CN106529676A - Deductive lattice and reasoning method based on deductive lattice - Google Patents

Deductive lattice and reasoning method based on deductive lattice Download PDF

Info

Publication number
CN106529676A
CN106529676A CN201610985211.4A CN201610985211A CN106529676A CN 106529676 A CN106529676 A CN 106529676A CN 201610985211 A CN201610985211 A CN 201610985211A CN 106529676 A CN106529676 A CN 106529676A
Authority
CN
China
Prior art keywords
node
lattice
rule
condition
rules
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.)
Granted
Application number
CN201610985211.4A
Other languages
Chinese (zh)
Other versions
CN106529676B (en
Inventor
胡启平
胡煜州
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Individual
Original Assignee
Individual
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Individual filed Critical Individual
Priority to CN201610985211.4A priority Critical patent/CN106529676B/en
Publication of CN106529676A publication Critical patent/CN106529676A/en
Application granted granted Critical
Publication of CN106529676B publication Critical patent/CN106529676B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N5/00—Computing arrangements using knowledge-based models
    • G06N5/04—Inference or reasoning models

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Computing Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Physics & Mathematics (AREA)
  • Computational Linguistics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Artificial Intelligence (AREA)
  • Devices For Executing Special Programs (AREA)

Abstract

The invention discloses a deductive lattice and a reasoning method based on the deductive lattice. Firstly, all rules which are in accordance with a condition is searched in a rule database, then the rules generate a deductive lattice according to a deductive lattice construction algorithm, then the condition matching is carried out, and a rule can be rapidly matched, or all existing rules can not matched. The deductive lattice of the invention is a novel efficient and intelligent reasoning model, through collected criteria, combined with the matching of the rules in the rule database, the layer-by-layer deduction from a root node is started by using a similar lattice structure, and finally a deduction result is obtained at a leaf node. In the deduction process, the statistical and analysis result of criteria is also used to assist the deduction, while a determined deduction result is obtained, other suspected conclusions can be deducted, the suspected conclusions are sorted according to confidence levels, and thus comprehensive deductive information is provided for a decision maker for reference.

