CN112434813A - Multiple multidimensional language reasoning method based on attribute-oriented language concept lattice - Google Patents

Multiple multidimensional language reasoning method based on attribute-oriented language concept lattice Download PDF

Info

Publication number
CN112434813A
CN112434813A CN202011209073.3A CN202011209073A CN112434813A CN 112434813 A CN112434813 A CN 112434813A CN 202011209073 A CN202011209073 A CN 202011209073A CN 112434813 A CN112434813 A CN 112434813A
Authority
CN
China
Prior art keywords
language
rule
attribute
decision
oriented
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
CN202011209073.3A
Other languages
Chinese (zh)
Other versions
CN112434813B (en
Inventor
刘新
崔慧
邹丽
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.)
Liaoning Normal University
Original Assignee
Liaoning Normal University
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 Liaoning Normal University filed Critical Liaoning Normal University
Priority to CN202011209073.3A priority Critical patent/CN112434813B/en
Publication of CN112434813A publication Critical patent/CN112434813A/en
Application granted granted Critical
Publication of CN112434813B publication Critical patent/CN112434813B/en
Expired - Fee Related legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

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
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Software Systems (AREA)
  • Data Mining & Analysis (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Computational Linguistics (AREA)
  • Mathematical Physics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Machine Translation (AREA)

Abstract

The invention discloses a multiple multidimensional language reasoning method based on attribute-oriented language concept lattices, which is carried out according to the following steps: data processing, namely expressing the relation between an object and an attribute in a form background by using a language term set; defining an approximate operator with language information, and constructing an attribute-oriented language concept lattice; researching a language decision form background, and extracting a language decision rule based on an advantage relation; establishing a multi-dimensional language reasoning model through the acquired attribute-oriented language decision rule; inputting the multi-dimensional language inference model into a neural network, and directly utilizing the neural network to carry out self-adaptive learning on a language decision rule; and calculating the weight of each node in the rule base and carrying out uncertainty reasoning on the rule with the language information.

