CN101887531B - Flight data knowledge acquisition system and acquisition method thereof - Google Patents
Flight data knowledge acquisition system and acquisition method thereof Download PDFInfo
- Publication number
- CN101887531B CN101887531B CN2010102086422A CN201010208642A CN101887531B CN 101887531 B CN101887531 B CN 101887531B CN 2010102086422 A CN2010102086422 A CN 2010102086422A CN 201010208642 A CN201010208642 A CN 201010208642A CN 101887531 B CN101887531 B CN 101887531B
- Authority
- CN
- China
- Prior art keywords
- rule
- module
- flying quality
- data
- knowledge
- 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
Images
Abstract
The invention provides a flight data knowledge acquisition system and an acquisition method thereof. The flight data knowledge acquisition system mainly comprises a flight data acquisition module, a flight database module, a mode selection module, a data preparation module, a data display module, a task execution module, a rule output module, a rule processing module, a knowledge acquisition module and a knowledge base module; and the system inputs and processes flight data, extracts useful data forming rules through rule acquisition operation, and supplies the useful data forming rules serving as knowledge to a user. The flight data knowledge acquisition method mainly comprises the following steps of: processing the input flight data, digging association rules by adopting an Apriori algorithm or a Mulit-Rule method, and supplying the extracted rules serving as knowledge to the user. The system and the method reduce the maintenance difficulty of a knowledge base by eliminating redundancy rules and merging the rules, implement an automatic knowledge acquisition process by using an association rule method, are suitable for processing multi-dimensional and single-dimensional data, and reduce manual input and dependency on expert experiences.
Description
Technical field
The invention belongs to the flight data process field, be specifically related to a kind of flying quality knowledge acquisition system and acquisition methods thereof.
Background technology
Flying quality is a series of and aeroplane performance flight parameter relevant with state that aircraft is noted by flight data recording equipment from fly to the landing process, and flare maneuver identification, equipment performance trend analysis and aircraft accident are identified to have important effect.
Utilize flying quality to carry out fault diagnosis and trend prediction at present, realize aircraft to look the feelings maintenance be an important research direction, the realization of these two functions is to be basis with powerful knowledge base, knowledge base is the important foundation of structure flying quality expert system.Therefore, the problem of obtaining of knowledge becomes the bottleneck problem of flying quality expert system in the knowledge base.
By the automaticity of knowledge acquisition, can knowledge acquisition be divided into semi-automatic knowledge acquisition and automatic knowledge acquisition dual mode.
In automanual knowledge acquisition mode, knowledge acquisition is carried out in two steps: the knowledge engineer at first obtains knowledge from corresponding knowledge source, through knowledge edition software knowledge is input in the knowledge base then.This mode mainly is artificial input, depends on very much expertise simultaneously, if therefore can find a kind of method to realize obtaining automatically of knowledge, with improving the intelligent of expert system to a great extent, more can give play to expert's advantage of expert system.
Automatically knowledge acquisition is a kind of desirable knowledge acquisition mode; The finger system self has the ability of the knowledge obtained; Not only can be directly from the relevant rudimentary knowledge of knowledge source " study ", and can also from the operation practice of system self, sum up, summarize new knowledge, constantly improve its knowledge storehouse.Although still can not realize obtaining fully automatically of knowledge at present, for a long time, the knowledge acquisition research work concentrates on the automaticity that improves the knowledge acquisition process, alleviates knowledge engineer's burden, and oneself is through making significant progress.
Method about automatic knowledge acquisition mainly contains based on neural network, immune algorithm, genetic algorithm and rough set method at present.Neural network possesses the robustness and the study associative ability of large-scale parallel distribution process ability, Nonlinear Processing ability, height, but the problem that mainly faces is the generalization ability of confirming the complicacy of the structure of network, reduction e-learning and improving network.Immune algorithm is the algorithm of the biological rule of simulation, and it attempts to seek optimum solution, so running time of algorithm is long, efficient is not high, and in use effect can be not fine for the fuzzy information systems of a large amount of connection attributes.Genetic algorithm must be chosen suitable crossover probability and variation probability in solution procedure, otherwise can cause algorithm instability or convergence problem; Rough set theory is as a kind of mathematical tool of analyzing and handling out of true, inconsistent, imperfect information and knowledge, and it can derive the diagnostic rule with directive significance from lot of data, analyzes the useful rule information of finding.
Find after deliberation; Association rules method in the data mining has powerful ability in knowledge acquisition; It can excavate the ND contact between the data, is shown as rule to this contacts list, forms knowledge; Form knowledge base to a large amount of knowledge at last, for expert system is carried out fault diagnosis and trend prediction provides powerful knowledge foundation.Correlation rule is through analyzing frequent (recurrent incident) between given data centralization each item; Find significant knowledge; This method can be handled various types of data; And can handle the mass data of one-dimensional and hyperspace, be fit to flying quality is handled.Tentatively carried out both at home and abroad at present and utilized correlation rule to carry out the Research on Knowledge Acquirement Methods of flight data, but all be the processing of carrying out to the one-dimensional data, and do not formed corresponding system.
Summary of the invention
The objective of the invention is to propose a kind of flying quality knowledge acquisition system and method thereof; This system can not rely on expertise simultaneously according to the characteristics of flying quality, need not consider the characteristics of the inconsistency of data type; Can logarithm value type data, interval type data, character type data handle; Analyze useful information implicit in the flying quality, form knowledge, for carrying out fault diagnosis and trend prediction provides knowledge base.The present invention is fit to solve the knowledge acquisition problem of one-dimensional and multidimensional flying quality; Also can be applied to simultaneously the knowledge acquisition platform of other type; According to the characteristics of association rules method, this system can provide processing power fast to the flying quality collection of big data quantity in addition.
A kind of flying quality knowledge acquisition system of the present invention; Participate in by the user, mainly comprise flying quality acquisition module, flying quality library module, mode selection module, data preparation module, data disaply moudle, task execution module, regular output module, rule treatments module, knowledge acquisition module and base module.