Description

It is a kind of to deduce lattice and based on the inference method for deducing lattice
Technical field
The invention belongs to the application of computer system solving complexity problem, is related to a kind of inference method, and in particular to It is a kind of to deduce lattice and based on the inference method for deducing lattice.
Background technology
Existing inductive decision method includes forward reasoning and backward reasoning.
Forward reasoning is also referred to as forward chained reasoning, is a kind of inference mode that conclusion is released by condition.It goes out from one group of fact Send out, using certain rule of inference, prove the establishment of the target fact or proposition.General reasoning process is first to synthetic data Storehouse provides some initial known facts, and control system is matched with the knowledge in knowledge base using these data, is triggered Knowledge, using its conclusion as it is new the fact be added in integrated database.Repeat said process, with updated integrated database The fact that middle, a knowledge another with knowledge base was matched again, its conclusion was updated in integrated database, until not having to match New knowledge and till the fact that there is no longer new being added in integrated database.Then test whether to be solved, have solution then to return Solution, then points out operation failure without solution.
The inference step of forward reasoning is：
1) by initial, known true feeding integrated database；
2) solution of whether included problem in integrated database is checked, is solved if having and is terminated, otherwise perform next step；
3) the initial known fact is matched with the knowledge in knowledge base, if having, turns 4), otherwise turn 6)；
4) by all of knowledge architecture that the match is successful into a Knowledge Set；
If 5) Knowledge Set is not sky, selects a rule to make inferences by certain Strategy of Conflict Resolution, and be pushed out New fact be updated to integrated database, then turn 2)；If Knowledge Set is sky, turn 6)
6) the fact that can provide new is asked the user whether, integrated database is added it to if having, turned 3)；Otherwise represent Problem solving fails, and exits.
The advantage of forward reasoning control strategy is the relevant information (new fact) that user can provide problem on one's own initiative, and And reaction is given in time.Many operations unrelated with problem are performed in being disadvantageous in that solution procedure, have certain blindness, It is less efficient, many and unrelated sub-goal of problem may be released in reasoning process.
Backward reasoning is also referred to as reverse strand reasoning, also known as backward inference, target reasoning.Be one kind by conclusion, to test The correctness for demonstrate,proving the conclusion goes the inference mode of evidence is looked in knowledge base, once evidence is found in knowledge base, then shows the knot By establishment.Its fundamental inference process is to propose a collection of hypothesis (target), verifies that these are assumed one by one using one group of knowledge correct Property.
The inference step of backward reasoning is：
1) being given needs the target of checking；
2) target is checked whether in integrated database, if successfully releasing, otherwise, turning next step；
3) judge whether the target is evidence, i.e., being whether should be by the initial fact of user-approved.If so, then inquiry is used Family, otherwise, turns next step；
4) the be possible to rule for deriving the target is found out in knowledge base, form the Knowledge Set being suitable for, then turn next Step；
5) rule is selected from Knowledge Set, and 2) then the former piece of the knowledge is turned as new hypothesis target.
The advantage of backward reasoning control strategy is that purposiveness is strong, it is not necessary to find the information unrelated with hypothesis and knowledge.It is this Strategy provides the more accurate rule (knowledge) explained, tell that user to be reached target is used to reasoning process.In addition, this control System strategy is especially suitable under the less problem solving environment of solution space, and it is beneficial to provide a user with solution procedure.Shortcoming is The selection of initial target has blindness, it is impossible to operated by the useful information of user's offer.Generally, user requires to rapidly input Corresponding problem domain, if not meeting reality, repeatedly propose it is assumed that affecting system effectiveness.Compared with forward reasoning, inversely The purposiveness of reasoning is very strong, is generally used for verifying whether a certain specific knowledge is set up.
The content of the invention