Description

Multiple multidimensional language reasoning method based on attribute-oriented language concept lattice
Technical Field
The invention belongs to the intelligent information processing technology, and particularly relates to a multi-dimensional language reasoning method based on an attribute-oriented language concept lattice.
Background
The reasoning method is that the computer simulates the thinking mode of human beings to carry out intelligent decision making, so that the dependence on human expert experience can be reduced, the automation degree of intelligent decision making is improved, and the accuracy and the reliability of the decision making are improved.
The formal concept analysis is an inference method proposed by Wille in 1982 based on formal context analysis concept hierarchy, and the core of the method is formal concepts and concept lattices, and a data source is formal context. The formal concept is a binary group formed by the common attributes (connotation) of an object (extension) and the object, and the concept lattice completes the visualization of a graph structure among all discovered concepts in the form of super concept and sub concept hierarchy through Galois link. The method essentially describes the relationship between objects and attributes, displays the generalization and specialization relationship between concepts, and has attracted extensive attention and research in many fields such as artificial intelligence, software engineering, data mining, information retrieval and the like.
At present, scholars at home and abroad have obtained remarkable results on the research of formal concept analysis, Godin and the like propose an incremental concept forming method for updating Galois lattices and corresponding graphs, which is superior to batch processing algorithms in most cases and generates concept lattices more efficiently and simply. Blohlavek researches the theory of concept lattices from the perspective of fuzzy logic, gives out fuzzy values of objects with attributes, provides fuzzy form backgrounds and constructs fuzzy hierarchical structures. Subsequently, Yao et al combine the thought of three-branch decision, expand the concept lattice from two aspects of positive and negative operators, propose three-branch concept lattices and three-branch concept lattices derived based on attributes (objects), and analyze the relationship between the three-branch concept lattices. After rough set was proposed by z.pawlak in 1982, Yao integrated the ideas of rough set theory into formal concept analysis, building an object-oriented concept lattice. In object-oriented concept lattice research, some new advances have been made in recent years. For example, Ma et al have introduced a novel approach to obtaining object-oriented concept lattices by discussing the nature of hierarchical expansion sets. Qi and the like research the relationship between formal concept analysis and reduction on a rough set, and in order to find implicit knowledge among data more easily and show the data more simply, a new method for attribute reduction by using irreducible elements, namely an object (attribute) -oriented concept lattice reduction method is provided, and the efficiency of a concept lattice construction method is further improved.
In the processing of information in many complex systems, many types of bugs are often encounteredDeterminism describes knowledge, which may be numerical, linguistic, etc. Probabilistic reasoning, evidence reasoning, fuzzy reasoning, etc. are all effective reasoning methods for handling uncertain knowledge. To more accurately illustrate the uncertainty in estimating the age distribution of a parent population, type et al describe a new method of inferring a set of Probabilistic Models (PME) from a sample of clastic zircon. Soares et al incorporate evidence reasoning to integrate incomplete fuzzy information and incorporate top priority of ideal solution (TOPSIS) into emergency response decisions to cope with order-free vessels. In the fuzzy control problem, fuzzy reasoning is always an important theoretical basic research method, Zadeh firstly proposes a fuzzy separation algorithm (FMP) in 1973, and a composition reasoning rule (CRI) proposed by Mamdani and the like is a basic method of the fuzzy reasoning method and is widely applied at present. Chinese (Chinese) to study language truth intuitive fuzzy proposition logic L2nP (S) automatic reasoning method and language truth intuitive fuzzy proposition logic system L2nP (S). Three-segment theory of T operation and F operation is researched, and a linguistic value intuitive fuzzy reasoning method is provided. Liu and the like introduce language information into formal concept analysis, and provide a language value intuitive fuzzy concept lattice and a fuzzy language concept lattice to solve the problem of uncertainty inference, which provides a new idea for the field of formal concept analysis. In the past two years, the inference method based on machine learning has a wide research prospect. Zhou proposes a traceback learning that combines the perceptibility of the neural network with the reasoning ability of the symbolic AI, and can process both sub-symbolic data (e.g., raw pixels) and symbolic knowledge. Retrospective learning is the first framework specifically designed for simultaneous reasoning and perception, opening up a new direction for exploring AIs that approach the human level learning ability. Ristic et al covers the main methods of knowledge representation and uncertainty reasoning, especially Bayesian probability theory, belief function-based reasoning, inexact probability