The flying quality acquisition module obtains flying quality, and the flying quality that obtains is sent to the flying quality library module; The flying quality library module stores the flying quality that obtains in the database table into; Data preparation module is carried out data processing to the flying quality that writes down in the database table in the flying quality library module; Make the data after the processing can be used in association rule mining, and the data after will handling offer data disaply moudle and task execution module as data source; Data preparation module adopts character marking to handle to the one-dimensional flying quality; To the multidimensional flying quality; Data hierarchy is carried out in numerical value interval according to flying quality; Adopt the character label that the data interval of each layer is identified, before carrying out data hierarchy, the numerical exception of flying quality is pressed the front and back Mean Method and handle; Data disaply moudle is shown to the user with the flying quality that receives;
The user selects the knowledge acquisition pattern through mode selection module: the knowledge acquisition of multidimensional flying quality is carried out in knowledge acquisition or the selection selecting to carry out the one-dimensional flying quality, and mode selection module notifies the knowledge acquisition pattern that chooses to task execution module;
Task execution module is carried out rule to the flying quality after the processing of data preparation module and is obtained work according to the knowledge acquisition pattern, obtains the useful information that lies in the flying quality, and formation rule, offers regular output module;
The rule that the rule output module will obtain is shown to the user, and the rule that will obtain offers the rule treatments module; The rule treatments module is handled the rule that regular output module provides, and mainly comprises the processing of redundancy rule and carries out the merging of rule, and should rule offer knowledge acquisition module as knowledge;
Knowledge acquisition module is shown to the user with the knowledge that receives, and simultaneously knowledge is offered base module; The knowledge that base module provides knowledge acquisition module is according to storing in the database.
Said task execution module adopts Apriori algorithm process one-dimensional flying quality; Adopt the Mulit-Rule method to handle the multidimensional flying quality; Wherein, the Multi-Rule method is to the present invention is based on a kind of method of carrying out the multidimensional association rule mining that the Apriori algorithm proposes.
The step that task execution module adopts the Apriori algorithm to carry out association rule mining is: step a, parameter setting: the user rule of thumb sets the minimum support threshold value and the minimal confidence threshold of rule; Step b, 1 collection of generation candidate: all items in the flying quality knowledge acquisition system scanning flying quality library module; Number of times to every appearance is counted; According to the minimum support threshold value that is provided with among the step a; With all item counts 1 collection of item formation candidate greater than the minimum support threshold value, described is the flying quality after handling through data preparation module; Step c, generate frequent item set: it is the process of iteration progressively that frequent item set generates; Through search one by one to each item relevance in the flying quality library module; Accomplish final iterative process, form frequent item set, the minimal confidence threshold that the item count of frequent item set is provided with greater than the user; Steps d, rule output: the frequent item set that step c produces is exactly the rule that finally obtains through association rule mining; Step e, rule reduction and preservation: the rule of output is the character mark form, is reduced into its represented data object to sign, the rule after the preservation reduction.
The concrete steps that task execution module adopts the Mulit-Rule method to carry out association rule mining are: steps A, tables of data generate: the flying quality after data preparation module is handled; The hierarchical information and the label that comprise every kind of flying quality are deposited in the database again; And the formation tables of data, this tables of data of formation mainly comprises the title of the affiliated system of flying quality, the label of flying quality, the title of flying quality, the type of flying quality, the identifier and the corresponding interval range of layering; Step B, parameter setting: minimum support threshold value and minimal confidence threshold are set; Step C, generate candidate's 1 predicate collection: all dimension in the scan database; Number of times to every dimension occurs is counted; All are satisfied the set formation candidate 1 predicate collection of dimension counting greater than the minimum support threshold value, and described dimension is the flying quality that utilizes the formalization denotational description that obtains after handling through data preparation module; Step D, generation candidate K predicate collection: through search one by one to each predicate relevance in the database; Accomplish final iterative process, form candidate K predicate collection, wherein; K represents the maximal value of multidimensional predicate, the minimal confidence threshold that the dimension counting of candidate K predicate collection all is provided with greater than the user; Step e, rule output: the candidate K predicate collection that generates among the step D is exactly the rule that obtains through association rule mining; Step F, the reduction of the rule of step e output and preserve: be reduced into its represented data object to character mark, the rule after will reducing is simultaneously preserved.
A kind of flying quality knowledge acquisition method based on the flying quality knowledge acquisition system of the present invention may further comprise the steps:
Step 1, obtain flying quality: the flying quality acquisition module obtains flying quality through the data reading software of flight parameter registering instrument; Perhaps obtain flying quality through flight data file, the flying quality acquisition module stores the flying quality that obtains in the flying quality library module into;
Step 4, association rule mining: the knowledge acquisition pattern that task execution module is selected according to the step 2 user, to the data after the step 3 processing, obtain the useful information that lies in the flying quality through association rule mining, and with these information formation rules;
Step 7, knowledge acquisition: the knowledge of knowledge acquisition module receiving step six, arrive base module with this knowledge store, and be shown to the user.
Advantage of the present invention and good effect are:
(1) not only is applicable to the knowledge acquisition of one-dimensional flying quality, has advantage aspect the multidimensional data knowledge acquisition equally;
When (2) carrying out knowledge acquisition, the data of polytype (like numeric type, interval type) can be handled simultaneously, the inconsistent problem of data type need not be considered;
(3) utilize association rules method to carry out the work of obtaining of knowledge, this method provides processing power fast to the flying quality collection of big data quantity;
(4) can handle the rule that obtains, eliminate redundancy rule and the rule that comprises is each other merged, the MAINTENANCE OF KNOWLEDGE BASE difficulty is reduced;
(5) utilize the knowledge acquisition system of association rules method development, can realize the process that knowledge is obtained automatically, reduce artificial input and the dependence of expertise.
Description of drawings
Fig. 1 is the flying quality knowledge acquisition system structural drawing that the present invention is based on association rules method;
Fig. 2 is the flying quality knowledge acquisition basic flow sheet that the present invention is based on association rules method;
Fig. 3 is the process flow diagram that the present invention handles one-dimensional flying quality knowledge acquisition method;
Fig. 4 is the process flow diagram that the present invention handles multidimensional flying quality knowledge acquisition method;
Fig. 5 is the process flow diagram of rule treatments resume module process of the present invention.
Among the figure: 1. flying quality acquisition module 2. flying quality library modules 3. mode selection modules 4. data preparation module 5. data disaply moudles 6. task execution module 7. regular output module 8. rule treatments modules 9. knowledge acquisition modules 10. base modules 11. users
Embodiment
To combine accompanying drawing and embodiment that the present invention is done further detailed description below.
