CN101751399A - Decision tree optimization method and optimization system - Google Patents
Decision tree optimization method and optimization system Download PDFInfo
- Publication number
- CN101751399A CN101751399A CN200810183275A CN200810183275A CN101751399A CN 101751399 A CN101751399 A CN 101751399A CN 200810183275 A CN200810183275 A CN 200810183275A CN 200810183275 A CN200810183275 A CN 200810183275A CN 101751399 A CN101751399 A CN 101751399A
- Authority
- CN
- China
- Prior art keywords
- subtree
- decision tree
- node
- level
- title
- 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
Images
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention discloses a decision tree optimization method and an optimization system; wherein the decision tree optimization method comprises the following steps: subtrees in the decision tree, which have the same name and are positioned under the same father node, are inquired, and the subtrees are stored; when the stored subtrees are repeated subtrees at the same layer, the repeated subtrees at the same layer are combined. In the decision tree optimization method and the optimization system, by combining the repeated subtrees at the same layer in the decision tree and storing the repeated subtrees, the repeated subtree problem is greatly reduced on the premise of ensuring the precision of the decision tree, so as to lead the structure of the decision tree to be compact and cause the operation efficiency to be high.
Description
Technical field
The present invention relates to a kind of decision tree learning technology, relate in particular to a kind of decision tree optimization method and optimization system.
Background technology
Method based on decision tree is by the study to one group of training data, construct the representation of knowledge of form of decision tree, carry out the comparison of property value and judge from the downward branch of this node at the inside of decision tree node, obtain conclusion at the decision tree leaf node according to different property value.So the paths from the root to the leafy node just corresponding a rule, whole decision tree just corresponding one group of expression formula rule of extracting.A biggest advantage based on the decision tree learning algorithm is exactly that it does not need the user to understand a lot of background knowledges in learning process, and its classical algorithm has ID3, C4.5, C5.0, Cart etc.
The too complicated reason of decision tree is because the incorrect of descriptive language caused; Another reason is to have noise in the training example set.In general system is based upon under the Utopian hypothesis prerequisite, sufficient training example is promptly arranged, and the training example is entirely true, yet in fact train and often contain noise data in the example, the quantity of perhaps training example can not produce the representational sampling of objective function very little.When if above-mentioned situation has a kind of the appearance, the tree that algorithm produced will the overfitting training examples.In addition, noise data can cause irrelevant example to be entrained in the selected training set, and this will cause " meaningless modeling " phenomenon.
All there are very serious iteron tree problem in the ID3 and the J48 algorithm (the improvement algorithm of C4.5) of the intelligent auditing strategy of 4A (unified account, authentication, mandate, audit) management system employing at present, and this also is to adopt decision Tree algorithms institute unavoidable problem.Redundant rule makes stores had regular understandable with tree construction, and it is so unobvious that advantages such as differentiation runnability height become, and serious redundancy repeats even deviated from our original intention: adopt tree construction obviously to be better than the rule set form.Therefore, the decision tree after generating is carried out structure optimization, the solution iteron tree problem of very big degree, it is imperative that decision tree is pruned.
Pruning decision tree has a variety of methods, is divided into such five kinds usually:
(1) pre-beta pruning and back beta pruning,
Wherein, back beta pruning is that a decision tree that grows is fully carried out cutting, the input that is algorithm is a decision tree T without beta pruning, and output is a decision tree T ' after the beta pruning, and to be algorithm leave out the result who obtains according to certain principle or standard with one among the T or several subtrees to T '.In the process of beta pruning, replace deleted subtree with leafy node.In the algorithm of beta pruning, the class under the leaf node is set the affiliated classes of great majority training examples with this stalk and is replaced, and marks the shared ratio of training example of affiliated this class on leafy node after some.But,, thereby no longer be 0 for the error rate of training example set through the decision tree T ' of beta pruning because the part subtree of decision tree is by beta pruning.
(2) decision tree is changed into another kind of data structure,
This scheme comprises following step: decision tree is converted to decision diagram; Decision tree is converted to the rule set form removes redundancy rule again; Use complex characteristic, the logical combination of atomic features or arithmetic combination.The simple employing is converted to decision diagram with decision tree, can solve the iteron tree problem greatly to a certain extent being converted to tree structure, but because " or link problems " can reduce tactful precision greatly, and this is that actual audit process institute is unacceptable.Shown in Fig. 1 a and Fig. 1 b, " or link problems " is meant that the excessive merging owing to attribute makes codomain extended in the process of carrying out the merging of repetition subtree.For example, account number acc=a1 is arranged on it for destination host ip=a; Other has a destination host ip=b, and account number acc=b1 is arranged on it; Following decision tree is carried out can occurring or link problems after the decision diagram storage merges.Through can occurring the situation that ip=a and acc=b1 obtain x after merging, and in fact ip does not have account number b1 on the main frame of a, as shown in Figure 1.
(3) extend testing collection, at first constituting by feature is the data-driven or the difference of hypothesis driven, with the characteristics combination set up or cut apart, introduces the multivariate test set then on this basis.But these methods are effectively expanded the decision tree collection when adjusting the decision tree expression;
(4) select to comprise different test set evaluation functions, by improving the description of continuous feature, or revise searching method realization itself;
(5) use constraint database, promptly come reduced decision tree by cutting down database or case description feature set.
Pruning method to decision tree in the prior art can not be taken into account compact conformation of decision tree and the precision of decision tree simultaneously, and " or link problems " greatly reduces the precision of decision tree.Therefore, provide a kind of and can simplify to the structure of decision tree effectively and guarantee that the decision tree optimization method and the optimization system of the precision of decision tree become the technical matters of being badly in need of solution in the prior art.
Summary of the invention
The objective of the invention is to, defective at can not take into account the precision of the compact conformation of decision tree and decision tree in the prior art simultaneously to the pruning method of decision tree provides a kind of decision tree optimization method and the optimization system that can simplify and guarantee the precision of decision tree effectively to the structure of decision tree.
This decision tree optimization method comprises: it is identical and be positioned at subtree under the same father node to inquire about title in the described decision tree, and stores described subtree; When the subtree of described storage is when repeating subtree with layer, merge described with a layer repetition subtree.
The operation of storage subtree specifically comprises: when described subtree is leaf node, store the nodename of described subtree; When described subtree is a father node, when its parent-offspring's node is leaf node, store the title of nodename and all parent-offspring's nodes thereof of described subtree.
When described subtree is leaf node, the nodename of Cun Chu described subtree relatively; When described subtree is a father node, when its parent-offspring's node is leaf node, the title of the nodename of more described subtree and all parent-offspring's nodes thereof.
This decision tree optimization system comprises: enquiry module is used to search and is positioned at the subtree that title is identical and father node ID is identical; The subtree memory module is used to store described subtree; Comparison module, the subtree that is used for relatively storing; Merge processing module, be used for when described subtree for layer repetition subtree the time, described repetition subtree is merged.
Decision tree optimization method of the present invention and optimization system, by being stored the back, the subtree of the same layer in the decision tree judges whether it is the repetition subtree, merge again, under the prerequisite of the precision that guarantees decision tree, reduce the iteron tree problem greatly, make the structure of decision tree more compact, operational efficiency is higher.
Description of drawings
Fig. 1 a, Fig. 1 b are respectively the forward and backward structural drawing of prior art decision tree optimization;
Fig. 2 is the structural drawing of first embodiment of the invention decision tree optimization system;
Fig. 3 is the process flow diagram of first embodiment of the invention decision tree optimization method;
Fig. 4 a is the structural drawing of the second embodiment of the invention decision tree that need be optimized;
Fig. 4 b, Fig. 4 c and Fig. 4 d are in the second embodiment of the invention decision tree optimization process and the structural drawing after optimizing;
Fig. 5 is the decision tree after third embodiment of the invention of the present invention is optimized;
Fig. 6 is the subtree of second embodiment of the invention node D;
Fig. 7 is the subtree of second embodiment of the invention Node B.
Embodiment
The present invention is described in detail below in conjunction with accompanying drawing.
As shown in Figure 2, first embodiment of the invention decision tree optimization system comprises:
Enquiry module 12 is used to search and is positioned at the subtree that title is identical and father node ID is identical;
Subtree memory module 14 is used to store described subtree, when the iteron tree node is designated 0, directly stores this repetition subtree.When the iteron tree node is designated 1, and its parent-offspring's node node identification is 0 o'clock, store the title of nodename and all parent-offspring's nodes thereof of this repetition subtree, concrete storage organization for this node name claims, leaf node title 1, leaf node title 2 ..., leaf node title N}.When the iteron tree node is designated 1, and its parent-offspring's node node identification is 1 o'clock, stores its parent-offspring's node, concrete storage organization for this node name claims, leaf node title 1, leaf node title 2 ..., leaf node title N}.