theory, which combines uncertainty information and the rules required for decision-making tasks under uncertainty and provides a solution for MATLAB. Yang et al propose a confidence Rule base Inference method based on an evidence Inference method (Rule-base Inference method Using the Evidense learning, RIMER) to deal with different types of uncertainty knowledge by establishing rules through confidence distribution representation. Watada et al propose a Gaussian-PSO method combining structure learning and fuzzy inference to train a neural network, effectively shortening the computation time by modifying the network structure. In order to solve fuzzy inference under the condition of multidimensional sparse rules and process the interrelation among a plurality of variables, Hassan et al proposes a multi-multidimensional fuzzy inference method based on a CMAC neural network to reduce errors of inference results.
Many researches are conducted on the problem of whether an object has an attribute in formal concept analysis, but the research on the group situation between the object and the attribute is few, especially the research on two important topics of decision of formal background and multi-dimensional reasoning, and besides, because a concept lattice still cannot process language information, information loss is easily caused.
Disclosure of Invention
The invention aims to solve the technical problems in the prior art and provides a multiple multidimensional language reasoning method based on attribute-oriented language concept lattice
The technical solution of the invention is as follows: a multi-dimensional reasoning method based on attribute-oriented language concept lattices is characterized by comprising the following steps:
step 1: data acquisition and preprocessing
Setting a language term set S ═ S which represents a language value relationship between an object and an attributelL is 0,1, …, g is even number, and has the following sequence relation s0<s1<...<sg-1<sgA language form background (U, a, S) is formed, and the object U ═ xiI | ═ 1,2, …, n }, and attribute a ═ a { (a }j|j＝1,2,…,m}，
Figure BDA0002758042460000031
I.e. S (x)i,aj)＝sl；
Step 2: constructing attribute-oriented language concept lattices
Based on the language form background (U, A, S), let ASIs defined as a set of linguistic terms SLanguage value set on attribute set A for object subset
Figure BDA0002758042460000032
And a subset of language values on the attribute
Figure BDA0002758042460000033
Defining an operator:
Figure BDA0002758042460000034
Figure BDA0002758042460000035
constructing attribute-oriented language concepts and attribute-oriented language concept lattices;
and step 3: obtaining language decision rule on attribute-oriented language concept lattice
Step 3.1 obtains language decision rules by attribute-oriented language decision concepts: the decision-making attribute is added to the system,
extending the linguistic form context (U, A, S) to a linguistic decision form context (U, A, S, C, K), where C ═ C is the decision attribute set and K is the linguistic relationship between U and C, i.e.
Figure BDA0002758042460000036
Is provided with CKDefining a set of language values on a set of decision attributes C for a set of language terms K, for a subset of objects
Figure BDA0002758042460000037
With a subset of language values on decision attribute C
Figure BDA0002758042460000038
Defining an operator:
Figure BDA0002758042460000039
constructing attribute-oriented language decision concept and attribute-oriented language decision concept lattices, obtaining a language decision rule B → D under the condition that an object is not an empty set by the attribute-oriented language decision concept (X, B, D), and recording all the language decision rules into omega (U);
step 3.2, obtaining an extension rule: for the benefit-type consistent rule, that is, the rule with the larger language value of the condition attribute, the larger language value of the decision attribute value, if the following rule antecedents appear, the more language decision rules are updated according to the dominance relationship among the rule antecedents as follows:
4) when in use
Figure BDA0002758042460000041
The language decision result corresponding to the back-part of the rule is
Figure BDA0002758042460000042
5) When in use
Figure BDA0002758042460000043
And is
Figure BDA0002758042460000044
The language corresponding to the back-part of the rule
The decision result is
Figure BDA0002758042460000045
6) When in use
Figure BDA0002758042460000046
The language decision result corresponding to the back-part of the rule is
Figure BDA0002758042460000047
Wherein,
Figure BDA0002758042460000048
is a rear part of
Figure BDA0002758042460000049
The front piece of (a) is,
Figure BDA00027580424600000410
is a rear part of
Figure BDA00027580424600000411
The smallest of the front-piece parts of (a),
Figure BDA00027580424600000412
is a rear part of
Figure BDA00027580424600000413
Is the largest front piece of
Figure BDA00027580424600000414
That is, if a rule front part is greater than or equal to a back part
Figure BDA00027580424600000415
The rule front piece of (2), then the back piece corresponding to the rule is also
Figure BDA00027580424600000416
If the front part is located at the rear part according to a certain rule
Figure BDA00027580424600000417
Between the largest and smallest front piece, the rear piece corresponding to the rule is
Figure BDA00027580424600000418
If the front part is less than or equal to the back part of a rule
Figure BDA00027580424600000419
The rule front piece of (2), then the back piece corresponding to the rule is also
