CN101593109A - A kind of satisfiability of targets is qualitative, quantificational expression and inference method - Google Patents

Abstract

A kind of being applied under the software systems engineering environment, relating to object module represents and target reasoning field, based on the satisfiability of targets representation model of cloud model theory with based on the satisfiability of targets inference method of OWA operator, be mainly used in the integrating representation problem and satisfiability of targets reasoning and the The reasoning results rationality problem thereof that solve the satisfiability of targets qualitative, quantitative.Technical essential is as follows: (1) uses a five-tuple to represent target based on cloud model objective definition satisfiability representation model, makes target have qualitative and quantitative satisfiability value simultaneously; (2), obtain the evidence source that father's target is decomposed the context satisfiability by these two kinds of operators based on the contextual AND operator of exploded, OR operator in the OWA operator definitions target reasoning process; (3) use the mode of giving weight to reflect the user for the different preference of target, the different evidences source that utilizes the ENV operator in the inference method to synthesize satisfiability of targets in the evidence context; (4) show complete object construction figure after the reasoning terminal procedure, each target all comprises qualitative and quantitative satisfiability value, and it is cognitive more intuitively to make that the user has target.Advantage of the present invention is the satisfiability of targets representation model, can provide satisfiability is carried out the ability that fine granularity is distinguished when keep the semantic understanding ability; Inference method can avoid based on pure logical, " or " the too The reasoning results of " bigoted "; Simultaneously, the satisfaction degree of father's target reacts human general thought characteristic preferably between the minimum and maximal value of sub-goal satisfiability.

Description

A kind of satisfiability of targets is qualitative, quantificational expression and inference method

Technical field

The present invention relates under the software systems engineering environment, based target model representation and target reasoning field particularly, can be expressed as a kind of satisfiability of targets representation model and a kind of satisfiability of targets inference method based on the OWA operator based on cloud model.

Background technology

Object-oriented requirement engineering (Goal-Oriented Requirements Engineering:GORE) is the important research branch of requirement engineering, target is a sign, a kind of logic mechanism of tissue and verifying software demand, target is by commerce, the high-rise purpose (Objectives) of tissue or system is formed, catch the reason that software systems are fabricated, and the design decision of different levels instructed. object-oriented demand modeling, from the higher level of abstraction target, according to vertical And between target, the Or exploded relationship, abstract object is successively refined, leaf target up to low abstraction hierarchy, the leaf target can be used as the software allocation of duties and gives different Agent and realization. on the basis of exploded relationship, can increase horizontal binary relation Satisfy and Deny simultaneously, be respectively applied for the forward of expressing between target, negative sense influence relation. according to relationship by objective (RBO), given leaf satisfiability of targets initial value, according to satisfiability inference method from bottom to top, can derive and draw the satisfaction degree of high-rise target. requirement engineering teacher and software engineer are according to the satisfiability of target, detect the potential conflict of demand model, and optional software design scheme weighed, finish every design decision of software systems.

Satisfiability represents it is the important prerequisite of design satisfiability inference method, satisfiability can be expressed as the quantitative value in continuous number field interval, or one group of qualitative rank. quantificational expression can satisfy the requirement of expression precision, but introduced the ambiquity of subjective cognition. for example, 0.65 essential implication what is? is the degree that target is satisfied a height or low? semantic being difficult to usually of notion explained with simple numerical value. for the satisfaction degree of one group of leaf target, be difficult to represent by qualitativing concept or quantitative value uniformly. for example, part leaf satisfiability of targets can't accurate Calculation draw, and usually according to user or domain expert's subjective experience and analysis result, with such as " general satisfy ", the mode of language values such as " very satisfied " is expressed. and qualitative representation can be distinguished the semantic difference between the satisfiability of different stage, yet be easy to understand the degree that target is satisfied in notion and cognitive level., qualitative representation but brings the loss of expression precision, can not carry out fine granularity ground relatively to satisfiability, also being unfavorable for the application of quantitative analysis method. object module is as demand analysis, consult, and software design, realize the important carrier of decision-making, software relates to many often hope and can analyze and estimate the satisfaction degree of target with different granularities. and current, take relatively independent satisfiability quantitatively or qualitatively to represent mode in the GORE research field, can't take into account the accuracy of quantificational expression simultaneously, and the definite semantic of qualitative representation.

The satisfiability of targets reasoning, it is the semanteme of representing mode and relationship by objective (RBO) according to satisfiability of targets, finish satisfiability transmission or synthetic process between different target. existing document is pointed out from the angle of the adequate condition of judging father's target satisfaction degree, semanteme to And, Or exploded relationship makes an explanation, in the And exploded relationship, the satisfied of all sub-goals is the adequate condition of judging that father's target satisfies; Concern for Or, satisfying of arbitrary sub-goal all can be used as the adequate condition of passing judgment on father's target satisfaction degree. and this explanation does not clearly provide the general principle that realizes the satisfiability reasoning under the exploded relationship, the result who causes is different researchers according to separately to the understanding of relationship by objective (RBO) semanteme, design different inference methods. for example use pure logical, OR operation is finished qualitative reasoning, only pay close attention to the influence of sub-goal satisfiability minimum or maximal value to father's target. this method is not considered " compensation " or " inhibition " effect between sub-goal, for example, target g is decomposed into one group of sub-goal g1 by And, ..., gn, and except that g1 is " general satisfied ", other sub-goals are " satisfied fully ". at this moment, the satisfaction degree of g should only not determined separately by g1, suitably consider " compensation " effect of other sub-goal to the satisfaction degree of g, just can make rationally, judge all sidedly. adopt for another example under the probability meaning " with ", OR operation is finished quantitative reasoning, yet the satisfaction degree of father's target is determined jointly by the probability that satisfies of all sub-goals., inference method based on probability, the satisfaction degree of father's target may much smaller than or greater than the minimum or the maximal value of sub-goal satisfiability. in fact, relate to the crowd as software with subjective judgement ability, be not with pure logical OR based on probability " with ", " or " the mode of thinking, the The reasoning results of cognitive father's satisfiability of targets. generally speaking, " compensation " or " inhibition " effect between sub-goal with due regard to, make the satisfaction degree of father's target between the minimum and maximal value of sub-goal satisfiability, more meet the characteristics of the human general mode of thinking.

Therefore, many disadvantages at above-mentioned appearance, need a kind of satisfiability representation model of reasonable target more and based on the satisfiability of targets inference method of this model, can qualitative, quantitative ground the satisfiability of expression object module, and, can make the result of reasoning reflect user's cognitive thinking better based on the reasoning process that this representation model carries out.

Summary of the invention

At above-mentioned satisfiability of targets represent with inference method in the limitation that exists, the present invention designs a kind of satisfiability of targets representation model based on cloud model, finishes the unified expression of satisfiability of targets qualitative, that quantitatively combine.On this model based, a kind of satisfiability inference method based on OWA operator core concept is proposed.This method considers that simultaneously " compensation " and " inhibition " effect is to the influence of father's target satisfaction degree between sub-goal, and The reasoning results can better reflect people's subjective thought characteristic.

Concrete grammar of the present invention is as follows:

Step 1:

Give the qualitative rank of satisfiability of targets representation model.

Step 2:

Give the satisfiability value of leaf node.

Step 3:

Definition is based on the satisfiability of targets representation model of cloud model, and based on this model the leaf target among the target figure carried out pre-service, makes each leaf node have qualitative and quantitative satisfiability value simultaneously.

Step 4:

Traversal is sought the father node that does not carry out reasoning, if exist, enters step 5; If do not exist, enter step 7.

Step 5:

AND operator, OR operator based in the core concept definition inference method of OWA algorithm carry out reasoning to father's target satisfaction degree in the exploded context, form an evidence source of father's satisfiability of targets.

Step 6:

Horizontal relationship operator (ENV) in the definition inference method, the different evidences source of satisfiability of targets in the synthetic evidence context.

Step 7:

Handle father's destination node based on the satisfiability of targets representation model that defines in the step 2, make all father's destination nodes also have qualitative and quantitative satisfiability value simultaneously.

Step 8:

Finish reasoning process, and complete display-object structural drawing.

Step 9:

Satisfiability of targets is represented and reasoning process finishes.

Description of drawings:

Fig. 1: prototype system structural representation

Fig. 2: satisfiability of targets reasoning process flow diagram

Embodiment

Understand and enforcement the present invention the existing embodiments of the invention of describing in conjunction with the accompanying drawings for ease of persons skilled in the art.

As shown in Figure 1, for implementing the prototype system structural representation of method described in the present invention, by at least one user input apparatus 1, at least one object module satisfiability is represented and reasoning center 2, and at least one object construction map display device 3 is formed.The function that user input apparatus 1 is realized is to give qualitative rank and to the satisfiability assignment of leaf destination node, assignment not only is confined to quantitative value, also can qualitative other qualitative value of level to the satisfiability of targets representation model.User input apparatus 1 can represent with the satisfiability of targets representation model and reasoning center 2 in qualitative, the quantitative conversion equipment 201 of satisfiability of targets exchange, give qualitative, quantitative conversion equipment to handle the satisfiability value of leaf node, purpose is that the satisfiability value of all leaf nodes before reasoning is carried out has qualitative, quantitative values simultaneously, quantitative values is beneficial to the carrying out of reasoning process, and qualitative value is beneficial to the overall cognition of user.Device 201 carries out data transmission with satisfiability reasoning arithmetic unit 202, finishes the reasoning calculating process of all father's targets in 202 devices; Install 202 simultaneously and transmit the reasoning data with qualitative, the quantitative conversion equipment 201 of satisfiability of targets once more, processing through device 201, father node also has qualitative, quantitative satisfiability value simultaneously, at last the satisfiability value (qualitative value, quantitative values) of all nodes is given object construction map display device 3, clearly demonstrate object construction figure, in structural drawing, comprise each node relation and satisfiability value of qualitative, quantitative separately each other.

As shown in Figure 2, existing in conjunction with prototype system structural representation shown in Figure 1 for implementing the object representation and the reasoning process flow diagram of method described in the present invention, concrete implementation step is described in detail as follows:

Step 1:

The user gives the qualitative rank of satisfiability of targets representation model by user input apparatus 1, qualitative rank comprises the qualitativing concept number of setting for object module, and three numerical characteristic Ex, En, the He of the cloud of each qualitativing concept correspondence, its essence is exactly to utilize cloud model that qualitativing concept and one group of quantitative value are linked together, and can change mutually in needs.

Step 2:

The user gives the satisfiability value of leaf node by input media 1, the user can be according to the preference of oneself, both can give quantitative satisfiability numerical value, as 0.5,0.8 etc., also can give satisfiability value qualitatively, for example generally satisfy, satisfy very much etc., but will be limited in the qualitative level range of giving in the step 1 when giving qualitative value, can not fabricate qualitative value without foundation.If leaf node satisfiability numerical value is quantitative, then can be directly used in reasoning process; If qualitatively, then can utilize qualitative, the quantitative conversion equipment of satisfiability of targets in the step 3 that qualitative value is converted to quantitative values, carry out the process of reasoning again.

Step 3:

The leaf node satisfiability value of in qualitative, the quantitative conversion equipment 201 of satisfiability of targets step 2 being given is carried out pre-service, makes leaf node both comprise quantitative numerical value, also has satisfiability value qualitatively.Quantitative numerical value is used for follow-up reasoning calculating process; Qualitative value can allow the user that destination node is had the understanding of a macroscopic view, and this is because the subjectivity of people's thinking, and numerical value 0.5 satisfies with general, and the user can more have a preference for and generally satisfy this expression way.The definition and the preprocessing process thereof of objectives satisfiability representation model are as follows:

Satisfiability of targets model based on cloud: CSat=＜L, X, μ, CFLX, CFXL 〉. wherein, L is that X is the corresponding sub-domain of L by the qualitative rank of the numerical characteristic sign of cloud $U_L\⋐U$ Interior arbitrary satisfiability quantitative values, μ is degree of certainty .CFLX:L * μ → X that X is under the jurisdiction of L, CFXL:X → L * μ is the mapping function between L and X, is used to realize the mutual conversion of the two.

Preprocessing process is: handle the satisfiability value of leaf node, make leaf node have quantitative satisfiability numerical value, can carry out the reasoning calculating process; Simultaneously also have other satisfiability value of qualitativing concept level, help generally cognition leaf node.

Can take into account the definite semantic and the accuracy of satisfiability based on the CSat model of cloud model, can integrate the ambiguity of cognition and randomness, better realize qualitative, the quantitative unified expression of satisfiability, for the satisfiability reasoning provides rational semantic basis and accuracy representing.

Step 4:

Traversal object construction figure seeks whether there is the father node that does not carry out reasoning, if exist, enters step 5, if do not exist, enters step 7.

Step 5:

Utilize AND operator in the reasoning algorithm, OR operator, the father node to not reasoning in device 202 carries out the contextual reasoning of exploded.If father node exploded relationship correspondence is the AND relation, then use the AND operator; If father node exploded relationship correspondence is the OR relation, then use the OR operator; Finally obtain the evidence source that father's target local context concerns satisfiability.

Specifically as follows based on AND, the OR operator calculating process of OWA:

Given one group of real number A=(a ₁..., a _n) use the OWA operator general step as follows:

Step 1: to (a ₁..., a _n) rearrangement, generate new real number set B=(b ₁..., b _n)

Step 2: according to set (b ₁..., b _n), generate weight set W=(w ₁..., w _n)

Step 3: calculate F (b ₁..., b _nThe polymerization result of)=WT * B, wherein WT is the transposition of W

The formula of AND operator, the concrete definition of OR operator is as follows:

$F_{\mathrm{and}}(g_1,g_2,...,g_n)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\ω}}_i\×g_i,$ And (ω _i∈ [0,1])

$F_{\mathrm{or}}(g_1,g_2,...,g_n)=\underset{i=1}{\overset{n}{\mathrm{\Σ}}}{\mathrm{\ω}}_i\×g_i,$ And (ω _i∈ [0,1])

For unified statement and analysis F _AndAnd F _OrOperator, the present invention's employing mode from small to large is to (a ₁..., a _n) rearrangement, promptly for any b _i, b _j∈ B when satisfying i＜j, must have b _i≤ b _jThe OWA operator is controlled pair set (b by the adjustment to weight parameter ₁..., b _n) carry out polymerization " with ", " or " degree, work as w ₁≤ w ₂≤ ...≤w _nThe time, orness (W) 〉=0.5; Work as w ₁〉=w ₂〉=... 〉=w _nThe time, andness (W) 〉=0.5.As seen, when giving big weight for bigger numerical, the OWA operator becomes the or-like operator; When giving big weight than fractional value, the OWA operator becomes the and-like operator.Therefore, given And＜g, { g ₁..., g _n＞or Or＜g, { g ₁..., g _n}＞, F _AndAnd F _OrThe operator weights are determined under the method: given one group of satisfiability of targets (x ₁..., x _n), x wherein _i∈ [0,1], and for any i and j, when satisfying i＜j, x _i≤ x _jAt this moment, for arbitrary x _i, have function f (x _i) 〉=0, and w _iDetermine by following formula:

${\mathrm{\ω}}_i=\frac{f\left(x_i\right)}{{\mathrm{\Σ}}_{j=1}^{n}f\left(x_j\right)}$

Determine the mentioned above principle of and-like and or-like operator according to OWA, and the semantic interpretation of And, Or exploded relationship, F _AndAnd F _OrF (the x of operator _i) shown in the following formula of function:

$f_{\mathrm{and}}\left(x_i\right)=\left(\begin{array}{cc}1,& x_i=0\\ {(1/x_i)}^{\mathrm{\&lambda;}},& x_i>0\end{array}\right.,(0<\mathrm{\&lambda;}\&le;1)$

f _or(x _i)＝(x _i) ^λ，(0＜λ≤1)

λ in above-mentioned two formula is used to regulate F _AndAnd F _OrOperator is to quantitative reasoning result's differentiation precision.For F _AndOperator, in the And local context, the satisfied of all sub-goals is the adequate condition of judging that father's target satisfies.Therefore, work as x _i=0 o'clock, the adequate condition that father's target satisfies was false W=[w ₁=1, w ₂=0..., w _n=0].At this moment, F _AndOperator is pure logical, and father's satisfiability of targets equals 0.Work as x _i＞0 o'clock, f _And(x _i) be x _iSubtraction function, i.e. x _iMore little then its weight is big more, is consistent with the and-like principle of OWA.Similar with it, in the Or local context, satisfying of arbitrary sub-goal all can be used as the adequate condition that judge father target satisfies.According to the or-like principle of OWA, f _Or(x _i) be x _iIncreasing function, i.e. x _iBig more then its weight is big more.

Step 6:

Utilize the different evidences source of satisfiability of targets in the synthetic evidence context of ENV operator in the reasoning algorithm.Both there had been exploded relationship longitudinally in for example high-rise target, there are horizontal promotion or the anti-relation that promotes again, the operation that need synthesize for different sources of evidence, what the present invention adopted is the mode of common weighting, by the hobby of user according to oneself, come to give different weights to the evidence source of the contextual evidence of exploded source and horizontal relationship, the module that exists a user to import different weights according to preference in the device 202 is finished this function, before carrying out synthetic operation, can the prompting user carry out the input of weights, adopt the mode of weighting to synthesize different evidence source in the evidence context at last.

Concrete ENV operator formula is:

$F_{\mathrm{env}}(e_1,e_2,...,e_n)=\underset{i=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&omega;}}_i\&times;e_i,$ And (ω _i∈ [1,1])

Step 7:

The quantitative satisfiability data of destination node that satisfiability reasoning arithmetic unit 202 finishes to obtain with reasoning are imported in qualitative, the quantitative conversion equipment 201 of another one satisfiability of targets and are handled, make father node also comprise simultaneously quantitatively and numerical value qualitatively, be beneficial to the user has integral body directly perceived to the father node satisfiability cognition.

Step 8:

To comprise father node, child node and the complete demonstration of relationship by objective (RBO) figure of relation each other thereof by object construction map display device 3, each node all should comprise quantitatively and satisfiability value qualitatively among the figure, for example: node A quantitative values is: 0.5, and qualitative value is: general satisfied.Make the user that the assurance of an integral body be arranged object construction figure.

Step 9:

Object representation and reasoning process thereof finish.

The present invention is on the basis of existing achievement in research, relate to many angles from software with artificial main body, use for reference the existing achievement in research of uncertain artificial intelligence field, the learning requirements of people to satisfiability satisfied in design, and the satisfiability that meets human general mode of thinking characteristics is represented and inference method. at first, it is qualitative to the present invention is based on satisfiability, the benefit and limitation of quantificational expression, a kind of satisfiability of targets representation model based on cloud model is proposed, this model is with qualitative, quantificational expression is considered as unified integral body, can when keep the semantic understanding ability, also provide satisfiability is carried out the ability that fine granularity is distinguished.Secondly, a kind of inference method based on OWA operator core concept is proposed, this method avoid based on pure logical, " or " the too The reasoning results of " bigoted "; Simultaneously, father's target satisfaction degree, between the minimum and maximal value of sub-goal satisfiability, the human general thought characteristic of reaction preferably.

Therefore; (five-tuple of object module is represented based on the satisfiability of targets model of cloud; other gives qualitative level, qualitative, quantitative values conversion method etc.), all belong to the protection domain of claim of the present invention based on the target inference method (AND, OR, the design of ENV operator and implementation algorithm etc.) of OWA.

Claims (7)

1. One kind be applied under the software systems engineering environment based on the satisfiability of targets representation model of cloud model with based on the satisfiability of targets inference method of OWA operator, it is characterized in that: (1) is based on cloud model objective definition satisfiability representation model, use a five-tuple to represent target, make target have qualitative and quantitative satisfiability value simultaneously; (2), obtain the evidence source that father's target is decomposed the context satisfiability by these two kinds of operators based on the contextual AND operator of exploded, OR operator in the OWA operator definitions target reasoning process; (3) use the mode of giving weight to reflect the user for the different preference of target, the different evidences source that utilizes the ENV operator in the inference method to synthesize satisfiability of targets in the evidence context; (4) show complete object construction figure after the reasoning terminal procedure, each target all comprises qualitative and quantitative satisfiability value, and it is cognitive more intuitively to make that the user has target.
2. 2. according to claim 1 a kind of be applied under the software systems engineering environment based on the satisfiability of targets representation model of cloud model with based on the satisfiability of targets inference method of OWA operator, five-tuple CSat=＜the L that it is characterized in that satisfiability of targets representation model in (1), X, μ, CFLX, CFXL 〉, L is that X is the corresponding sub-domain of L by the qualitative rank of the numerical characteristic sign of cloud $U_L\&Subset;U$ Interior arbitrary satisfiability quantitative values, μ is degree of certainty .CFLX:L * μ → X that X is under the jurisdiction of L, CFXL:X → L * μ is the mapping function between L and X, is used to realize the mutual conversion of the two.
3. 3. according to claim 1 a kind of be applied under the software systems engineering environment based on the satisfiability of targets representation model of cloud with based on the satisfiability of targets inference method of OWA operator, it is characterized in that (1) further comprises: other gives qualitative level and the proper vector＜Ex of the corresponding cloud of each qualitativing concept, En, He 〉, and the directed quantitatively function C FLX of conversion that utilizes cloud algorithms such as former piece cloud, consequent cloud to carry out, CFXL etc.
4. 4. according to claim 1 a kind of be applied under the software systems engineering environment based on the satisfiability of targets representation model of cloud model with based on the satisfiability of targets inference method of OWA operator, the processing that it is characterized in that utilizing in (1) the satisfiability of targets representation model of definition that target is carried out makes target have qualitative, quantitative satisfiability value simultaneously.
5. 5. according to claim 1 a kind of be applied under the software systems engineering environment based on the satisfiability of targets representation model of cloud model with based on the satisfiability of targets inference method of OWA operator, it is characterized in that (2) further comprise: the precision etc. of distinguishing The reasoning results when AND operator, OR operator design, operator are used by regulatory factor.
6. 6. according to claim 1 a kind of be applied under the software systems engineering environment based on the satisfiability of targets representation model of cloud model with based on the satisfiability of targets inference method of OWA operator, it is characterized in that (3) further comprise: the design that the evidence context is synthetic, promptly give weight and reflect different preferences by the user; ENV operator design etc.
7. 7. according to claim 1 a kind of be applied under the software systems engineering environment based on the satisfiability of targets representation model of cloud model with based on the satisfiability of targets inference method of OWA operator, it is characterized in that (4) further comprise: the complete display mode of object construction figure, display-object and relation time each other thereof, show the satisfiability value that each target is qualitative, quantitative simultaneously, the user can have more intuitive cognition for target thus.

Images