In order to solve above-mentioned technical problem, the invention provides a kind of deduce lattice and based on the inference method for deducing lattice, energy Analysis and the diagnosis system for being responsible for challenge is built by intellectuality, visualization, digitized, interactive device.
A kind of deduction lattice that the present invention is provided, it is characterised in that：
Define first：
(1) lattice：Consider any one partial ordering set (L ,≤), if to arbitrary element a, b in set L so that a, b exist There is an infimum and supremum in L, then (L ,≤) it is lattice；
(2) rule：C1∧C2∧...∧Cn→h；
Wherein, C1、C2、…、CnIt is condition, condition has true-false value, when for true time, condition then says that condition meets, h is knot By；Work as condition C1、C2、…、CnWhen being true, corresponding conclusion h is just drawn；
If there is rule C1∧C2∧(Ci∨…∨Ci+n) → h, then need for the rule to split into multiple rules, respectively： C1∧C2∧Ci→ h ..., C1∧C2∧Ci+n→h；
If there is rule C1∧C2∧C3→h1And h1∧C4∧C5→ h, then need to merge the two rules, after merging For C1∧C2∧C3∧C4∧C5→h；
If presence takes corresponding actions according to condition, first reached a conclusion using rule, accordingly moved further according to conclusion Make；
(3) knowledge base：Include the fact that storehouse and rule base；Knowledge is included the fact that and rule, in the known integrated circuit it is a fact that Jing man-machine interactivelies are input into Or be stored in factbase；The collection of strictly all rules is collectively referred to as rule base, is designated as RuleBase；
Then lattice structure is deduced in setting：
Deduce lattice and one lattice with hierarchical relationship is constituted by n finite element, the element in lattice is referred to as node, wherein, n >1；Node includes root node, intermediate node and leaf node；Node without forerunner is referred to as root node, and root node is unique；No Follow-up node is referred to as leaf node；Remaining node is referred to as intermediate node, and intermediate node is divided into two types：A kind of node its Follow-up or intermediate node, referred to as pure intermediate node, another kind of intermediate node its it is follow-up be leaf node, referred to as leaf forerunner section Point, root node and intermediate node it is corresponding be rule condition, it is the conclusion of rule that leaf node is corresponding, from root node to leaf A rule in child node rule of correspondence storehouse；
Strictly all rules composition rule collection of the screening with common conditions C from rule base, this rule set is constituted and is deduced Lattice, its root node are exactly condition C；For arbitrary non-leaf nodes Node, it is assumed that its son node number is k, respectively C1, C2..., Ck, C1It is first child node of Node, C2It is second child node ... of Node, CkIt is k-th son section of Node Point, TreeSet (Ci) it is with CiFor root node and CiThe node set of the tree that follow-up intermediate node is constituted, not including leaf section Point, CiMeet：
Finally build and deduce lattice：
The strictly all rules that condition C is met in assuming RuleBase has n bars, and the rule set RuleSet that these rules are constituted has：
RuleName1：C∧C11∧C12∧…∧C1m1->h1
RuleName2：C∧C21∧C22∧…∧C2m2->h2
…………
RuleNamen：C∧Cn1∧Cn2∧…∧Cnmn->hnmn
Alternative condition C is used as root node, and is designated as C ' as present node, then remaining condition is merged public , it is assumed that there are k public keys in remaining condition, to each public keys Ci：
(1) there is C in counting RuleSet rule setsiNumber of times be designated as Ci.Count；
(2) include C in RuleSetiRule name constitute set be designated as Ci.RuleNames；
(3) by Ci.Count by being ranked up from big to small, i=1 ..., k, the result of sequence are designated as C1.Count ..., Ck.Count, and to Ci.RuleNames do correspondence order to adjust, the result of adjustment is designated as C1.RuleNames ..., Ck.RuleNames；
From the beginning of j=1, from Cj.RuleNames take out RuleName in successively and judge Cl.RulesNames whether include RuleName, such as the C including ifl.Count=Cl.Count-1, and by RuleName from Cl.RuleNames remove, l=j+ 1 ..., k, not comprising not processing；Till j=k；After so processing, C in k public keysi.Count what is be not zero is public Also have k ' individual, it is clear that k '≤k, be thus the individual disjoint set of k ' RuleSet point, each of which set it is maximum public Item is respectively C1..., Ck', by C1..., Ck' child node in order from left to right as C '；C1..., Ck' constitute brother's section Point, what they were from left to right ordered into；
Respectively with C1..., Ck' as present node C ', respectively using the individual disjoint sets of this k ' as rule set RuleSet, Above-mentioned principle circular treatment is pressed to remaining condition in RuleSet, until all of Ci.Count be zero, lattice are so deduced with regard to structure Build and complete.