Figure BDA00027580424600000420
For the situation that the languages in the front piece are not comparable, selecting a plurality of maximum values and minimum values, and extracting language decision rules respectively according to the method in the step 3.2;
step 3.3, eliminating unreasonable rules under the background of big data: comparing all rules acquired by the attribute-oriented language decision concept and the extended rule with the original rules under the background of the language decision form, and if the front part of the rule is larger than the original rules and the back part of the rule is small, determining that the rule is an unreasonable rule and rejecting the rule; if the rule back piece is smaller than the original rule and the back piece is large, the rule is regarded as an unreasonable rule and is rejected; the updating rule is counted in omega (U);
step 4, constructing a multi-dimensional language reasoning model and reasoning the known conditions
Step 4.1 for the known fact with unknown result, the known attribute-oriented language decision rule is used to establish a multi-dimensional language inference model:
Figure BDA00027580424600000421
Figure BDA0002758042460000051
wherein B is11,B12,...,B1m,B21,B22,...,B2m,...,Bv1,Bv2,...,BvmIs a linguistic value, D, describing the property of a condition in a antecedent1,D2,...,DvIs a linguistic value that describes the decision attribute in the back-piece,
Figure BDA0002758042460000052
Figure BDA0002758042460000053
is the language value, D, corresponding to the input condition attribute*Is the language value corresponding to the output decision attribute;
step 4.2 reasoning with neural networks
Inputting the v language decision rules in the multi-dimensional language reasoning model into a BP neural network, continuously correcting each condition attribute and the weight of each corresponding node through reverse error propagation, and utilizing gradientThe descent method minimizes the error value and finally inputs the known fact
Figure BDA0002758042460000054
Outputting D through a neural network*And outputting different prediction results according to different conditions by corresponding language values.
The attribute-oriented language decision rule provided based on the advantage relation successfully expands the language decision rule set, reflects wider language decision knowledge, applies the rule obtained by the attribute-oriented language concept lattice to the inference problem of uncertain knowledge, has certain accuracy and persuasion, and better accords with the actual problem. A multi-dimensional language reasoning model is established, objective node weight is given through a neural network, and reasonable reasoning on a language decision rule is achieved. The language value is added into the reasoning problem, so that the method is more consistent with the normal thinking of human beings, avoids information loss and improves the accuracy of a prediction result.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention.
FIG. 2 is a diagram of a concept lattice of an attribute-oriented language according to an embodiment of the present invention.
FIG. 3 is a diagram of a concept lattice for attribute-oriented language decision according to an embodiment of the present invention.
Detailed Description
The embodiment of the invention applies the method to language programming ability prediction of college students and deduces the final evaluation result of the whole language programming level of the students and laboratories, and the flow is shown as the figure 1.
Step 1: data acquisition and preprocessing
Setting a language term set S ═ S which represents a language value relationship between an object and an attributelL is 0,1, …, g is even number, and has the following sequence relation s0<s1<...<sg-1<sgA language form background (U, a, S) is formed, and the object U ═ xiI | ═ 1,2, …, n }, and attribute a ═ a { (a }j|j＝1,2,…,m}，
Figure BDA0002758042460000055
I.e. S (x)i,aj)＝sl。
E.g. seven students of the object set U ═ { x1,x2,x3,x4,x5,x6,x7And the four language programming software level of the attribute set a ═ a }1Python level, a2C language level, a3C + + level, a4When the evaluation is performed in the JAVA level, a language term set g indicating a language value relationship between an object and an attribute is set to 4, and S is { S ═ S }0Failing grid, s1Passing, s2Medium, s3Good as s4Excellence } and has the following order relation s0<s1<...<s3<s4Forming a decision form background (U, A, S) for the student language programming learning situation and
Figure BDA0002758042460000061
the details are shown in Table 1.
TABLE 1
Figure BDA0002758042460000062
Step 2: constructing attribute-oriented language concept lattices
Based on the language form background (U, A, S), let ASDefining a set of language values on a set of attributes A for a set of language terms S, for a subset of objects
Figure BDA0002758042460000063
And a subset of language values on the attribute
Figure BDA0002758042460000064
Defining an operator:
Figure BDA0002758042460000065
Figure BDA0002758042460000066
the attribute-oriented language concept and the attribute-oriented language concept lattice are constructed, and the results are shown in table 2 and fig. 2.
TABLE 2
Figure BDA0002758042460000067
Figure BDA0002758042460000071
And step 3: obtaining language decision rule on attribute-oriented language concept lattice
Step 3.1 obtains language decision rules by attribute-oriented language decision concepts: the decision-making attribute is added to the system,