The structure of a kind of flying quality knowledge acquisition system based on correlation rule of the present invention is as shown in Figure 1, mainly comprises flying quality acquisition module 1, flying quality library module 2, mode selection module 3, data preparation module 4, data disaply moudle 5, task execution module 6, regular output module 7, rule treatments module 8, knowledge acquisition module 9 and base module 10.
Wherein, Flying quality acquisition module 1 obtains flying quality and passes to flying quality library module 2 and be kept in the database table; Data preparation module 4 handles the flying quality of preserving in the flying quality library module 2 so that this flying quality can be used in association rule mining, and data preparation module 4 sends the data of handling well to data disaply moudle 5 and task execution module 6.Data disaply moudle 5 is given user 11 with data presentation.User 11 is through the pattern of mode selection module 3 selection knowledge acquisition, and mode selection module 3 is with the knowledge acquisition pattern notice task execution module of selecting 6.Task execution module 6 is carried out association rule mining according to the knowledge acquisition pattern to the flying quality after handling through data preparation module 4, and the Useful Information formation rule is exported to regular output module 7.The rule that rule output module 7 will receive sends rule treatments module 8 and user 11 to.User 11 can check the rule of regular output module 7 outputs.The rule that 8 pairs of regular output modules 7 of rule treatments module provide is carried out the redundancy rule processing and is merged rule treatments, and should rule offer knowledge acquisition module 9 as knowledge.Knowledge acquisition module 9 is shown to user 11 with the knowledge that receives, and simultaneously knowledge is offered base module 10.Base module 10 stores the knowledge that obtains in the corresponding database into according to the database format that designs.
Flying quality acquisition module 1 is the interface of knowledge acquisition system of the present invention and present various flight parameter registering instrument data-switching.Flying quality acquisition module 1 reads flying quality, and flying quality obtains through the data reading software of flight parameter registering instrument, perhaps directly obtains through flight data file, and the flying quality that reads is sent to flying quality library module 2.
Flying quality library module 2 will store into the corresponding database table from the flying quality that flying quality acquisition module 1 obtains; The essential information of this database table storage flying quality and the concrete numerical information of flying quality; The essential information of flying quality comprises title, the label of flying quality, the title of flying quality and the type of flying quality of system under the flying quality, and the concrete numerical information of flying quality comprises like the writing time of flying quality and corresponding data values.All flying qualities can adopt unified data storehouse list structure.
In 4 pairs of flying quality library modules 2 of data preparation module in the database table data recorded carry out data processing; Make the data after the processing become the data that association rule mining can be used, the data after handling are offered data disaply moudle 5 and task execution module 6 as data source.The original flying quality of record is not complete data in the flying quality library module 2; Generally all comprise noise, exceptional value or inconsistent data; Therefore data preparation module 4 need be handled the original flying quality of record in the flying quality library module 2 according to the characteristics of original flying quality accordingly, as eliminating noise, rejecting abnormalities value, carrying out type conversion work.Concrete processing mode is: to the one-dimensional flying quality mainly is that data type property according to flying quality carries out character marking; To the multidimensional flying quality, the numerical exception of flying quality is pressed the front and back Mean Method handle, carry out data hierarchy according to the numerical value interval of flying quality again, adopt corresponding character label that the data interval of each layer is identified.Described front and back Mean Method is exactly to get the previous point and a back point of exceptional value institute corresponding point, then these two some value corresponding is sued for peace and is averaged.
Data disaply moudle 5 receives the flying quality that data preparation module 4 is handled well, and it is shown to user 11, is convenient to 11 pairs of data of user and carries out observation analysis.
The knowledge acquisition pattern that task execution module 6 is selected according to mode selection module 3; And utilize flying quality after the processing that data preparation module 4 provides to carry out rule and obtain work; Obtain the useful information that lies in the flying quality, and formation rule, regular output module 7 offered.Under the knowledge acquisition pattern of one-dimensional flying quality, task execution module 6 adopts the Apriori algorithm to carry out association rule mining; Under the knowledge acquisition pattern of multidimensional flying quality, task execution module 6 adopts the Mulit-Rule method to carry out association rule mining.Said Apriori algorithm is a kind of alternative manner of search successively, and K item collection is used for search (K+1) collection in the Apriori algorithm.At first find out the set of frequent 1 collection, this set note is made L1, and L1 is used to look for the set L2 of frequent 2 collection, and L2 is used to look for the set L3 of frequent 3 collection, and recursion like this up to can not finding K item collection, thereby is accomplished the excacation of rule.Wherein, K is the integer greater than 0.
Described Mulit-Rule method is to the present invention is based on a kind of method of carrying out the multidimensional association rule mining that the Apriori algorithm proposes, and also is a kind of alternative manner of search successively.K item collection is used for search (K+1) collection in the Mulit-Rule method.At first find out the set of frequent 1 collection, this set note is made L1, and L1 is used to look for the set L2 of frequent 2 collection, and L2 is used to look for the set L3 of frequent 3 collection, and recursion like this up to can not finding K item collection, thereby is accomplished the excacation of rule.
Rule output module 7 will be shown to user 11 from the rule that task execution module 6 is obtained, be convenient to 11 pairs of rules that obtain of user and watch, and the rule that will obtain simultaneously offers rule treatments module 8.
The rule that 8 pairs of regular output modules 7 of rule treatments module provide is handled, and mainly comprises the processing of redundancy rule and carries out the merging work of rule.Rule through after 8 processing of rule treatments module is the simplest, is independently a kind of describing mode simultaneously.Specifying information according to the related flying quality of the prerequisite of each bar rule and conclusion; Every rule is associated with the affiliated system of corresponding flight parameter; The go forward side by side semantic description of line discipline is described corresponding rule as a kind of knowledge, the actual knowledge representation mode that meets promptly is provided; And this knowledge offered knowledge acquisition module 9; This knowledge is obtained after handling through correlation rule automatically, in corresponding expert system application process, can be used as a kind of expertise and instructs fault diagnosis and trend prediction work.
Knowledge acquisition module 9 receives the knowledge that rule treatments module 8 provides, and this knowledge is shown to user 11, and user 11 can check knowledge.Knowledge acquisition module 9 offers base module 10 with knowledge simultaneously.
The knowledge store that base module 10 provides knowledge acquisition module 9 can adopt the unified data structure to describe to the knowledge that provides in corresponding database.This data structure comprises the label of knowledge, the prerequisite of knowledge, the conclusion of knowledge and the reliability information of knowledge.These knowledge are that the expert system of utilizing flying quality to carry out fault diagnosis and trend prediction provides expertise.