It is a kind of based on the inference method for deducing lattice that the present invention is provided, it is characterised in that comprises the following steps：
Step 1：User is selected with the phenomenon being actually consistent as initial condition from multiple phenomenons；
Step 2：System is searched from rule base RuleBase and meets the strictly all rules formation rule collection of initial condition and be designated as RuleSet；
Step 3：System deduces lattice to the strictly all rules dynamic construction in RuleSet, and using root node as present node；
Step 4：The child node for deducing present node in lattice is formed system problem and condition is unsatisfactory for being supplied to above User, selects for user；
Step 5：User selects one of them as response input system according to practical situation；If noodles in user's selection Part is false, and at this moment represents that system convention storehouse does not meet the rule of correlated condition, this Flow ends；
Step 6：System carries out condition coupling to all child nodes of present node according to the input of user, and the match is successful adjusts Whole present node is the child node that the match is successful, repeat step 4-6 until reaching leaf node, with it is concluded that.
Preferably, it is true that matching described in step 6 is exactly condition value.
Preferably, thick system of reaching a conclusion in step 6 can also be made corresponding actions to meet reality according to actual needs When the needs that control.
Present invention is mainly used for the analysis of challenge and solution procedure.Core is based on the intelligent decision mistake for deducing lattice Journey, compares with existing forward reasoning with backward reasoning mode, deduces lattice and is more applicable for the deduction of complexity problem and solved Journey, the aspect such as comprehensive of the efficiency, inference conclusion when its advantage is mainly reflected in solve problem：
1st, obtain true from user using interactive approach.The real-time interactive mode that deduction lattice are adopted, fast and easy, The fact that obtain user input in time, so as to facilitate inference machine to be differentiated；
2nd, the matching process complexity for deducing lattice is low.Forward reasoning and backward inference are required for when matching every time will be whole Individual regular library searching one time, and lattice are deduced when deduction activity is carried out, need the rule of matching gradually decrease, so as to subtract as far as possible The complexity and resource consumption of few matching；
3rd, the matching process for deducing lattice is more scientific rationally, to criterion (fact) and regular consideration more comprehensively.Deducing lattice will Criterion (fact) is divided into two parts of classification and pattern, first classification is matched when matching, then pattern is entered Row matching, determines the action of next step according to the matching result of classification and pattern；
4th, matching efficiency height, the consumption resource for deducing lattice is few.In the matching process, deduce lattice constantly to enter matching degree Row is calculated, for individual path of the matching degree less than threshold value carries out " beta pruning ", it is to avoid the consumption for no reason of resource, there is provided matching effect Rate；
5th, the branched structure for deducing lattice is more flexible.The matching for deducing lattice is not only consideration the path that the match is successful, for Type matching success, the unsuccessful branch of pattern match will also continue to deduce, it is to avoid the disappearance of highly doubtful conclusion；
6th, by using statistical data result, improve and deduce lattice deduction accuracy.During the deduction for deducing lattice, pass through Result statistical data is introduced, strengthens the accuracy of deduction process, and can be to the branch node of deduction and conclusion according to statistics Data are ranked up, and these are that forward reasoning and backward inference method cannot possess.
Description of the drawings
Fig. 1：The structural representation of the deduction lattice of the embodiment of the present invention.
Specific embodiment
Understand for the ease of those of ordinary skill in the art and implement the present invention, below in conjunction with the accompanying drawings and embodiment is to this It is bright to be described in further detail, it will be appreciated that enforcement example described herein is merely to illustrate and explains the present invention, not For limiting the present invention.
A kind of deduction lattice that the present invention is provided,
Define first：
(1) lattice：Consider any one partial ordering set (L ,≤), if to arbitrary element a, b in set L so that a, b exist There is an infimum and supremum in L, then (L ,≤) it is lattice；
(2) rule：C1∧C2∧...∧Cn→h；