extending the linguistic form context (U, a, S) to a linguistic decision form context (U, a, S, C, K), see table 3, where C ═ C is the decision attribute set and K is the linguistic relationship between U and C, i.e.
Figure BDA0002758042460000072
Is provided with CKDefining a set of language values on a set of decision attributes C for a set of language terms K, for a subset of objects
Figure BDA0002758042460000073
With a subset of language values on decision attribute C
Figure BDA0002758042460000074
Defining an operator:
Figure BDA0002758042460000075
the attribute-oriented language decision concept and attribute-oriented language decision concept lattice are constructed, and the results are shown in table 4 and fig. 3. The language decision rule B → D under the condition that the object is not an empty set is obtained from the attribute-oriented language decision concept (X, B, D), and all the language decision rules are recorded into omega (U), see Table 5.
TABLE 3
Figure BDA0002758042460000076
TABLE 4
Figure BDA0002758042460000077
Figure BDA0002758042460000081
TABLE 5
Figure BDA0002758042460000082
Step 3.2, obtaining an extension rule: for the benefit-type consistent rule, that is, the rule with the larger language value of the condition attribute, the larger language value of the decision attribute value, if the following rule antecedents appear, the more language decision rules are updated according to the dominance relationship among the rule antecedents as follows:
7) when in use
Figure BDA0002758042460000083
The language decision result corresponding to the back-part of the rule is
Figure BDA0002758042460000084
8) When in use
Figure BDA0002758042460000085
And is
Figure BDA0002758042460000086
The language decision result corresponding to the back-part of the rule is
Figure BDA0002758042460000087
9) When in use
Figure BDA0002758042460000088
The language decision result corresponding to the back-part of the rule is
Figure BDA0002758042460000089
Wherein,
Figure BDA00027580424600000810
is a rear part of
Figure BDA00027580424600000811
The front piece of (a) is,
Figure BDA00027580424600000812
is a rear part of
Figure BDA00027580424600000813
The smallest of the front-piece parts of (a),
Figure BDA00027580424600000814
is a rear part of
Figure BDA0002758042460000091
Is the largest front piece of
Figure BDA0002758042460000092
That is, if a rule front part is greater than or equal to a back part
Figure BDA0002758042460000093
The rule front piece of (2), then the back piece corresponding to the rule is also
Figure BDA0002758042460000094
If the front part is located at the rear part according to a certain rule
Figure BDA0002758042460000095
Between the largest and smallest front piece, the rear piece corresponding to the rule is
Figure BDA0002758042460000096
If the front part is less than or equal to the back part of a rule
Figure BDA0002758042460000097
The rule front piece of (2), then the back piece corresponding to the rule is also
Figure BDA0002758042460000098
For the situation that the languages in the front piece are not comparable, selecting a plurality of maximum values and minimum values, and extracting language decision rules respectively according to the method in the step 3.2;
step 3.3, eliminating unreasonable rules under the background of big data: comparing all rules acquired by the attribute-oriented language decision concept and the extended rule with the original rules under the background of the language decision form, and if the front part of the rule is larger than the original rules and the back part of the rule is small, determining that the rule is an unreasonable rule and rejecting the rule; if the rule back piece is smaller than the original rule and the back piece is large, the rule is regarded as an unreasonable rule and is rejected; the update rule is included in Ω (U), and the update rule is included in Ω (U), as shown in Table 6.
TABLE 6
Figure BDA0002758042460000099
Step 4, constructing a multi-dimensional language reasoning model and reasoning the known conditions
Step 4.1 for the known fact with unknown result, the known attribute-oriented language decision rule is used to establish a multi-dimensional language inference model:
Figure BDA0002758042460000101
wherein B is11,B12,...,B1m,B21,B22,...,B2m,...,Bv1,Bv2,...,BvmIs a linguistic value, D, describing the property of a condition in a antecedent1,D2,...,DvIs to describe decision genus in the back-partThe value of the language of the sex is,
Figure BDA0002758042460000102
Figure BDA0002758042460000103
is the language value, D, corresponding to the input condition attribute*Is the language value corresponding to the output decision attribute;
the method comprises the following specific steps: if each language programming level is a1---s3，a2---s2，a3---s3，a4---s1To predict the comprehensive capability level, the following multi-dimensional language inference model can be built according to the above 24 attribute-oriented language decision rules:
Figure BDA0002758042460000104
step 4.2 reasoning with neural networks
Inputting the multi-dimensional language reasoning model formed by the 24 language decision rules into a BP neural network, continuously correcting each condition attribute and the weight of each corresponding node through reverse error propagation, minimizing the error value by using a gradient descent method, and finally inputting the known fact s3,s2,s3,s1Output D via a neural network*Corresponding to a linguistic value of s2That is, if the level of Python of a student or a laboratory is good, the level of C language is medium, the level of C + + is good, and the level of JAVA is passing, the comprehensive level of language programming ability of the student or the laboratory is medium.