The present invention is based on a kind of flying quality knowledge acquisition method of above-mentioned knowledge acquisition system based on correlation rule, as shown in Figure 2, accomplish by following step:
Step 1: obtain flying quality;
It is the prerequisite of carrying out knowledge acquisition that flying quality obtains, and this part work is accomplished by flying quality acquisition module 1.Said flying quality comprises the essential information of flying quality, as: title, label, title and the type of flying quality and the concrete numerical information of flying quality of system under the flying quality, as: the writing time of flying quality and corresponding data values.Flying quality acquisition module 1 can obtain flying quality through the data reading software of flight parameter registering instrument, also can be directly flight data file through corresponding format obtain flying quality.Flying quality acquisition module 1 is stored in the flying quality library module 2 after obtaining these flying qualities.
Step 2: select the knowledge acquisition pattern;
The user is according to the demand of reality, selects the pattern that need obtain through mode selection module 3.User 11 can select to carry out the knowledge acquisition of one-dimensional flying quality, also can select to carry out the knowledge acquisition of multidimensional flying quality.
Step 3: data are prepared;
It mainly is that the data that store in the step 1 in the flying quality library module 2 are handled that data are prepared, and this part work is accomplished by data preparation module 4.Data prepare mainly to comprise noise, type conversion and the data hierarchy work eliminated.Handle through data preparation module 4, make the data in the flying quality library module 2 become the data that association rule mining can be used.According to the difference of the knowledge acquisition pattern of confirming in the step 2, the detailed process that data are prepared is also different.
When selecting the knowledge acquisition of one-dimensional flying quality; It mainly is that data type property according to flying quality carries out character marking that its data are prepared; Convert flying quality into computing machine and be convenient to the forms of treatment symbol, represent high pressure rotor rotating speed, I3 to represent Fuel Oil Remaining, I4 to represent barometer altitude as represent rotational speed of lower pressure turbine rotor, I2 with I 1.
When elected majority is tieed up the knowledge acquisition of flying quality; Mean Method was handled before and after its data set-up procedure was at first pressed the numerical exception of flying quality; Carry out data hierarchy according to the numerical value interval of flying quality on this basis then; Adopt corresponding character label that the data interval of each layer is identified at last, be convenient to carry out post-processed.As concerning barometer altitude, its numerical value interval is (1-30000), behind data hierarchy, with A1 representative (0-500), A2 representative (500-5000), A3 representative (5000-10000) A4 representative (10000-30000).Said front and back Mean Method is a previous point and a back point of getting exceptional value institute corresponding point, then these two some value corresponding is sued for peace and is averaged.
The data set-up procedure is the prerequisite of carrying out association rule mining, and it is convenient to Computer Processing and recognition data basis for association rule mining provides.
Step 4: association rule mining;
Association rule mining is the core link that carries out knowledge acquisition.This part work is mainly accomplished by task execution module 6.Task execution module 6 is obtained the useful information that lies in the flying quality according to the knowledge acquisition pattern that user 11 selects through association rule mining.
When selecting the knowledge acquisition of one-dimensional flying quality, then utilize the Apriori algorithm to carry out association rule mining.Said Apriori algorithm is a kind of alternative manner of search successively, and K item collection is used for search (K+1) collection.At first find out the set of frequent 1 collection, this set note is made L1, and L1 is used to look for the set L2 of frequent 2 collection, and L2 is used to look for the set L3 of frequent 3 collection, and recursion like this up to can not finding K item collection, thereby is accomplished the excacation of rule.Wherein, K is the integer greater than 0.
When in step 2, selecting the knowledge acquisition of one-dimensional flying quality, association rule mining is specifically realized through following step, and is as shown in Figure 3:
Step a, parameter setting;
It mainly is that minimum support and min confidence to rule is provided with that parameter is provided with, and is a prerequisite step when carrying out association rule mining work.Two kinds of tolerance that the support of rule and degree of confidence are regular interest-degree, they reflect the serviceability and the determinacy of the rule of being found respectively.If rule satisfies condition simultaneously: greater than the minimum support threshold value with greater than minimal confidence threshold, then this rule is significant.If do not set minimum support and min confidence; All all will be stored as significant rule through the rule that association rule mining obtains so; But in fact not all rule that obtains all is significant; Cause the huge of rule base so on the one hand, also can reduce the work efficiency of expert system on the other hand.
The minimum support and the min confidence of rule are rule of thumb set by user 11; The threshold value of regular minimum support and min confidence does not have unified regulation for different expert systems; In the present invention with the threshold value of minimum support and min confidence as open interface; Set by the user; System according to the invention provides the corresponding default value simultaneously, and the minimum support threshold value is traditionally arranged to be the arbitrary value in the scope of [0.2-0.5] closed interval, and minimal confidence threshold is traditionally arranged to be the arbitrary value in the scope of [0.5-0.9] closed interval.
Parameter is set to carry out rule and obtains decision condition is provided, and is the prerequisite of carrying out following step.
Step b, 1 collection of generation candidate;
The iterative process first time that 1 collection of candidate is the Apriori algorithm.Basic process is to scan items all in the flying quality library module 2 by the flying quality knowledge acquisition system that this paper relates to, and the number of times of every appearance is counted.Described is the flying quality after handling through the data set-up procedure, promptly utilizes the flying quality of being convenient to Computer Processing of formalization denotational description.According to the minimum support that parameter is provided with, obtain the item of all item counts greater than minimum support, form 1 collection of candidate.The item count of 1 each concentrated element of candidate is all greater than minimum support.
Step c, generation frequent item set;
Frequent item set is meant through iteration finds out the relevance between the most recurrent each parameter, just the relevance between item and the item.Frequent 2 collection are described two simultaneous set, and frequent 3 collection are described three simultaneous set, and the like.It is the process of iteration progressively that frequent item set generates, and through the search one by one to each item relevance in the flying quality library module 2, accomplishes final iterative process, forms frequent item set, and the item count of frequent item set is greater than the minimal confidence threshold of user's setting.
Steps d, rule output;
Through the minimal confidence threshold that parameter among the step a is provided with, produce at step c on the basis of frequent item set, the output correlation rule, the degree of confidence of the correlation rule that this moment is all is all greater than minimal confidence threshold.