Wherein, C1、C2、…、CnIt is condition, h is conclusion；Work as condition C1、C2、…、CnWhen being true, corresponding conclusion is just drawn h；
If there is rule C1∧C2∧(Ci∨…∨Ci+n) → h, then need for the rule to split into multiple rules, respectively： C1∧C2∧Ci→ h ..., C1∧C2∧Ci+n→h；
If there is rule C1∧C2∧C3→h1And h1∧C4∧C5→ h, then need to merge the two rules, after merging For C1∧C2∧C3∧C4∧C5→h；
If there is C1∧C2∧C3This kind of systems of → action do not allow exist it is regular when, then first derive conclusion, then root Corresponding actions are carried out according to conclusion；
(3) knowledge base：Knowledge base：Include the fact that storehouse and rule base；Knowledge is included the fact that and rule, in the known integrated circuit it is a fact that Jing people's industry and traffic Mutually it is input into or has been stored in factbase；The collection of strictly all rules is collectively referred to as rule base, is designated as RuleBase；
Then lattice structure is deduced in setting：
Deduce lattice and one lattice with hierarchical relationship is constituted by n finite element, the element in lattice is referred to as node, wherein, n >1；Node includes root node, intermediate node and leaf node；Node without forerunner is referred to as root node, and root node is unique；No Follow-up node is referred to as leaf node；Remaining node is referred to as intermediate node, and intermediate node is divided into two types：A kind of node its Follow-up or intermediate node, referred to as pure intermediate node, another kind of intermediate node its it is follow-up be leaf node, referred to as leaf forerunner section Point, leaf predecessor node only have a leaf node, root node and intermediate node it is corresponding be rule condition, leaf node pair Answer be rule conclusion, from root node in the leaf node rule of correspondence storehouse rule；
Strictly all rules composition rule collection of the screening with common conditions C from rule base, this rule set is constituted and is deduced Lattice, its root node are exactly condition C；For arbitrary non-leaf nodes Node, it is assumed that its son node number is k, respectively C1, C2..., Ck, C1It is first child node of Node, C2It is second child node ... of Node, CkIt is k-th son section of Node Point, TreeSet (Ci) it is with CiFor root node and CiThe node set of the tree that follow-up intermediate node is constituted, not including leaf section Point, CiMeet：
Such deduction lattice meet following property：
It is assumed that starting to present node Node conditions all to meet from root node condition, then next condition can be in C1..., CkMiddle searching, if Ci meets, C1..., Ci-1, Ci+1..., CkIt is unsatisfactory for；
By (1) formula it is recognised that due to CiDo not existMiddle appearance, then CiNext step when meeting Should be just in CiSubtree in find the condition that can meet；
Finally build and deduce lattice：
The strictly all rules that condition C is met in assuming RuleBase has n bars, and the rule set RuleSet that these rules are constituted has：
RuleName1：C∧C11∧C12∧…∧C1m1->h1
RuleName2：C∧C21∧C22∧…∧C2m2->h2
…………
RuleNamen：C∧Cn1∧Cn2∧…∧Cnmn->hnmn
Alternative condition C is used as root node, and is designated as C ' as present node, then remaining condition is merged public , it is assumed that there are k public keys in remaining condition, to each public keys Ci：
(1) there is C in counting RuleSet rule setsiNumber of times be designated as Ci.Count；
(2) include C in RuleSetiRule name constitute set be designated as Ci.RuleNames；
(3) by Ci.Count by being ranked up from big to small, the result of sequence is designated as C1.Count ..., Ck.Count, and To Ci.RuleNames do correspondence order to adjust, the result of adjustment is designated as C1.RuleNames ..., Ck.RuleNames；
From the beginning of j=1, from Cj.RuleNames take out RuleName in successively and judge Cl.RulesNames whether include RuleName, such as the C including ifl.Count=Cl.Count-1, and by RuleName from Cl.RuleNames remove, l=j+ 1 ..., k, not comprising not processing；Till j=k；After so processing, C in k public keysi.Count what is be not zero is public Also have k ' individual, it is clear that k '≤k, be thus the individual disjoint set of k ' RuleSet point, each of which set it is maximum public Item is respectively C1..., Ck’, by C1..., Ck’Child node from left to right as C ' in order；C1..., Ck' brotgher of node is constituted, What they were from left to right ordered into；
Respectively with C1..., Ck’As present node C ', respectively using the individual disjoint sets of this k ' as rule set RuleSet, Above-mentioned principle circular treatment is pressed to remaining condition in RuleSet, until all of Ci.Count be zero, lattice are so deduced with regard to structure Build and complete.
It is a kind of based on the inference method for deducing lattice that the present invention is provided, and comprises the following steps：