Claims (1)

1. A multi-dimensional reasoning method based on attribute-oriented language concept lattices is characterized by comprising the following steps:
step 1: data acquisition and preprocessing
Setting a language term set S ═ S which represents a language value relationship between an object and an attributelL is 0,1, …, g is even and has the same form asSequence relation of0<s1<...<sg-1<sgA language form background (U, a, S) is formed, and the object U ═ xiI | ═ 1,2, …, n }, and attribute a ═ a { (a }j|j＝1,2,…,m}，
Figure FDA0002758042450000011
I.e. S (x)i,aj)＝sl；
Step 2: constructing attribute-oriented language concept lattices
Based on the language form background (U, A, S), let ASDefining a set of language values on a set of attributes A for a set of language terms S, for a subset of objects
Figure FDA0002758042450000012
And a subset of language values on the attribute
Figure FDA0002758042450000013
Defining an operator:
Figure FDA0002758042450000014
Figure FDA0002758042450000015
constructing attribute-oriented language concepts and attribute-oriented language concept lattices;
and step 3: obtaining language decision rule on attribute-oriented language concept lattice
Step 3.1 obtains language decision rules by attribute-oriented language decision concepts: the decision-making attribute is added to the system,
extending the linguistic form context (U, A, S) to a linguistic decision form context (U, A, S, C, K), where C ═ C is the decision attribute set and K is the linguistic relationship between U and C, i.e.
Figure FDA0002758042450000016
Is provided with CKDefining a set of language values on a set of decision attributes C for a set of language terms K, for a subset of objects
Figure FDA0002758042450000017
With a subset of language values on decision attribute C
Figure FDA0002758042450000018
Defining an operator:
Figure FDA0002758042450000019
constructing attribute-oriented language decision concept and attribute-oriented language decision concept lattices, obtaining a language decision rule B → D under the condition that an object is not an empty set by the attribute-oriented language decision concept (X, B, D), and recording all the language decision rules into omega (U);
step 3.2, obtaining an extension rule: for the benefit-type consistent rule, that is, the rule with the larger language value of the condition attribute, the larger language value of the decision attribute value, if the following rule antecedents appear, the more language decision rules are updated according to the dominance relationship among the rule antecedents as follows:
1) when in use
Figure FDA0002758042450000021
The language decision result corresponding to the back-part of the rule is
Figure FDA0002758042450000022
2) When in use
Figure FDA0002758042450000023
And is
Figure FDA0002758042450000024
The language decision result corresponding to the back-part of the rule is
Figure FDA0002758042450000025
3) When in use
Figure FDA0002758042450000026
The language decision result corresponding to the back-part of the rule is
Figure FDA0002758042450000027
Wherein,
Figure FDA0002758042450000028
is a rear part of
Figure FDA0002758042450000029
The front piece of (a) is,
Figure FDA00027580424500000210
is a rear part of
Figure FDA00027580424500000211
The smallest of the front-piece parts of (a),
Figure FDA00027580424500000212
is a rear part of
Figure FDA00027580424500000213
Is the largest front piece of
Figure FDA00027580424500000214
That is, if a rule front part is greater than or equal to a back part
Figure FDA00027580424500000215
The rule front piece of (2), then the back piece corresponding to the rule is also
Figure FDA00027580424500000216
If the front part is located at the rear part according to a certain rule
Figure FDA00027580424500000217
Between the largest and smallest front piece, the rear piece corresponding to the rule is
Figure FDA00027580424500000218
If the front part is less than or equal to the back part of a rule
Figure FDA00027580424500000219
The rule front piece of (2), then the back piece corresponding to the rule is also
Figure FDA00027580424500000220
For the situation that the languages in the front piece are not comparable, selecting a plurality of maximum values and minimum values, and extracting language decision rules respectively according to the method in the step 3.2;
step 3.3, eliminating unreasonable rules under the background of big data: comparing all rules acquired by the attribute-oriented language decision concept and the extended rule with the original rules under the background of the language decision form, and if the front part of the rule is larger than the original rules and the back part of the rule is small, determining that the rule is an unreasonable rule and rejecting the rule; if the rule back piece is smaller than the original rule and the back piece is large, the rule is regarded as an unreasonable rule and is rejected; the updating rule is counted in omega (U);
step 4, constructing a multi-dimensional language reasoning model and reasoning the known conditions
Step 4.1 for the known fact with unknown result, the known attribute-oriented language decision rule is used to establish a multi-dimensional language inference model:
Figure FDA00027580424500000221
wherein B is11,B12,...,B1m,B21,B22,...,B2m,...,Bv1,Bv2,...,BvmIs a linguistic value, D, describing the property of a condition in a antecedent1,D2,...,DvIs a linguistic value that describes the decision attribute in the back-piece,
Figure FDA0002758042450000031
is the language value, D, corresponding to the input condition attribute*Is the language value corresponding to the output decision attribute;
step 4.2 reasoning with neural networks
Inputting the v language decision rules in the multi-dimensional language reasoning model into a BP neural network, continuously correcting each condition attribute and the weight of each corresponding node through reverse error propagation, minimizing the error value by using a gradient descent method, and finally inputting the known fact
Figure FDA0002758042450000032
Outputting D through a neural network*And outputting different prediction results according to different conditions by corresponding language values.
CN202011209073.3A 2020-11-03 2020-11-03 Multi-multidimensional language reasoning method based on attribute-oriented language concept lattice Expired - Fee Related CN112434813B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011209073.3A CN112434813B (en) 2020-11-03 2020-11-03 Multi-multidimensional language reasoning method based on attribute-oriented language concept lattice

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011209073.3A CN112434813B (en) 2020-11-03 2020-11-03 Multi-multidimensional language reasoning method based on attribute-oriented language concept lattice

Publications (2)

Publication Number Publication Date
CN112434813A true CN112434813A (en) 2021-03-02
CN112434813B CN112434813B (en) 2023-07-11

Family

ID=74695229

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011209073.3A Expired - Fee Related CN112434813B (en) 2020-11-03 2020-11-03 Multi-multidimensional language reasoning method based on attribute-oriented language concept lattice

Country Status (1)

Country Link
CN (1) CN112434813B (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113539375A (en) * 2021-07-16 2021-10-22 河北大学 Biological information class extraction method based on three-branch semi-concept
CN113553399A (en) * 2021-07-16 2021-10-26 山东建筑大学 Text search method and system based on fuzzy language approximate concept lattice
CN114049956A (en) * 2022-01-12 2022-02-15 山东建筑大学 Disease diagnosis prediction system for deriving fuzzy object language concept lattice based on attribute

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150019462A1 (en) * 2013-07-15 2015-01-15 Senscio Systems Systems and methods for semantic reasoning
US20160078345A1 (en) * 2014-09-12 2016-03-17 Chevron U.S.A. Inc. Linguistic Goal Oriented Decision Making
CN107578106A (en) * 2017-09-18 2018-01-12 中国科学技术大学 A kind of neutral net natural language inference method for merging semanteme of word knowledge
CN111597217A (en) * 2020-05-07 2020-08-28 辽宁师范大学 Personalized recommendation method based on fuzzy object language concept lattice

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150019462A1 (en) * 2013-07-15 2015-01-15 Senscio Systems Systems and methods for semantic reasoning
US20160078345A1 (en) * 2014-09-12 2016-03-17 Chevron U.S.A. Inc. Linguistic Goal Oriented Decision Making
CN107578106A (en) * 2017-09-18 2018-01-12 中国科学技术大学 A kind of neutral net natural language inference method for merging semanteme of word knowledge
CN111597217A (en) * 2020-05-07 2020-08-28 辽宁师范大学 Personalized recommendation method based on fuzzy object language concept lattice

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
崔慧;曹仪铭;罗思元;邹丽;: "模糊语言概念格及其规则提取", 高师理科学刊, no. 03 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113539375A (en) * 2021-07-16 2021-10-22 河北大学 Biological information class extraction method based on three-branch semi-concept
CN113553399A (en) * 2021-07-16 2021-10-26 山东建筑大学 Text search method and system based on fuzzy language approximate concept lattice
CN113539375B (en) * 2021-07-16 2022-03-15 河北大学 Biological information class extraction method based on three-branch semi-concept
CN113553399B (en) * 2021-07-16 2022-05-27 山东建筑大学 Text search method and system based on fuzzy language approximate concept lattice
CN114049956A (en) * 2022-01-12 2022-02-15 山东建筑大学 Disease diagnosis prediction system for deriving fuzzy object language concept lattice based on attribute
CN114049956B (en) * 2022-01-12 2022-06-24 山东建筑大学 Disease diagnosis prediction system for deriving fuzzy object language concept lattice based on attribute

Also Published As

Publication number Publication date
CN112434813B (en) 2023-07-11

Similar Documents

Publication Publication Date Title
CN112434813A (en) Multiple multidimensional language reasoning method based on attribute-oriented language concept lattice
Konar et al. Reasoning and unsupervised learning in a fuzzy cognitive map
US11651216B2 (en) Automatic XAI (autoXAI) with evolutionary NAS techniques and model discovery and refinement
Sharma Designing and modeling fuzzy control Systems
Tan et al. Application of genetic algorithm for fuzzy rules optimization on semi expert judgment automation using Pittsburg approach
Cui et al. Multiple multidimensional linguistic reasoning algorithm based on property-oriented linguistic concept lattice
Gegov et al. Rule base simplification in fuzzy systems by aggregation of inconsistent rules
CN112700099A (en) Resource scheduling planning method based on reinforcement learning and operation research
Cao et al. Fast and explainable warm-start point learning for AC Optimal Power Flow using decision tree
CN111460275A (en) Social network-oriented dynamic network representation learning method and system
Zhang et al. BRN: A belief rule network model for the health evaluation of complex systems
Chernyshev et al. Integration of building information modeling and artificial intelligence systems to create a digital twin of the construction site
Razak et al. Decomposing Conventional Fuzzy Logic Systems to Hierarchical Fuzzy Systems
CN112862211A (en) Method and device for assigning orders of dynamic ring defects of communication management system
Zhang et al. Construction and application of Bayesian networks in flood decision supporting system
CN117474522A (en) Power grid substation equipment operation and detection auxiliary decision-making method based on natural language reasoning
Balazs et al. Hierarchical-interpolative fuzzy system construction by genetic and bacterial memetic programming approaches
CN115440387A (en) Artificial intelligence-based resident respiratory infectious disease monitoring and early warning system and method
CN115563225A (en) Power grid equipment fault diagnosis method and system based on knowledge graph relation reasoning
Ren et al. Fuzzy multi-state fault tree analysis based on fuzzy expert system
Chen et al. An Attribute Reduction Algorithm Based on Rough Set Theory and an Improved Genetic Algorithm.
CN111950691A (en) Reinforced learning strategy learning method based on potential action representation space
Fryc et al. Balanced Fuzzy Petri Nets
Kurochkin et al. Fuzzy logic inference ruleset augmentation with sample data in medical decision-making systems
Zhao et al. Fuzzy rule learning during simulation of manufacturing resources

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant
CF01 Termination of patent right due to non-payment of annual fee

Granted publication date: 20230711