Comparison module 16, the subtree that is used for relatively storing;
Merge processing module 18, be used for when described subtree when repeating subtree with layer, merge with a layer repetition subtree described.
Preferably, the decision tree optimization system also comprises decision tree memory module 11, and storage decision tree information comprises: the ID of each node, nodename, nodal value, node identification, level and father node ID.The concrete storage organization of storage decision tree information is as follows in the subtree memory module 14:
typedef?struct?tree
{
Int nodeid; // node ID
String nodename; // nodename
Int parented; // father node ID
String nadevalue; // nodal value
Int flag; // sign
Int level; // level
}TREE；
Wherein, nodal value is an expression formula; Node identification, being used to distinguish this node is leaf node or father node, is provided with when node identification is 0, represents that this node is a leaf node, node identification is 1 o'clock, represents that this node is a father node; Level is set to, and level increases from top to bottom, and the hierarchical value that root node is is 0.
Preferably, enquiry module 12 is further used for searching level and differs one-level and the identical subtree of some title; Subtree memory module 14 is further used for storing described level and differs one-level and the identical subtree of some title; Comparison module 16, the level that is further used for more described storage differs one-level and the identical subtree of some title; Merge processing module 18, be further used for when described subtree is different layer of repetition subtree, after the subtree that hierarchical value is big promotes one-level, more described different layer is repeated subtree and merge.
As shown in Figure 3, the first embodiment decision tree optimization method comprises:
The first embodiment decision tree optimization method also comprises step 301, storage decision tree information, and described decision tree information comprises: the ID of each node, nodename, nodal value, node identification, level and father node ID.
Preferably, also comprise step 310, search level and differ one-level and the identical subtree of some title,
Step 312, storage level differ one-level and some title identical subtree;
Step 314 is judged when whether described subtree is different layer of repetition subtree;
Step 316 merges described different layer and repeats subtree.
Step 316 specifically comprises: step 316a, and the repetition subtree that hierarchical value is big promotes one-level; Step 316b merges repeating subtree.
The concrete operations of step 316a are as follows:
Step b1 seeks different layer and repeats the pairing father node of subtree, represents the father node of the repetition subtree that hierarchical value is less with p1, represents the father node of the subtree that hierarchical value is bigger with P2;
Step b2 carries out nodal value for the big repetition subtree of hierarchical value and handles, and seeks this repetition subtree with all brotghers of node under the father node p2, and itself and brotgher of node value are made an amendment, and changes into: the value of father node p2; ﹠amp; Self nodal value;
Step b3 carries out nodal value for the less subtree of hierarchical value and handles, and seeks all brothers under the p1 node, and its nodal value is made an amendment, and changes into: the Zhi ﹠amp of father node p1; ﹠amp; Self nodal value;
Step b4 deletes its father node p1, p2, structure new node new, and each property value of node new all equates with each property value of p1 except nodename;
Step b5 all points to new node new with the numbering of the father node under all former p1, p2, and subtree promotes and finishes.
Second embodiment of the invention is removed the iteron tree optimization to the decision tree shown in Fig. 4 a and is handled, and specifically comprises the steps:
1, the repetition subtree with layer is merged, comprise the steps:
(1) for subtree X, the subtree of node D shown in Figure 6, the subtree of Node B shown in Figure 7 go to judge whether to exist the repetition subtree earlier
(2) be that leaf node then can directly compare for subtree X, whether under with layer, equal leaf node arranged, have directly to merge then to become Fig. 4 b;
(3) carry out special storage earlier for the subtree of node D, judge whether again to exist and repeat subtree, have then to merge.It is stored as { D, 1 ‖ 3,2, " " }, all one-level subtrees all can be converted into this form storage, and whether equate to come two subtrees of comparison whether to equate by comparing two one-dimension array.There is the repetition subtree in the subtree of node D in this example, need carry out subtree and merge, and merges the back as Fig. 4 c;
(4) subtree for Node B be stored as B, " ", C, D, 1 ‖ 3,2} is not through more find repeating subtree.
2, the repetition subtree to different layer merges, and comprises the steps:
(1) father node of node Y is respectively node D and Node B, changes the nodal value of node D lower node Y into B=G ‖ H﹠amp; ﹠amp; D=5; The nodal value of Node B lower node Y changes B=F into;
(2) shown in Fig. 4 d, construct new node new, the node Y under node D and the Node B is merged, the nodal value that merges posterior nodal point Y is (B=F) ‖ (B=G ‖ H﹠amp; ﹠amp; D=5); Deletion of node node D and Node B;
(3) because two nodes X differ two-stage, should this not merge.
After the merging of process with layer merging of repetition subtree and different layer repetition subtree, the decision tree after being optimized is shown in Fig. 4 d.
Third embodiment of the invention, the decision tree optimization method of the present invention of the application in the 4A management platform.In in the 4A management platform, auditing decision tree being optimized, can solve auditing decision tree, problem such as branch time-like performance is not high because to repeat the structure of the tree that subtree causes at random.
In charge system for the database access permission grant of account number root on the database bill1 and file1 relation, wherein account number root for detailed single _ 1 and detailed single _ 2 have operating right and be: select, create, alter, drop, insert, update and delete authority; Account number root for custom1 and custom2 without any operating right; Account number file1 all has the select operating right for all detailed single classes and Customer Location info class form.As shown in the table for object authority mandate relation.
Table 1
Mandate relation according to above can produce following strategy by the intelligent auditing decision tree method among the 4A:
Audit is from account number=root
| operand=detailed single: sensitive data visit
| operand=Customer Location information: unauthorized user operation
Audit is from account number=fil1
| audit operation=DROP: authorized user illegal operation
| audit operation=UPDATE: authorized user illegal operation
| audit operation=CREATE: authorized user illegal operation
| audit operation=DELETE: authorized user illegal operation
| audit operation=INSERT: authorized user illegal operation
| audit operation=ALTER: authorized user illegal operation
| audit operation=SELECT: sensitive data visit
For the shown auditing decision tree of last figure, there is very serious iteron tree problem, as follows:
Audit is from account number=fil1
| audit operation=DROP: authorized user illegal operation
| audit operation=UPDATE: authorized user illegal operation
| audit operation=CREATE: authorized user illegal operation
| audit operation=DELETE: authorized user illegal operation
| audit operation=INSERT: authorized user illegal operation
| audit operation=ALTER: authorized user illegal operation
| audit operation=SELECT: sensitive data visit
As shown in Figure 5, the regular expression of authorized user illegal operation is in the audit operation after the optimization:
Audit from account number==fil1﹠amp; ﹠amp; Audit operation
＝＝DROP‖UPDATE‖CREATE‖DELETE‖INSERT‖ALTER
Adopt decision tree optimization method of the present invention and optimization system, can reduce the iteron tree problem greatly under the prerequisite that guarantees nicety of grading, make the structure of decision tree more compact, operational efficiency is higher.
It should be noted that: above embodiment is only unrestricted in order to explanation the present invention, and the present invention also is not limited in above-mentioned giving an example, and all do not break away from the technical scheme and the improvement thereof of the spirit and scope of the present invention, and it all should be encompassed in the claim scope of the present invention.
Claims (9)
1. a decision tree optimization method is characterized in that, comprising:
It is identical and be positioned at subtree under the same father node to inquire about title in the described decision tree, and stores described subtree;
When the subtree of described storage is when repeating subtree with layer, merge described with a layer repetition subtree.
2. decision tree optimization method according to claim 1 is characterized in that, the operation of the described subtree of described storage specifically comprises:
When described subtree is leaf node, store the nodename of described subtree;
When described subtree is a father node, when its parent-offspring's node is leaf node, store the title of nodename and all parent-offspring's nodes thereof of described subtree.
3. decision tree optimization method according to claim 2 is characterized in that, also comprises after the operation of the described subtree of described storage:
When described subtree is leaf node, the nodename of Cun Chu described subtree relatively;
When described subtree is a father node, when its parent-offspring's node is leaf node, the title of the nodename of more described subtree and all parent-offspring's nodes thereof.
4. according to any described decision tree optimization method in the claim 1 to 3, it is characterized in that, also comprise before the operation of the subtree in the described decision tree of described inquiry:
Store described decision tree information, described decision tree information comprises: the ID of each node, nodename, nodal value, node identification, level and father node ID.
5. decision tree optimization method according to claim 4 is characterized in that, also comprises:
Search level and differ one-level and the identical subtree of some title, and store described subtree;
When described subtree is different layer of repetition subtree, merges described different layer and repeat subtree.
6. decision tree optimization method according to claim 5 is characterized in that, the operation that the described different layer of described merging repeats subtree specifically comprises:
After the subtree that hierarchical value is big promotes one-level, more described different layer is repeated subtree and merge.
7. a decision tree optimization system is characterized in that, comprising:
Enquiry module is used to search and is positioned at the subtree that title is identical and father node ID is identical;
The subtree memory module is used to store described subtree;
Comparison module, the subtree that is used for relatively storing;
Merge processing module, be used for when described subtree when repeating subtree with layer, merge with a layer repetition subtree described.
8. decision tree optimization according to claim 7 system, it is characterized in that, also comprise the decision tree memory module, be used to store described decision tree information, described decision tree information comprises: the ID of each node, nodename, nodal value, node identification, level and father node ID.
9. decision tree optimization according to claim 8 system is characterized in that, described enquiry module is further used for searching level and differs one-level and the identical subtree of some title;
Described subtree memory module is further used for storing described level and differs one-level and the identical subtree of some title;
Described comparison module, the level that is further used for more described storage differs one-level and the identical subtree of some title;
Described merging processing module is further used for when described subtree is different layer of repetition subtree, after the subtree that hierarchical value is big promotes one-level, more described different layer is repeated subtree and merges.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810183275A CN101751399A (en) | 2008-12-12 | 2008-12-12 | Decision tree optimization method and optimization system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200810183275A CN101751399A (en) | 2008-12-12 | 2008-12-12 | Decision tree optimization method and optimization system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN101751399A true CN101751399A (en) | 2010-06-23 |
Family
ID=42478390
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200810183275A Pending CN101751399A (en) | 2008-12-12 | 2008-12-12 | Decision tree optimization method and optimization system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101751399A (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102214213A (en) * | 2011-05-31 | 2011-10-12 | 中国科学院计算技术研究所 | Method and system for classifying data by adopting decision tree |
| CN102662939A (en) * | 2011-11-17 | 2012-09-12 | 福建榕基软件股份有限公司 | Tree-form data structure displaying method and system |
| CN103530379A (en) * | 2013-10-15 | 2014-01-22 | 福建榕基软件股份有限公司 | Tree structure data display method and device |
| CN104035923A (en) * | 2013-03-04 | 2014-09-10 | 阿里巴巴集团控股有限公司 | Data inquiry method and device |
| CN104111920A (en) * | 2013-04-16 | 2014-10-22 | 华为技术有限公司 | Decision-making tree based prediction method and device |
| CN104143152A (en) * | 2014-07-04 | 2014-11-12 | 北京思特奇信息技术股份有限公司 | Business relation binary tree node pruning method and device |
| CN104537010A (en) * | 2014-12-17 | 2015-04-22 | 温州大学 | Component classifying method based on net establishing software of decision tree |
| CN104657447A (en) * | 2015-02-05 | 2015-05-27 | 上海达梦数据库有限公司 | Database management system-oriented plan tree optimizing method and device |
| CN104778608A (en) * | 2015-04-13 | 2015-07-15 | 合一信息技术（北京）有限公司 | N+ advertisement putting and optimizing method |
| CN105159758A (en) * | 2015-08-07 | 2015-12-16 | 北京思特奇信息技术股份有限公司 | Relationship object recursion merging calculation method and system |
| CN107293298A (en) * | 2016-04-05 | 2017-10-24 | 富泰华工业（深圳）有限公司 | Speech control system and method |
| WO2023246146A1 (en) * | 2022-06-23 | 2023-12-28 | 上海淇玥信息技术有限公司 | Target security recognition method and apparatus based on optimization rule decision tree |
-
2008
- 2008-12-12 CN CN200810183275A patent/CN101751399A/en active Pending
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102214213A (en) * | 2011-05-31 | 2011-10-12 | 中国科学院计算技术研究所 | Method and system for classifying data by adopting decision tree |
| CN102662939A (en) * | 2011-11-17 | 2012-09-12 | 福建榕基软件股份有限公司 | Tree-form data structure displaying method and system |
| CN104035923A (en) * | 2013-03-04 | 2014-09-10 | 阿里巴巴集团控股有限公司 | Data inquiry method and device |
| CN104035923B (en) * | 2013-03-04 | 2017-09-01 | 阿里巴巴集团控股有限公司 | Inquire about the method and device of data |
| CN104111920A (en) * | 2013-04-16 | 2014-10-22 | 华为技术有限公司 | Decision-making tree based prediction method and device |
| CN104111920B (en) * | 2013-04-16 | 2018-03-09 | 华为技术有限公司 | A kind of Forecasting Methodology and device based on decision tree |
| CN103530379B (en) * | 2013-10-15 | 2017-01-04 | 福建榕基软件股份有限公司 | tree structure data display method and device |
| CN103530379A (en) * | 2013-10-15 | 2014-01-22 | 福建榕基软件股份有限公司 | Tree structure data display method and device |
| CN104143152A (en) * | 2014-07-04 | 2014-11-12 | 北京思特奇信息技术股份有限公司 | Business relation binary tree node pruning method and device |
| CN104143152B (en) * | 2014-07-04 | 2017-12-19 | 北京思特奇信息技术股份有限公司 | A kind of business relations y-bend tree node pruning method and device |
| CN104537010A (en) * | 2014-12-17 | 2015-04-22 | 温州大学 | Component classifying method based on net establishing software of decision tree |
| CN104657447B (en) * | 2015-02-05 | 2018-04-03 | 上海达梦数据库有限公司 | The plan tree optimization method and apparatus of oriented database management system |
| CN104657447A (en) * | 2015-02-05 | 2015-05-27 | 上海达梦数据库有限公司 | Database management system-oriented plan tree optimizing method and device |
| CN104778608A (en) * | 2015-04-13 | 2015-07-15 | 合一信息技术（北京）有限公司 | N+ advertisement putting and optimizing method |
| CN105159758A (en) * | 2015-08-07 | 2015-12-16 | 北京思特奇信息技术股份有限公司 | Relationship object recursion merging calculation method and system |
| CN105159758B (en) * | 2015-08-07 | 2018-11-13 | 北京思特奇信息技术股份有限公司 | A kind of relationship object recurrence merges acquiring method and system |
| CN107293298A (en) * | 2016-04-05 | 2017-10-24 | 富泰华工业（深圳）有限公司 | Speech control system and method |
| WO2023246146A1 (en) * | 2022-06-23 | 2023-12-28 | 上海淇玥信息技术有限公司 | Target security recognition method and apparatus based on optimization rule decision tree |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101751399A (en) | Decision tree optimization method and optimization system | |
| AU2002229734B2 (en) | Database system and query optimiser | |
| Romero et al. | GEM: Requirement-driven generation of ETL and multidimensional conceptual designs | |
| CN106372177B (en) | Support the correlation inquiry of mixed data type and the enquiry expanding method of fuzzy grouping | |
| US20130006968A1 (en) | Data integration system | |
| CN105830064A (en) | Scenario generation device and computer program therefor | |
| CN105302803B (en) | A kind of product BOM variance analysis and synchronous updating method | |
| US20120209855A1 (en) | Bit-string key classification/distribution apparatus, classification/distribution method, and program | |
| CN106104519A (en) | Phrase is to collection device and is used for its computer program | |
| CN102893281A (en) | Information retrieval device, information retrieval method, computer program, and data structure | |
| KR101987915B1 (en) | System for generating template used to generate query to knowledge base from natural language question and question answering system including the same | |
| US7640266B2 (en) | Computer program product and associated methods for searching a database of objects, connecting objects in such a database, and exporting data from at least one arbitrary database | |
| CN106104524A (en) | Complex predicate template collection device and be used for its computer program | |
| CN115329504B (en) | BOM construction method based on complex product structure | |
| US7159171B2 (en) | Structured document management system, structured document management method, search device and search method | |
| CN110389950A (en) | A kind of big data cleaning method quickly run | |
| CN101174316A (en) | Device and method for cases illation based on cases tree | |
| US20070220420A1 (en) | System and method for efficient maintenance of indexes for XML files | |
| CN101216845A (en) | Database automatic classification method | |
| JP2007535009A (en) | A data structure and management system for a superset of relational databases. | |
| Pradeepini et al. | Tree-based incremental association rule mining without candidate itemset generation | |
| Karegar et al. | Data-mining by probability-based patterns | |
| CN106294652A (en) | Web page information search method | |
| CN106168982A (en) | Data retrieval method for particular topic | |
| Bai et al. | Optimized algorithm for skyline community discovery in multi-valued networks |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C12 | Rejection of a patent application after its publication | ||
| RJ01 | Rejection of invention patent application after publication |
Open date: 20100623 |