Step 1：User is selected with the phenomenon being actually consistent as initial condition from multiple phenomenons；
Step 2：System is searched from rule base RuleBase and meets the strictly all rules formation rule collection of initial condition and be designated as RuleSet；
Step 3：System deduces lattice to the strictly all rules dynamic construction in RuleSet, and using root node as present node；
Step 4：The child node for deducing present node in lattice is formed problem and is supplied to user by system, is selected for user；
Step 5：User selects one of them as response input system according to practical situation；If noodles in user's selection Part is false, and at this moment represents that system convention storehouse does not meet the rule of correlated condition, this Flow ends；
Step 6：System carries out condition coupling according to the input of user to all child nodes of present node, and (matching is exactly bar Part value is true), adjustment present node is the child node that the match is successful that the match is successful, repeat step 4-6 until reaching leaf node, To obtain inference conclusion.
Fig. 1 is asked for an interview, after condition A is input to system, system can be searched out from rule base RuleSet in all former pieces Regular composition rule collection comprising A, addition have following rule to meet：
Rule1:A∧B∧C∧D→h1
Rule2:A∧C∧D∧E→h2
Rule3:A∧D∧E∧F→h3
Rule4:A∧G∧H→h3
A is the most important condition, used as the whole root node for deducing lattice；In Rule1, Rule2, in Rule3, in addition to root node A, Frequency of occurrences highest is criterion D, therefore D and G is used as next layer of intermediate node；The high criterion of the frequency of occurrences time has C and E, For twice, therefore C and E is the child node of D, and C and E is in same level, forms the remaining part for deducing lattice.
First from top to bottom, then from left to right the Main Inference order for deducing lattice is.By taking Fig. 1 as an example, A is primary bar Part, so as the root node for deducing lattice.So the rule comprising A conditions from rule set searches for out, this deduction is formed Rule set, and deduction lattice this rule set being built into shown in Fig. 1.
D and G are the conditions for next needing matching, if the matching of condition D, is matched under D branches, under G branches Ignore, condition C is matched, if C is matched continue first to match E downwards, if E is mismatched, match B；If D is matched It is unsuccessful, then just give up D branches, need to scan for G branches matching, if G matchings are unsuccessful, illustrate nothing in rule set The rule of matching.
The reasoning results for deducing lattice have two：
1) deduction process finds leaf node, that is, have found uniqueness conclusion, and deduction terminates；
2) find that certain layer of intermediate node cannot all be matched during deducing, that is, lack the Rule content of present problems, need Supplement new rule to deduce again, deduction terminates.
The technological core of the present invention is intelligent decision, and intelligent decision is data mining, pattern recognition, specialist system, artificial One of most active research contents in the field such as intelligence and machine learning, it has by which to the automatic study of data, Cong Zhongti Take wherein implicit rule or model, and make the great ability of intelligent decision, have a very wide range of applications in reality (example Such as：Diagnostic analysiss, Financial Risk Forecast, management decision-making etc.).
For comparing forward reasoning and backward reasoning, it is a kind of new and effective, intelligent inference pattern to deduce lattice, by receiving The criterion of collection, with reference to the matching to rule in rule base, starts successively to deduce from root node using similar grating texture, and most Obtain at leaf node eventually and deduce result.During deduction, can also be aided in using the statistic analysis result of some criterions Deduce, other doubtful conclusions can be also deduced out while the deduction result for determining is obtained, and by these doubtful conclusions according to which Confidence level is ranked up, and refers to for policymaker so as to provide comprehensive deduction information.
It should be appreciated that the part that this specification is not elaborated belongs to prior art.
It should be appreciated that the above-mentioned description for preferred embodiment is more detailed, therefore can not be considered to this The restriction of invention patent protection scope, one of ordinary skill in the art are being weighed without departing from the present invention under the enlightenment of the present invention Under the protected ambit of profit requirement, replacement can also be made or deformed, be each fallen within protection scope of the present invention, this It is bright scope is claimed to be defined by claims.

