condition |
CN111666519A (en) * | 2020-05-13 | 2020-09-15 | 中国科学院软件研究所 | Dynamic mining method and system for network access behavior feature group under enhanced condition |
CN112948187A (en) * | 2021-01-04 | 2021-06-11 | 杭州恒朴电子科技有限公司 | Multi-index fluctuation correlation analysis method for volume production link |
CN114513449A (en) * | 2022-02-23 | 2022-05-17 | 北京慧橙信息科技有限公司 | Method and system for optimizing routing in domain |
CN114513449B (en) * | 2022-02-23 | 2024-04-26 | 北京慧橙信息科技有限公司 | Intra-domain routing optimization method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Moreira et al. | Polynomial time verification of decentralized diagnosability of discrete event systems | |
DE69838835T2 (en) | Method for testing and displaying hardware by decomposition and partitioning | |
Georgiadis et al. | Continuous outlier detection in data streams: an extensible framework and state-of-the-art algorithms | |
CN109150619B (en) | Fault diagnosis method and system based on network flow data | |
CN110740054B (en) | Data center virtualization network fault diagnosis method based on reinforcement learning | |
CN112559237B (en) | Operation and maintenance system troubleshooting method and device, server and storage medium | |
CN107579844A (en) | It is a kind of that failure method for digging is dynamically associated based on service path and frequency matrix | |
CN109936479A (en) | Control plane failure diagnostic system and its implementation based on Differential Detection | |
Dallal et al. | On most permissive observers in dynamic sensor activation problems | |
CN103793589B (en) | High-speed train fault handling method | |
Lin et al. | Facgraph: Frequent anomaly correlation graph mining for root cause diagnose in micro-service architecture | |
CN103914386B (en) | Software defect positioning method based on input parameter characteristic spectrum | |
Mo et al. | Network simplification and k-terminal reliability evaluation of sensor-cloud systems | |
CN115544519A (en) | Method for carrying out security association analysis on threat information of metering automation system | |
CN115118621A (en) | Micro-service performance diagnosis method and system based on dependency graph | |
Levi et al. | Constructing near spanning trees with few local inspections | |
Bonacina et al. | On Variable-inactivity and Polynomial $\mathcal T $-Satisfiability Procedures | |
Luo et al. | A new algorithm for testing diagnosability of fuzzy discrete event systems | |
He et al. | A distributed network alarm correlation analysis mechanism for heterogeneous networks | |
Hagiya et al. | Analysis of synchronous and asynchronous cellular automata using abstraction by temporal logic | |
CN108319517B (en) | Fault diagnosis method and device | |
CN110995747A (en) | Distributed storage security analysis method | |
CN111240881A (en) | Web service combination abnormity diagnosis method based on Bayesian network | |
Aribi et al. | Multiple fault localization using constraint programming and pattern mining | |
Takan et al. | Mutant selection by using Fourier expansion |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180112 |
|
RJ01 | Rejection of invention patent application after publication |