The frequent item set that step c produces has just been represented the rule that finally obtains through association rule mining.
Step e, rule reduction and preservation;
Owing in the data set-up procedure, be the character mark form with data-switching, therefore the rule of output is the character mark form, need reduce to it, promptly is reduced into its represented object to sign.Rule after will reducing is simultaneously preserved, and is convenient to the later stage handle.As in the data set-up procedure, representing rotational speed of lower pressure turbine rotor, then in this process, I1 is replaced with rotational speed of lower pressure turbine rotor with I1.
When elected majority is tieed up the knowledge acquisition of flying quality; Then utilize the Mulit-Rule method to carry out association rule mining; The Multi-Rule method is a kind of method of carrying out the multidimensional association rule mining that the Apriori algorithm proposes that the present invention is based on, and therefore also is a kind of alternative manner of search successively.K item collection is used for search (K+1) collection.At first find out the set of frequent 1 collection, this set note is made L1, and L1 is used to look for the set L2 of frequent 2 collection; L2 is used to look for the set L3 of frequent 3 collection, and recursion like this is up to not finding K item collection; Thereby accomplish the excacation of rule, so a lot of definition of Multi_Rule algorithm are all same or similar with the Apriori algorithm, like minimum support; Min confidence, but it need carry out data hierarchy aspect the data preparation.Wherein, K is the integer greater than 0.
When in step 2, selecting the knowledge acquisition of multidimensional flying quality, utilize the Mulit-Rule method to carry out association rule mining and specifically realize through following step, as shown in Figure 4:
Steps A, tables of data generate;
With in the step 3 through the flying quality information of data hierarchy, promptly the hierarchical information of every kind of flying quality and label are deposited in the database again, and form tables of data, for post-processed provides foundation.The content of the tables of data of said formation mainly comprises the title of the affiliated system of flying quality, the label of flying quality, title and type, the identifier of layering and the interval range of correspondence of flying quality.
Step B, parameter setting;
It mainly is that minimum support and min confidence to rule is provided with that parameter is provided with.The parameter setting is a prerequisite step when carrying out association rule mining work.Two kinds of tolerance that the support of rule and degree of confidence are regular interest-degree.They reflect the serviceability and the determinacy of the rule of being found respectively.If correlation rule satisfies minimum support threshold value and minimal confidence threshold simultaneously, then this rule is significant.
Being provided with of minimum support threshold value and minimal confidence threshold with described in the step a that the one-dimensional flying quality is carried out association rule mining.System according to the invention provides the corresponding default value, and the minimum support threshold value is traditionally arranged to be the arbitrary value in the scope of [0.2-0.5] closed interval, and minimal confidence threshold is traditionally arranged to be the arbitrary value in the scope of [0.5-0.9] closed interval.
Parameter is set to carry out rule and obtains decision condition is provided, and is the prerequisite of carrying out following step.
Step C, generation candidate 1 predicate collection;
Predicate is each attribute in the multidimensional flying quality storehouse, and as comprising barometer altitude, radio altitude, rotational speed of lower pressure turbine rotor attribute in the flying quality storehouse, then each attribute is promptly represented a predicate.Also can be called dimension to different predicates, like this one-dimensional just only relates to a predicate, promptly has only an attribute.And multidimensional just relates to a plurality of predicates, i.e. multidimensional property.
Candidate's 1 predicate collection is the iterative process first time of Multi-Rule algorithm.Basic process is all dimensions in the scanning flying quality library module 2, and dimension is the flying quality of being convenient to Computer Processing that utilizes the formalization denotational description after step 3 is handled in this process, and the number of times that every dimension occurs is counted.According to the minimum support that parameter is provided with, obtain all and satisfy the set of counting greater than minimum support, form candidate's 1 predicate collection.Every dimension counting that candidate's 1 predicate is concentrated is all greater than minimum support.
Step D, generation candidate K predicate collection;
Generate candidate K predicate collection and be the progressively process of iteration,, accomplishes final iterative process through search one by one to each predicate relevance in the database, formation candidate K predicate collection, K represents the maximal value of multidimensional predicate.The minimal confidence threshold that the dimension counting of candidate K predicate collection all is provided with greater than the user.
Step e, rule output;
Through the minimal confidence threshold that step B is provided with, produce on the basis of candidate K predicate collection the output correlation rule at step D.To the set that minimum support requires of satisfying that obtains among the step C, step D further limits on this basis, is met the set of min confidence requirement, and then obtains final regular collection.The candidate K predicate collection that forms among the step D has just been represented the rule that finally obtains through association rule mining.
Step F, rule reduction and preservation;
Owing in the data set-up procedure, data hierarchy is carried out in the numerical value interval of flying quality, and adopt corresponding character label that the data interval of each layer is identified, therefore the rule of output is the character mark form, need reduce to it.This step is reduced into its represented data object to character mark, and the rule after will reducing is simultaneously preserved, and is convenient to the later stage handle.As in the data set-up procedure, utilizing A1 to represent barometer altitude between 0-500, in this process, A is replaced with 0-500 so.
Step 5: rule output;
Rule output is accomplished by regular output module 7, and regular output module 7 is shown to user 11 with the rule of association rule mining through man-machine interface, is convenient to the user and analyzes and judge.
Step 6: rule treatments;
Rule treatments mainly is that redundancy rule is handled, and simultaneously some rules that comprise is each other merged.This part work is mainly accomplished by rule treatments module 8.After the rule that 8 pairs of regular output modules 7 of rule treatments module provide is handled, the rule after handling is offered knowledge acquisition module 9 as knowledge.
The process that the rule that 8 pairs of regular output modules 7 of rule treatments module provide is specifically handled is accomplished by following step, and is as shown in Figure 5:
Step 6.1: import rule;
The rule that the one-dimensional flying quality knowledge acquisition of rule treatments module 8 receiving steps five outputs or the knowledge acquisition of multidimensional flying quality obtain.
Step 6.2 redundancy rule is handled;
Redundancy rule can reduce the efficient of system, causes the unnecessary increase of knowledge base, and the MAINTENANCE OF KNOWLEDGE BASE difficulty is strengthened.The redundancy rule that flying quality knowledge acquisition method of the present invention carries out is handled and is comprised following several kinds:
I: redundancy rule of equal value;
The condition that redundancy rule of equal value is meant a rule and the condition of conclusion and another rule and conclusion be equivalence fully.As:
R1：if?A?and?B?then?C；
R2：if?B?and?A?then?C；
In the foregoing description, R1, R2 represent a rule respectively, the prerequisite of A, B delegate rules, the conclusion of C delegate rules.Rule R1 refers to obtain conclusion C by prerequisite A and B, and regular R2 refers to also obtain conclusion C by prerequisite B and A.
If satisfy this situation through importing two rules that rule obtains, handle can deletion any rule wherein for redundancy rule so.
II: " with " redundancy rule that comprises of condition;
" with " redundancy rule that comprises of condition be meant a rule " with " constraint of condition comprise another rule " with " constraint of condition, but their conclusion is identical.As:
R1：if?A?and?B?then?C；
R2：if?A?then?C；
In the foregoing description, R1, R2 represent a rule respectively, the prerequisite of A, B delegate rules, the conclusion of C delegate rules.Rule R1 refers to obtain conclusion C by prerequisite A and B, and regular R2 refers to obtain conclusion C by prerequisite A.
If satisfy this situation through importing two rules that rule obtains, handle can deletion rule R1 for redundancy rule so.
III: " or " redundancy rule that comprises of condition;
" or " redundancy rule that comprises of condition be meant a rule " or " constraint of condition comprise another rule " or " constraint of condition, but their conclusion is identical.As:
R1：if?A?OR?B?then?C；
R2：if?A?then?C；
In the foregoing description, R1, R2 represent a rule respectively, the prerequisite of A, B delegate rules, the conclusion of C delegate rules.Rule R1 refers to obtain conclusion C by prerequisite A or B, and regular R2 refers to obtain conclusion C by prerequisite A.
If satisfy this situation through importing two rules that rule obtains, handle can deletion rule R2 for redundancy rule so.
IV: " with " redundancy rule that comprises of result;
" with " redundancy rule that comprises of result is that the condition of these two rules is identical, but wherein the result of a rule " with " the condition number more than another regular result " with " the condition number.As:
R1：if?A?then?B?and?C?and?D；
R2：if?A?then?B?and?C；
In the foregoing description, R1, R2 represent a rule respectively, the prerequisite of A delegate rules, the conclusion of B, C, D delegate rules.Rule R1 refers to obtain conclusion B and C and D by prerequisite A, and regular R2 refers to obtain conclusion B and C by prerequisite A.
If satisfy this situation through importing two rules that rule obtains, handle can deletion rule R2 for redundancy rule so.
V: " or " redundancy rule that comprises of result.
" or " redundancy rule that comprises of result is that the conditional number of two rules equates, but wherein the result of a rule " or " condition more than the result of another rule " or " the condition number.As:
R1：if?A?then?B?or?C?or?D；
R2：if?A?then?B?or?C；
In the foregoing description, R1, R2 represent a rule respectively, the prerequisite of A delegate rules, the conclusion of B, C and D delegate rules.Rule R1 refers to obtain conclusion B or C or D by prerequisite A, and regular R2 refers to obtain conclusion B or C by prerequisite A.
If satisfy this situation through importing two rules that rule obtains, handle can deletion R2 for redundancy rule so.
Step 6.3: rule merges;
It mainly is that the rule that comprises is each other handled that rule merges.If there are two rules, their conditions are identical, conclusion comprises or condition comprises, conclusion is identical, then can merge these two rules specifically.Rule merges the rule that is divided into following situation:
I: condition is identical, the conclusion condition of different;
R1:A?and?B=>C
R2:A?and?B=>D
In the foregoing description, R1, R2 represent a rule respectively, the prerequisite of A, B delegate rules, the conclusion of C and D delegate rules.Rule R1 refers to can obtain conclusion C by prerequisite A and B, and regular R2 refers to can obtain conclusion D by prerequisite A and B.
If satisfy this situation through importing two rules that rule obtains so, rule merges and can regular R1 and R2 be deleted so, and forms a new rule:
R3：A?and?B=>C?or?D;
In the foregoing description, R3 represents a rule, the prerequisite of A, B delegate rules, the conclusion of C and D delegate rules.Rule R3 refers to can obtain conclusion C or D by prerequisite A and B.
Ii: condition is different, the situation that conclusion is identical;
R1:A=>B?and?C
R2:D=>B?and?C
In the foregoing description, R1, R2 represent a rule respectively, the prerequisite of A, D delegate rules, the conclusion of B and C delegate rules.Rule R1 refers to can obtain conclusion B and C by prerequisite A, and regular R2 refers to can obtain conclusion B and C by prerequisite D.
If satisfy this situation through importing two rules that rule obtains so, rule merges and can regular R1 and R2 be deleted so, and forms a new rule:
R3：A?or?D=>B?and?C
In the foregoing description, R3 represents a rule, the prerequisite of A, D delegate rules, the conclusion of B and C delegate rules.Rule R1 refers to can obtain conclusion B and C by prerequisite A or D.
Step 6.4: rule is preserved;
The rale store that to pass through above-mentioned steps 6.1 ~ step 6.3 processing back formation is in base module 10.
Step 7: knowledge acquisition;
Knowledge acquisition is last link of system, is mainly accomplished by knowledge acquisition module 9.Knowledge acquisition obtains is rule letter, that can embody data characteristics simultaneously after handling, and with these rules as knowledge store in base module 10, be convenient to the research and the use in later stage.
Claims (7)
1. flying quality knowledge acquisition system; It is characterized in that, comprise flying quality acquisition module, flying quality library module, mode selection module, data preparation module, data disaply moudle, task execution module, regular output module, rule treatments module, knowledge acquisition module and base module;
The flying quality acquisition module obtains flying quality, and the flying quality that obtains is sent to the flying quality library module; The flying quality library module stores the flying quality that obtains in the database table into;
Data preparation module is carried out data processing to the flying quality that writes down in the database table in the flying quality library module; Make the data after the processing can be used in association rule mining, and the data after will handling offer data disaply moudle and task execution module as data source;
Data disaply moudle is shown to the user with the flying quality that receives;
The user selects the knowledge acquisition pattern through mode selection module: the knowledge acquisition of multidimensional flying quality is carried out in knowledge acquisition or the selection selecting to carry out the one-dimensional flying quality, and mode selection module notifies the knowledge acquisition pattern that chooses to task execution module;
Task execution module is according to the knowledge acquisition pattern; Flying quality after the processing of data preparation module is carried out rule obtain work; Under the knowledge acquisition pattern of one-dimensional flying quality; Adopt the Apriori algorithm to carry out association rule mining, under the knowledge acquisition pattern of multidimensional flying quality, adopt the Mulit-Rule method to carry out association rule mining;
The step that described task execution module adopts the Apriori algorithm to carry out association rule mining is: step a, parameter setting: the user rule of thumb sets the minimum support threshold value and the minimal confidence threshold of rule; Step b, 1 collection of generation candidate: all items in the flying quality knowledge acquisition system scanning flying quality library module; Number of times to every appearance is counted; According to the minimum support threshold value that is provided with among the step a; With all item counts 1 collection of item formation candidate greater than the minimum support threshold value, described is the flying quality after handling through data preparation module; Step c, generate frequent item set: it is the process of iteration progressively that frequent item set generates; Through search one by one to each item relevance in the flying quality library module; Accomplish final iterative process, form frequent item set, the minimal confidence threshold that the item count of frequent item set is provided with greater than the user; Steps d, rule output: the frequent item set that step c produces is exactly the rule that finally obtains through association rule mining; Step e, rule reduction and preservation: the rule of output is the character mark form, is reduced into its represented data object to sign, the rule after the preservation reduction;
The concrete steps that described task execution module adopts the Mulit-Rule method to carry out association rule mining are: steps A, tables of data generate: the flying quality after data preparation module is handled; The hierarchical information and the label that comprise every kind of flying quality are deposited in the database again; And the formation tables of data, this tables of data of formation mainly comprises the title of the affiliated system of flying quality, the label of flying quality, the title of flying quality, the type of flying quality, the identifier and the corresponding interval range of layering; Step B, parameter setting: minimum support threshold value and minimal confidence threshold are set; Step C, generate candidate's 1 predicate collection: all dimension in the scan database; Number of times to every dimension occurs is counted; All are satisfied the set formation candidate 1 predicate collection of dimension counting greater than the minimum support threshold value, and described dimension is the flying quality that utilizes the formalization denotational description that obtains after handling through data preparation module; Step D, generation candidate K predicate collection: through search one by one to each predicate relevance in the database; Accomplish final iterative process, form candidate K predicate collection, wherein; K represents the maximal value of multidimensional predicate, the minimal confidence threshold that the dimension counting of candidate K predicate collection all is provided with greater than the user; Step e, rule output: the candidate K predicate collection that generates among the step D is exactly the rule that obtains through association rule mining; Step F, the reduction of the rule of step e output and preserve: be reduced into its represented data object to character mark, the rule after will reducing is simultaneously preserved;
Task execution module is obtained the useful information that lies in the flying quality, and formation rule, offers regular output module;
The rule that the rule output module will obtain is shown to the user, and the rule that will obtain offers the rule treatments module; The rule treatments module is handled the rule that regular output module provides, and mainly comprises the processing of redundancy rule and carries out the merging of rule, and should rule offer knowledge acquisition module as knowledge;
Knowledge acquisition module is shown to the user with the knowledge that receives, and simultaneously knowledge is offered base module; The knowledge store that base module provides knowledge acquisition module is in database.
2. flying quality knowledge acquisition system according to claim 1 is characterized in that, said flying quality acquisition module obtains flying quality through the data reading software of flight parameter registering instrument, perhaps obtains flying quality through flight data file.
3. flying quality knowledge acquisition system according to claim 1; It is characterized in that described data preparation module adopts character marking to handle the one-dimensional flying quality; To the multidimensional flying quality; Data hierarchy is carried out in numerical value interval according to flying quality, adopts the character label that the data interval of each layer is identified, and before carrying out data hierarchy, the numerical exception of flying quality is pressed the front and back Mean Method and handles.
4. an application rights requires the flying quality knowledge acquisition method of 1 described flying quality knowledge acquisition system, it is characterized in that, specifically comprises the steps:
Step 1, obtain flying quality: the flying quality acquisition module obtains flying quality through the data reading software of flight parameter registering instrument; Perhaps obtain flying quality through flight data file, the flying quality acquisition module stores the flying quality that obtains in the flying quality library module into;
Step 2, selection knowledge acquisition pattern: the user is according to the demand of reality; Select the knowledge acquisition pattern that to carry out through mode selection module; Selection is carried out the knowledge acquisition of one-dimensional flying quality or is selected to carry out the knowledge acquisition of multidimensional flying quality, and mode selection module passes to task execution module with the knowledge acquisition pattern of selecting;
Step 3, data are prepared: data preparation module is handled the data in the flying quality library module; Mainly comprise noise, type conversion and the data hierarchy work eliminated; Make data in the flying quality library module become the data that association rule mining can usefulness, the data transfer after data preparation module will be handled is to task execution module;
Step 4, association rule mining: the knowledge acquisition pattern that task execution module is selected according to user in the step 2, to the data after the step 3 processing, obtain the useful information that lies in the flying quality through association rule mining, and with these information formation rules;
Step 5, rule output: regular output module offers the rule treatments module with the rule that step 4 forms, and simultaneously rule is shown to the user through man-machine interface;
Step 6, rule treatments: the rule treatments module is handled the rule that regular output module provides, and the rule after the processing offers knowledge acquisition module as knowledge;
Step 7, knowledge acquisition: the knowledge of knowledge acquisition module receiving step six, arrive base module with this knowledge store, and be shown to the user.
5. flying quality knowledge acquisition method according to claim 4 is characterized in that the data preparation module described in the step 3 is handled the data in the flying quality library module, to the different knowledge obtaining mode, specifically is treated to:
When selecting the knowledge acquisition of one-dimensional flying quality, flying quality is carried out character marking, convert computing machine into and be convenient to the forms of treatment symbol;
When elected majority is tieed up the knowledge acquisition of flying quality; Mean Method is handled before and after at first the numerical exception of flying quality being pressed; Carry out data hierarchy according to the numerical value interval of flying quality on this basis, adopt the character label that the data interval of each layer is identified at last; Described front and back Mean Method is exactly to get the previous point and a back point of exceptional value institute corresponding point, then these two some value corresponding is sued for peace and is averaged.
6. flying quality knowledge acquisition method according to claim 4; It is characterized in that; Parameter setting described in step a or the step B; The arbitrary value of minimum support threshold value default setting in the scope of [0.2-0.5] closed interval, the arbitrary value of minimal confidence threshold default setting in the scope of [0.5-0.9] closed interval.
7. flying quality knowledge acquisition method according to claim 4 is characterized in that, the rule treatments module described in the step 6 is handled the rule that regular output module provides, and concrete processing procedure is:
Step 6.1, importing rule: the rule treatments module receives the rule of regular output module input;
Step 6.2, redundancy rule are handled: several kinds of redundancy rules are handled below main the employing: redundancy rule of equal value; " with " redundancy rule that comprises of condition; " or " redundancy rule that comprises of condition; " with " redundancy rule that comprises of result; " or " redundancy rule that comprises of result;
Step 6.3, rule merge: mainly contain two kinds of situation: if the condition of two rules is identical, the conclusion difference then merges two rules; If the condition of two rules is different, conclusion is identical then to be merged two rules.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010102086422A CN101887531B (en) | 2010-06-13 | 2010-06-13 | Flight data knowledge acquisition system and acquisition method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN2010102086422A CN101887531B (en) | 2010-06-13 | 2010-06-13 | Flight data knowledge acquisition system and acquisition method thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101887531A CN101887531A (en) | 2010-11-17 |
| CN101887531B true CN101887531B (en) | 2012-12-19 |
Family
ID=43073445
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN2010102086422A Expired - Fee Related CN101887531B (en) | 2010-06-13 | 2010-06-13 | Flight data knowledge acquisition system and acquisition method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101887531B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103473409B (en) * | 2013-08-25 | 2016-06-01 | 浙江大学 | The FPGA automatic fault diagnosis method in a kind of knowledge based storehouse |
| CN107015809A (en) * | 2017-04-11 | 2017-08-04 | 中国人民解放军海军航空工程学院青岛校区 | A kind of airplane equipment electronic record system and application process |
| CN107332721B (en) * | 2017-08-27 | 2019-07-12 | 海南大学 | Internet of Things collection of resources transmission optimization system towards typing resource |
| CN108564136B (en) * | 2018-05-02 | 2019-05-03 | 北京航空航天大学 | A kind of airspace operation Situation Assessment classification method based on fuzzy reasoning |
| CN112990466A (en) * | 2021-03-31 | 2021-06-18 | 龙马智芯（珠海横琴）科技有限公司 | Redundancy rule detection method and device and server |
| CN113640699B (en) * | 2021-10-14 | 2021-12-24 | 南京国铁电气有限责任公司 | Fault judgment method, system and equipment for microcomputer control type alternating current and direct current power supply system |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1479249A (en) * | 2002-08-30 | 2004-03-03 | 厦门航空有限公司 | Method of precisely calculating continuouly using time of airplane auxiliary power device through establishing dynamic model |
| CN101083019A (en) * | 2006-12-31 | 2007-12-05 | 中国人民解放军63791部队 | Rapid evaluating system based on roomage state sensing |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6952693B2 (en) * | 2001-02-23 | 2005-10-04 | Ran Wolff | Distributed mining of association rules |
| US20030068617A1 (en) * | 2001-04-09 | 2003-04-10 | Jorng-Tzong Horng | Method for predicting regulatory elements in repetitive sequences using transcription factor binding sites |
| CN101042698A (en) * | 2007-02-01 | 2007-09-26 | 江苏技术师范学院 | Synthesis excavation method of related rule and metarule |
| CN101655857B (en) * | 2009-09-18 | 2013-05-08 | 西安建筑科技大学 | Method for mining data in construction regulation field based on associative regulation mining technology |
-
2010
- 2010-06-13 CN CN2010102086422A patent/CN101887531B/en not_active Expired - Fee Related
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1479249A (en) * | 2002-08-30 | 2004-03-03 | 厦门航空有限公司 | Method of precisely calculating continuouly using time of airplane auxiliary power device through establishing dynamic model |
| CN101083019A (en) * | 2006-12-31 | 2007-12-05 | 中国人民解放军63791部队 | Rapid evaluating system based on roomage state sensing |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101887531A (en) | 2010-11-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101887531B (en) | Flight data knowledge acquisition system and acquisition method thereof | |
| CN104239501B (en) | Mass video semantic annotation method based on Spark | |
| US8898168B2 (en) | Information searching apparatus, information searching method, and computer product | |
| CN107193967A (en) | A kind of multi-source heterogeneous industry field big data handles full link solution | |
| CN106446148A (en) | Cluster-based text duplicate checking method | |
| CN102779143B (en) | Visualizing method for knowledge genealogy | |
| Daoqing et al. | Parallel discrete lion swarm optimization algorithm for solving traveling salesman problem | |
| CN103390063B (en) | A kind of based on ant group algorithm with the search method of related feedback images of probability hypergraph | |
| CN108446391A (en) | Processing method, device, electronic equipment and the computer-readable medium of data | |
| CN106991446A (en) | A kind of embedded dynamic feature selection method of the group policy of mutual information | |
| CN101211344A (en) | Text message ergodic rapid four-dimensional visualization method | |
| CN104021156A (en) | Method for rapidly collecting multi-layer fact data based on SQL statements | |
| Danping et al. | The data mining of the human resources data warehouse in university based on association rule | |
| CN109325062A (en) | A kind of data dependence method for digging and system based on distributed computing | |
| CN107944465A (en) | A kind of unsupervised Fast Speed Clustering and system suitable for big data | |
| CN110597876B (en) | Approximate query method for predicting future query based on offline learning historical query | |
| CN106933844A (en) | Towards the construction method of the accessibility search index of extensive RDF data | |
| Min et al. | Data mining and economic forecasting in DW-based economical decision support system | |
| Li | Glowworm swarm optimization algorithm-and K-prototypes algorithm-based metadata tree clustering | |
| CN114238045A (en) | System and method for judging and automatically repairing integrity of multi-source measurement data of power grid | |
| CN101334863A (en) | Integrated session environment scheme generation system and method based on elemental analysis | |
| Zuo et al. | A tag-aware recommendation algorithm based on deep learning and multi-objective optimization | |
| Usman et al. | Multi level mining of warehouse schema | |
| Wang et al. | Graph transformer: Learning better representations for graph neural networks | |
| Qiuyu et al. | Research on user knowledge acquisition and application in software ecology |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20121219 Termination date: 20150613 |
|
| EXPY | Termination of patent right or utility model |