Claims (4)

  1. It is 1. a kind of to deduce lattice, it is characterised in that：
    Define first：
    (1) lattice：Consider any one partial ordering set (L ,≤), if to arbitrary element a, b in set L so that a, b are in L There is an infimum and supremum, then (L ,≤) it is lattice；
    (2) rule：C1∧C2∧...∧Cn→h；
    Wherein, C1、C2、…、CnIt is condition, condition has true-false value, when for true time, condition then says that condition meets；H is conclusion, when Condition C1、C2、…、CnWhen being true, corresponding conclusion h is just drawn；
    If there is rule C1∧C2∧(Ci∨…∨Ci+n) → h, then need for the rule to split into multiple rules, respectively：C1∧C2 ∧Ci→ h ..., C1∧C2∧Ci+n→h；
    If there is rule C1∧C2∧C3→h1And h1∧C4∧C5→ h, then need to merge the two rules, is C after merging1 ∧C2∧C3∧C4∧C5→h；
    If presence takes corresponding actions according to condition, first reached a conclusion using rule, corresponding actions are carried out further according to conclusion；
    (3) knowledge base：Include the fact that storehouse and rule base；Knowledge include the fact that and rule, in the known integrated circuit it is a fact that Jing man-machine interactivelies be input into or It has been stored in factbase；The collection of strictly all rules is collectively referred to as rule base, is designated as RuleBase；
    Then lattice structure is deduced in setting：
    Deduce lattice and one lattice with hierarchical relationship is constituted by n finite element, the element in lattice is referred to as node, wherein, n>1； Node includes root node, intermediate node and leaf node；Node without forerunner is referred to as root node, and root node is unique；After not having After node be referred to as leaf node；Remaining node is referred to as intermediate node, and intermediate node is divided into two types：A kind of node is thereafter After or intermediate node, referred to as pure intermediate node, another kind of intermediate node its it is follow-up be leaf node, referred to as leaf forerunner section Point, root node and intermediate node it is corresponding be rule condition, it is the conclusion of rule that leaf node is corresponding, from root node to leaf A rule in child node rule of correspondence storehouse；
    Strictly all rules composition rule collection of the screening with common conditions C from rule base, this rule set is constituted and deduces lattice, its Root node is exactly condition C；For arbitrary non-leaf nodes Node, it is assumed that its son node number is k, respectively C1, C2..., Ck, C1It is first child node of Node, C2It is second child node ... of Node, CkIt is k-th child node of Node, TreeSet (Ci) it is with CiFor root node and CiThe node set of the tree that follow-up intermediate node is constituted, not including leaf node, CiMeet：
    C i ∉ ∪ j = i + 1 k T r e e S e t ( C j ) , i = 1 , 2 , ... , k - 1 ;
    Finally build and deduce lattice；
    The strictly all rules that condition C is met in assuming RuleBase has n bars, and the rule set RuleSet that these rules are constituted has：
    RuleName1：C∧C11∧C12∧…∧C1m1->h1
    RuleName2：C∧C21∧C22∧…∧C2m2->h2
    …………
    RuleNamen：C∧Cn1∧Cn2∧…∧Cnmn->hnmn
    Alternative condition C is used as root node, and is designated as C ' as present node, then merge public keys to remaining condition, false If there is k public keys in remaining condition, to each public keys Ci：
    (1) there is C in counting RuleSet rule setsiNumber of times be designated as Ci.Count；
    (2) include C in RuleSetiRule name constitute set be designated as Ci.RuleNames；
    (3) by Ci.Count by being ranked up from big to small, i=1 ..., k, the result of sequence are designated as C1.Count ..., Ck.Count, and to Ci.RuleNames do correspondence order to adjust, the result of adjustment is designated as C1.RuleNames ..., Ck.RuleNames；
    From the beginning of j=1, from Cj.RuleNames take out RuleName in successively and judge Cl.RulesNames whether include RuleName, such as the C including ifl.Count=Cl.Count-1, and by RuleName from Cl.RuleNames remove, l=j+ 1 ..., k, not comprising not processing；Till j=k；After so processing, C in k public keysi.Count what is be not zero is public Also have k ' individual, it is clear that k '≤k, be thus the individual disjoint set of k ' RuleSet point, each of which set it is maximum public Item is respectively C1..., Ck’, by C1..., Ck' child node in order from left to right as C '；C1..., Ck’Constitute the brotgher of node, What they were from left to right ordered into；
    Respectively with C1..., Ck’It is as present node C ', respectively using the individual disjoint sets of this k ' as rule set RuleSet, right In RuleSet, remaining condition presses above-mentioned principle circular treatment, until all of Ci.Count it is zero, so deduces lattice and just build Complete.
  2. 2. it is a kind of based on the inference method for deducing lattice, it is characterised in that to comprise the following steps：
    Step 1：User is selected with the phenomenon being actually consistent as initial condition from multiple phenomenons；
    Step 2：System is searched from rule base RuleBase and meets the strictly all rules formation rule collection of initial condition and be designated as RuleSet；
    Step 3：System deduces lattice to the strictly all rules dynamic construction in RuleSet, and using root node as present node；
    Step 4：The child node formation problem of present node in deduction lattice and above condition are false and are supplied to user by system, Select for user；
    Step 5：User selects one of them as response input system according to practical situation；If user selects condition above equal It is false, at this moment represents that system convention storehouse does not meet the rule of correlated condition, this Flow ends；
    Step 6：System carries out condition coupling to all child nodes of present node according to the input of user, and the match is successful, and adjustment is worked as Front nodal point is the child node that the match is successful, repeat step 4-6 until reaching leaf node, to obtain inference conclusion.
  3. 3. according to claim 2 based on the inference method for deducing lattice, it is characterised in that：Match described in step 6 and be exactly Condition value is true.
  4. 4. according to claim 2 based on the inference method for deducing lattice, it is characterised in that：Inference conclusion is obtained in step 6 System is made corresponding actions as needed and meets real-time control needs afterwards.
CN201610985211.4A 2016-10-25 2016-10-25 It is a kind of to deduce lattice and based on the inference method for deducing lattice Active CN106529676B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201610985211.4A CN106529676B (en) 2016-10-25 2016-10-25 It is a kind of to deduce lattice and based on the inference method for deducing lattice

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201610985211.4A CN106529676B (en) 2016-10-25 2016-10-25 It is a kind of to deduce lattice and based on the inference method for deducing lattice

Publications (2)

Publication Number Publication Date
CN106529676A true CN106529676A (en) 2017-03-22
CN106529676B CN106529676B (en) 2018-11-23

Family

ID=58350344

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201610985211.4A Active CN106529676B (en) 2016-10-25 2016-10-25 It is a kind of to deduce lattice and based on the inference method for deducing lattice

Country Status (1)

Country Link
CN (1) CN106529676B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107491484A (en) * 2017-07-17 2017-12-19 阿里巴巴集团控股有限公司 A kind of data matching method, device and equipment
CN110276453A (en) * 2019-05-29 2019-09-24 武汉大学 A method of it is made inferences using the representation of knowledge based on feature
CN110826840A (en) * 2019-08-14 2020-02-21 东航技术应用研发中心有限公司 Flight plan recovery method and system

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101710393A (en) * 2009-11-25 2010-05-19 北京航空航天大学 Method for knowledge expressing and reasoning mechanism of expert system
CN103886377A (en) * 2014-02-21 2014-06-25 北京神舟航天软件技术有限公司 inspection method based on BDD
US20140279801A1 (en) * 2013-03-15 2014-09-18 International Business Machines Corporation Interactive method to reduce the amount of tradeoff information required from decision makers in multi-attribute decision making under uncertainty
CN105302112A (en) * 2015-10-23 2016-02-03 中国电子科技集团公司第十研究所 Intelligent fault diagnosis system for ICNI system
US9373086B1 (en) * 2015-01-07 2016-06-21 International Business Machines Corporation Crowdsource reasoning process to facilitate question answering

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101710393A (en) * 2009-11-25 2010-05-19 北京航空航天大学 Method for knowledge expressing and reasoning mechanism of expert system
US20140279801A1 (en) * 2013-03-15 2014-09-18 International Business Machines Corporation Interactive method to reduce the amount of tradeoff information required from decision makers in multi-attribute decision making under uncertainty
CN103886377A (en) * 2014-02-21 2014-06-25 北京神舟航天软件技术有限公司 inspection method based on BDD
US9373086B1 (en) * 2015-01-07 2016-06-21 International Business Machines Corporation Crowdsource reasoning process to facilitate question answering
CN105302112A (en) * 2015-10-23 2016-02-03 中国电子科技集团公司第十研究所 Intelligent fault diagnosis system for ICNI system

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107491484A (en) * 2017-07-17 2017-12-19 阿里巴巴集团控股有限公司 A kind of data matching method, device and equipment
CN107491484B (en) * 2017-07-17 2020-08-28 阿里巴巴集团控股有限公司 Data matching method, device and equipment
CN110276453A (en) * 2019-05-29 2019-09-24 武汉大学 A method of it is made inferences using the representation of knowledge based on feature
CN110276453B (en) * 2019-05-29 2022-06-14 武汉大学 Method for reasoning by using knowledge representation based on characteristics
CN110826840A (en) * 2019-08-14 2020-02-21 东航技术应用研发中心有限公司 Flight plan recovery method and system
CN110826840B (en) * 2019-08-14 2020-07-17 东航技术应用研发中心有限公司 Flight plan recovery method and system

Also Published As

Publication number Publication date
CN106529676B (en) 2018-11-23

Similar Documents

Publication Publication Date Title
Gasse et al. Exact combinatorial optimization with graph convolutional neural networks
Cintra et al. Genetic generation of fuzzy systems with rule extraction using formal concept analysis
CN106529676A (en) Deductive lattice and reasoning method based on deductive lattice
Keshtkar Langaroudi et al. Sports result prediction based on machine learning and computational intelligence approaches: A survey
Ban et al. From query tools to causal architects: Harnessing large language models for advanced causal discovery from data
Abdelhalim et al. A new method for learning decision trees from rules
Bontempi et al. Recursive lazy learning for modeling and control
Castro et al. An inductive learning algorithm in fuzzy systems
Nabwey et al. An integrated methodology of rough set theory and grey system for extracting decision rules
Nápoles et al. Hybrid model based on rough sets theory and fuzzy cognitive maps for decision-making
de Campos et al. Learning bayesian networks by ant colony optimisation: searching in two different spaces
Martignon et al. Simplicity and robustness of fast and frugal heuristics
Cintra et al. FuzzyDT—a fuzzy decision tree algorithm based on C4. 5
Horvat et al. Importance of the training dataset length in basketball game outcome prediction by using naïve classification machine learning methods
Bustince et al. The origin of fuzzy extensions
Karamyan Cardinality estimation of an SQL Query using recursive neural networks
Campos et al. Learning Bayesian networks by ant colony optimisation: searching in two different spaces
Mandal On deneutrosophication
CN110276453A (en) A method of it is made inferences using the representation of knowledge based on feature
Sharma et al. An optimize decision tree algorithm based on variable precision rough set theory using degree of β-dependency and significance of attributes
Al-Helali et al. Data imputation for symbolic regression with missing values: A comparative study
Meschino et al. Using SOM as a tool for automated design of clustering systems based on fuzzy predicates
Manzali et al. A new decision tree pre-pruning method based on nodes probabilities
Bonzon et al. An empirical and axiomatic comparison of ranking-based semantics for abstract argumentation
SCARPELLI et al. High level fuzzy Petri nets and backward reasoning

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant