CN101908019A - Bayesian network-based internetware dependability evaluation model - Google Patents

Abstract

The invention discloses a Bayesian network-based internetware dependability evaluation model. By aiming at the structural complexity of the internetware and the dynamic characteristics of a system, and researching the system structure of the internetware by using a theory of a software architecture, a structural pattern-based internetware dependability evaluation method is put forward; an architecture model and a semantic structure model of the internetware are established, and structural and semantic analyses are performed on the internetware from top to bottom; a structural pattern-based conditional probability matrix and the Bayesian network are established, a calculating method of system dependability and structure dependability evaluation indexes is put forward, and the dependability of the overall system is calculated from bottom to top according to the dependability of each entity; and the operational characteristics of the system are comprehensively evaluated by using the Bayesian network to perform integration so as to perfect the dependability evaluation system of the internetware. Practices prove that an internetware dependability evaluation method of the invention is reliable, has systematicness and completeness, and can be used in the fields of software dependability techniques and software evaluation techniques.

Description

A kind of internetware dependability evaluation model based on Bayesian network

[technical field]

The invention belongs to software trust technology and software evaluation technical field, be specifically related to a kind of internetware dependability evaluation model based on Bayesian network.

[background technology]

Software credibility (Dependability) is meant the service process proof that computer software provides, and is trustworthy.According to the difference of user to the systemic-function demand, the user often has different stressing to the system credibility demand, and each side all is credible attributes.Credibility is made of following attribute: availability (Availability), reliability (Reliability), anti-danger property (Safety), confidentiality (Confidentiality), integrality (Integrity) and maintainability (Maintainability).Software credibility more and more becomes the focus that people pay close attention to.

Entering 21 century, is the every aspect that the network of representative incorporates human society gradually with Internet, has greatly promoted the process of globalization, for infotech and application extension development space.On the other hand, Internet is growing into one " the unified computing machine " be made up of enormous amount and increasing computing equipment, compare with conventional computer system, on the amount of being supported in that Internet can provide for the application problem solving and the matter leap is arranged all.In order to adapt to the major transformation of these applications and infotech aspect, under technical supports such as object-oriented, software component, formed new software forms, i.e. network configuration software (Internetware) based on the Internet environment.Run on owing to network configuration software under the Internet environment of open, dynamic and difficult control, its credibility assessment just becomes extremely important.

At first, Internet is open, can freely connect, and the restriction in not free and space does not have geographic distance conception yet.As long as follow the procotol of regulation, anyone can add Internet whenever and wherever possible, and this open environment makes the credibility of network become the significant problem that merits attention.Network configuration software is as a kind of alliance by the one group of software entity that is distributed in the wide scope, is provided by the third party, and the assessment of its member and system credibility just becomes the important means of safeguards system safety, normal operation.

Secondly, the dynamic of Internet makes it be different from traditional computer hardware platforms as operation platform, and its node does not have unified control, has " very " distributivity and high degree of autonomy, people, equipment and software have multiple isomerism, and network connection environment has diversity etc.This makes the use-pattern of network configuration software be tending towards personalization and dirigibility, its exploitation, dispose, operation and environment of safeguarding from sealing, static, controlled change into open, dynamically, difficult control.This just need carry out comprehensive assessment to the credibility of each member and entire system, provides the thoroughly evaluating of each scheme.

The 3rd, the most members in the network configuration software all are to be provided by the third party, and its source code and inner structure are invisible usually, can only conduct interviews its credible difficult control by predefined interface specification.Its resistance to control has also been brought big difficulty to monitoring, tolerance and the analysis of member and entire system simultaneously.In addition, because the plug and play of member, the frequent upgrading of third party's member guarantees to bring certain difficulty also can for the credibility of network configuration software.

The thought of credible assessment originates from large-scale resource sharing and integrated application demand among the Internet.Some scholars make self-examination to traditional network security technology in resources integration, introduced credible assessment technology.In recent years because the development of credible component technology has formed the thought of the credible assessment of member with tolerance gradually on conceptual foundations such as traditional software quality, reliability.Existing research method is making certain gains aspect the credibility assessment of single member, can assess the credibility of member from many-sided methods such as code, quality model, trust models.But network configuration software be by a plurality of members by Internet carry out alternately, the collaborative integral body that constitutes, can not use the simple superposition of single member credibility to the assessment of entire system credibility, achievement in research also comparatively lacks in this respect at present.And existing system-level credible assessment models is primarily aimed at the traditional software system, does not consider the characteristic of network configuration software, and it is credible to calculate also black boxs that are based on more, does not go deep into the taking into account system structure.In addition, because the opening of internet, exist functional similarity that a large amount of different developers provide or identical service in the reality.How to select the highest credible service, and between the credibility of the compatibility between the credibility of single service, service, the entire system that forms by Services Combination, carry out balance, making optimum selection, also is the major issue that internetware dependability evaluation research institute will consider.

[summary of the invention]

The present invention is directed to the network configuration software state of the art, propose a kind of internetware dependability evaluation method based on Bayesian network.On the basis of analyzing the network configuration software architecture, set up system structure model, according to the credible evaluation system of this modelling, the total system of network configuration software and the credibility of respectively forming entity thereof are assessed based on the network configuration software of Bayesian network.

The credibility of internetware system is exactly to be made up of the credibility of each entity self credibility and connexon thereof, and is therefore as follows to the appraisal procedure of whole internetware system credibility:

(1) credibility of each entity of forming network configuration software is assessed;

(2) assess the credibility of each inter-entity structure dependence relation and semantic dependence relation;

(3) credibility of each element is combined into the credibility of total system according to the interdependent figure of structure.

In view of this, the internetware dependability evaluation model based on Bayesian network provided by the invention comprises:

1st, compare with member software, the fundamental element of forming network configuration software is not member or module, but software entity, it is the entity that can satisfy certain network configuration software required service that is provided by software developer or domestic consumer, and can independently realize concertedness between each entity.For guaranteeing the service quality of internetware system, when carrying out entity selection, need choose credible high entity after the composition system as far as possible.For assessing the credibility after each entity composition system, according to software architecture network configuration software is carried out structure analysis, define its structure dependence and semantic dependence, it forms the structure dependence figure of inter-entity relation, structure dependence shown in Figure 2 tree and semantic dependency matrix to obtain description shown in Figure 1; Wherein:

1.1st, the structure dependence is defined as in the entity sets of forming network configuration software, the difference of the function that realizes according to entity is divided into different subclass, and in use there is complementary relation in the two or more entities of certain in the subclass that the entity of same or similar function is formed;

1.2nd, semantic dependence is defined as in same entity subclass, for finishing certain function of application system, by the combination of several entities according to certain order;

1.3rd, according to the semantic dependence relation of each physical layer inside, construct semantic dependency matrix, if this physical layer has N entity, then matrix is N*N.

$\left[\begin{array}{cccc}{a}_{11}& a_{12}& ......& a_{1n}\\ a_{21}& a_{22}& ......& a_{2n}\\ ......& ......& ......& ......\\ a_{n1}& a_{n2}& ......& a_{\mathrm{nn}}\end{array}\right]a_{\mathrm{ij}}=\left(\mathrm{0,1}\right)$

A in physical layer A _iMust be at a _jWhen using before, a _Ij=1; Otherwise, a _Ij=0.

2nd, network configuration software is a kind of alliance by the one group of software entity that is distributed in the wide scope, is provided by the third party, and wherein each primary entity all at first is a member, can independently dispose, moves and develop to guarantee single entity.Therefore be applicable to the credible assessment models of member.But characteristics such as the independence of network configuration software, evolution, concertedness, polymorphism and reactivity have determined its primary entity can not be equal to member fully again.The characteristic that the isomerism of Internet " natural " has determined the network configuration software primary entity to be possessed is not quite similar, and its credible assessment also should comprise distinctive index.Set up internetware dependability evaluation index system (tree structure shown in Figure 3 according to the characteristic of network configuration software, and this tree structure of foundation is set up Bayesian network, each node in the network is represented a member entity in the net construction system, chain between the node is represented the dependence between them), the credibility of system is divided into the credible and integrated credibility of entity, and wherein: the entity credibility is defined as the credibility that each entity of forming network configuration software is assessed as independent member; Integrated credibility is defined as the credibility that two or more group of entities are fitted together the formed functional unit in back;

2.1st, comprise in the credible index of the entity of internetware system: priori index and evaluation index; Wherein:

The priori index definition is the initial assessment to the entity credibility, comprises traditional credible indexes of functional, reliability, ease for use, efficient, maintainability, portability and reusability, can adopt classic method to calculate and obtain;

Evaluation index is defined as by to assessing the index of obtaining with credible closely bound up software development methodology, software administration, the software arrangements factor of software, is used for the priori index is revised, so that credible tolerance is more accurate, objective;

2.2nd, comprise in the integrated credible index of internetware system: static index and dynamic indicator; Wherein:

Static index definition is from the system architecture aspect, the structure dependence of assessment inter-entity and the index of semantic dependence, index when also promptly designing; Described system architecture comprises inter-entity interfaces consistance and protocol compatibility;

Security when dynamic indicator is defined as from network configuration software deployment and operation, the index that stable aspect is assessed, index when also promptly moving;

3rd, calculate credible index based on Bayesian network

But the conditional probability of telecommunications services can be provided when the credibility of network configuration software is regarded as system and satisfied credible indexes, calculate credible index based on Bayesian network.Each entity in the network configuration software can be regarded independent member as, its basic credibility can utilize the credible index of member (as functional, reliability, ease for use, efficient, maintainability, portability and reusability etc.) to assess, and uses its quality to revise for the conditional probability that entity satisfies the quality standard requirement simultaneously.And in internetware system, the credibility of upper layer entity also should be considered the influence of its basic entity credibility.

The present invention defines the credible indexes group of each basic component, comprise and contain structure credibility, semantic credible static credible index, and the dynamic credible index that contains network environment credibility, deploy resources credibility, collect the parameter index of each basic component by several different methods, but when being regarded as system and satisfy credible indexes, credibility can provide the conditional probability of telecommunications services, set up assessment scene shown in Figure 4 according to the network configuration software architectural model, bottom-up from Bayesian network, calculate the credible index of total system.Comprise following steps:

Step 1, the credible calculating of basic entity

Calculate single basic entity credibility by following formula:

$P\left(D\right|M)=\frac{P\left(M_e\right|D)P\left(M_p\right)}{P\left(M_e\right)}$

P in the formula (D|M) is the entity credibility, P (M _p)={ P (M _Pi) | i=1,2 ..., n} is the priori index, P (M _e)={ P (M _Ei) | i=1,2 ..., n} is an evaluation index;

Step 2, the credible calculating of upper layer entity

Calculate the upper layer entity credibility that relies on basic entity by following formula:

$P\left(D\right|M)=\frac{P\left(M_e\right|D)P\left(M_p\right)}{P\left(M_e\right)}\·P\left(D\right|D_{\mathrm{Dp}})$

P in the formula (D|M) is the entity credibility, P (D|D _Dp) by the credibility of the basic entity of dependence current entity when credible, P (M _p)={ P (M _Pi) | i=1,2 ..., n} is the priori index, P (M _e)={ P (M _Ei) | i=1,2 ..., n} is an evaluation index;

Step 3, iterative process: each upper layer entity repeating step 2 is carried out iterative computation, the credible lower course of solid that relies on of upper layer entity, up to the credibility that draws top layer entity, the last node layer that on the level of the superiors' entity, increases again, by comprehensive top layer entity credibility, and obtain the credibility of total system by the integrated credible index corrected Calculation of inter-entity;

Step 4, get optimum solution: when system structure mode can be divided into many dependences trees, every dependence tree is all set up Bayesian network, and calculate final entire system credibility, make optimal selection.

Advantage of the present invention and good effect:

The present invention proposes a kind of internetware dependability evaluation model based on Bayesian network.The internetware dependability evaluation index system based on Bayesian network that this model comprises is that a cover can reflect the credible feature of network configuration software comprehensively, and has inner link, play the index set of complementation, not only comprise tolerance to the traditional software product, also comprise tolerance, thereby guarantee the credibility of evaluation result software development methodology, software administration, software arrangements these and product correlative factor.Simultaneously, the present invention gives the computing method of each index, can accurately assess internetware dependability, provides good foundation for the user makes the optimal selection scheme to third party's entity, to the design of network configuration software, exploitation and dispose and also have directive significance.

[description of drawings]

Fig. 1 is the structure dependence figure of internetware system.

Fig. 2 is the structure dependence tree of internetware system.

Fig. 3 is an internetware dependability evaluation index system synoptic diagram.

Fig. 4 is internetware dependability evaluation system assessment scene.

[embodiment]

Internetware dependability evaluation model based on Bayesian network provided by the invention specifically comprises:

Step 1, the credible calculating of basic entity.

At first single basic entity is assessed, the priori index of obtaining according to number of ways or method, and revised by the evaluation index that training data draws, can obtain the credibility of node.

Each entity in the network configuration software can be regarded independent member as, its credibility can utilize the credible index of member (as functional, maintainability, reliability, portability, ease for use, reusability and efficient) to assess, and uses its quality to revise for the conditional probability that entity satisfies the quality standard requirement simultaneously.That is: entity is carried out the back according to quality standard it is served the influence of believable probability to self credibility in exploitation, test, deployment.

Among Fig. 4, the credibility of entity C1 can be calculated by its priori index, uses evaluation index correction simultaneously again.If the credibility of C1 is D _C1, each priori index is M _p={ M _Pi| i=1,2 ..., n}, each evaluation index are M _e={ M _Ei| i=1,2 ..., n}, formula is as follows:

$P\left(D_{c1}\right|M_p,M_e)=\frac{P\left(M_e\right|D_{c1})\·P\left(M_p\right)}{P\left(M_e\right)}$

For the entity C1 among Fig. 4, the priori index that gets access to is functional 0.98, reliability 0.96, ease for use 0.96, efficient 0.95, maintainable 0.92, portable 0.90, reusability 0.94.Satisfy the difference of situation according to evaluation index and carry out following calculating respectively:

Situation 1: when previously described 5 kinds of evaluation indexes all satisfy.Calculate 5 kinds of evaluation index joint probabilities according to the evaluation index conditional probability table and can get P (M _e5 kinds of evaluation index joint probability P (M are satisfied in)=0.33 under the software trust condition _e| D)=0.24, data substitution formula 5 can be got:

$P\left(D\right|M)=\frac{0.33\×(0.98\×0.96\×0.96\×0.95\×0.92\×0.90\×0.94)}{0.24}=0.918$

Situation 2: when satisfying good test, standard programming, detailed documentation, process management in previously described 5 kinds of evaluation indexes.Calculate 5 kinds of evaluation index joint probabilities according to the evaluation index conditional probability table and can get P (M _e5 kinds of evaluation index joint probability P (M are satisfied in)=0.39 under the software trust condition _e| D)=0.31, data substitution formula 5 can be got:

$P\left(D\right|M)=\frac{0.39\×(0.98\×0.96\×0.96\×0.95\×0.92\×0.90\×0.94)}{0.31}=0.84$

Situation 3: when satisfying good test, standard programming, skilled employee in previously described 5 kinds of evaluation indexes.Calculate 5 kinds of evaluation index joint probabilities according to the evaluation index conditional probability table and can get P (M _e5 kinds of evaluation index joint probability P (M are satisfied in)=0.43 under the software trust condition _e| D)=0.37, data substitution formula 5 can be got:

$P\left(D\right|M)=\frac{0.43\×(0.98\×0.96\×0.96\×0.95\×0.92\×0.90\×0.94)}{0.37}=0.776$

Situation 4: when previously described 5 kinds of evaluation indexes do not satisfy.Calculate 5 kinds of evaluation index joint probabilities according to the evaluation index conditional probability table and can get P (M _e5 kinds of evaluation index joint probability P (M are satisfied in)=0.09 under the software trust condition _e| D)=0.13, data substitution formula 5 can be got:

$P\left(D\right|M)=\frac{0.09\×(0.98\×0.96\×0.96\×0.95\×0.92\×0.90\×0.94)}{0.13}=0.462$

The explanation of above-mentioned situation, to evaluation index satisfy situation not simultaneously, the joint distribution probability difference that obtains, evaluation index is also different to the correction degree of priori index, proves absolutely the objectivity and the credibility of this assessment models.

Step 2, the credible calculating of upper layer entity.

The upper layer entity that relies on basic entity need calculate the entity credibility equally, also should consider the influence of its basic entity credibility simultaneously, the influence that to be its basic entity that relies on produced by the structure dependence and the semantic dependence of inter-entity, use the integrated credible index of each inter-entity to revise, to guarantee the accuracy of upper layer entity credibility.Integrated credibility is divided into dynamic indicator and static index, can calculate the credibility of system dynamics aspect and the credibility of static aspect respectively.

(1) static credible index

Among Fig. 4, the credibility of entity B 2 also should be revised by the conditional probability of entity C1, C3 except calculating by the entity of self is credible.If entity B 2 credibilities are D _B2, the entity that it relied on is credible to be D _DP={ D _Dpi| i=1,2 ..., n}, inter-entity structure credibility is D _St={ D _Sti| i=1,2 ..., n}, the semantic credibility of inter-entity is D _Se={ S _Sei| i=1,2 ..., n}, computing formula is as follows:

$P\left(D_{B2}\right|M_p,M_e,D_{\mathrm{St}},D_{\mathrm{Se}},D_{\mathrm{Dp}})=\frac{P(M_e,D_{\mathrm{St}},D_{\mathrm{Se}},D_{\mathrm{Dp}}|D_{B2})\·P\left(M_p\right)}{P(M_e,D_{\mathrm{St}},D_{\mathrm{Se}},D_{\mathrm{Dp}})}$

Wherein, the structure credibility is the credibility of each inter-entity structure dependence, i.e. the credibility of system architecture aspect is as language compatibility, interface consistance.Semantic credibility is the credibility of each inter-entity semanteme dependence, i.e. the credibility of system semantics aspect is as data copy consistency, fault detect, fault recovery.This Bayesian network is by element a in the dependency matrix _{I, j}=1 entity constitutes, so structure is credible and the credible value of semanteme is 1, and then the credible computing formula of entity B 2 can be reduced to:

$P\left(D_{B2}\right|M_p,M_e,D_{\mathrm{Dp}})=\frac{P(M_e,D_{\mathrm{Dp}}|D_{B2})\·P\left(M_p\right)}{P(M_e,D_{\mathrm{Dp}})}$

(2) dynamic credible index

Because various credible situation about descending that causes may appear in the opening of network configuration software environment of living in, dynamic, resistance to control at any time in deployment and operational process, therefore, also should comprise following dynamic credible index in the internetware dependability evaluation.By the dynamic credible index the current service quality of entity is assessed, when finding credible decline, system can switch to the identical and credible higher entity of function from current entity and get on, and has demonstrated fully the characteristics such as independence, evolution, concertedness, polymorphism and reactivity of network configuration software.

Among Fig. 4, be example with entity B 2, entity C1, C3 credibility need be weighted correction by its dynamic indicator to the influence of B2.Dynamic indicator comprises security credibility, stable operation credibility, the unified dynamic credible index that comprehensively becomes.If the dynamic credible of entity B 2 is D _B2', dynamic credible is D _d, computing formula is as follows:

$P\left({D^{\′}}_{B2}\right|M_p,M_e,D_d)=\frac{P(M_e,D_d|{D^{\′}}_{B2})\·P\left(M_p\right)}{P(M_e,D_d)}$

Wherein, D _dBy entity B 2 the entity of dependence entity credible with its dynamic credible weighting after union, that is:

$D_d=\underset{i=1}{\overset{n}{\∪}}{D}_{\mathrm{di}}\·D_{\mathrm{pi}}$

The network environment credibility is carried out metric evaluation according to the inter-entity network condition to its credibility, and influence factor comprises bandwidth, delay.The trustworthy information of the hardware system that the deploy resources credibility will be disposed for entity, operating system, Database Systems resource is assessed.Be the probable value in [0,1] continuum, can record by the method that mistake is injected.

To entity B among Fig. 42, calculating this entity credibility is 0.94, and the credibility of its basic entity is respectively 0.918 and 0.928, and then revised entity credibility is:

P(D?|M)·P(D?|D′)＝0.94×0.928＝0.87

Step 3, iterative process.

Successively analogize calculating according to abovementioned steps 2, the credible lower course of solid that relies on of upper layer entity to the last draws the credibility of top layer entity.Increase by a node layer on the level of the superiors' entity again, i.e. the credibility of total system can be by comprehensive top layer entity credibility, and obtains by the integrated credible index corrected Calculation of inter-entity.

Among Fig. 4, can calculate the credibility of basic entity C1, C2, C3 according to preceding method, the credibility of entity B 2 can calculate by C1, C3, and the credibility of entity B 3 can be drawn by entity C2.In like manner can calculate the credibility of entity A 2 by the credibility of B2, B3, draw the entire system credibility at last.Computation process is as follows:

(1) each basic entity credibility:

$C1:P\left(D_{c1}\right|M_p,M_e)=\frac{P\left(M_e\right|D_{c1})\·P\left(M_p\right)}{P\left(M_e\right)}=0.918$

$C2:P\left(D_{c2}\right|M_p,M_e)=\frac{P\left(M_e\right|D_{c2})\·P\left(M_p\right)}{P\left(M_e\right)}=0.936$

$C3:P\left(D_{c3}\right|M_p,M_e)=\frac{P\left(M_e\right|D_{c3})\·P\left(M_p\right)}{P\left(M_e\right)}=0.928$

(2) second layer entity credibility:

$B2:P\left(D_{B2}\right|M_p,M_e,D_{C1},D_{C3})=\frac{P(M_e,D_{C1},D_{C3}|D_{B2})\·P\left(M_p\right)}{P(M_e,D_{C1},D_{C3})}=0.87$

$B3:P\left(D_{B3}\right|M_p,M_e,D_{C2})=\frac{P(M_e,D_{C2}|D_{B3})\·P\left(M_p\right)}{P(M_e,D_{C2})}=0.92$

(3) top layer entity credibility:

$A2:P\left(D_{A2}\right|M_P,M_e,D_{B2},D_{B3})=\frac{P(M_e,D_{B2},D_{B3}|D_{A2})\·P\left(M_p\right)}{P(M_e,D_{B2},D_{B3})}=0.916$

(4) entire system credibility:

$P\left(D_s\right|M_i,D_{A2})=\frac{P(M_i,D_{A2}|D_s)P\left(D_s\right)}{P(M_i,D_{A2})}=0.909$

Step 4, get optimum solution.

When system structure mode can be divided into many dependence trees, when promptly the composition of system can have multiple choices, we can set up Bayesian network to every dependence tree, and calculate final entire system credibility, so that make optimal selection.

Claims (1)

1. internetware dependability evaluation model based on Bayesian network is characterized in that this model comprises:

1st, according to software architecture network configuration software is carried out structure analysis, define its structure dependence and semantic dependence, obtain to describe its structure dependence figure that forms the inter-entity relation, structure dependence tree and semantic dependency matrix; Wherein:

1.1st, the structure dependence is defined as in the entity sets of forming network configuration software, the difference of the function that realizes according to entity is divided into different subclass, and in use there is complementary relation in the two or more entities of certain in the subclass that the entity of same or similar function is formed;

1.2nd, semantic dependence is defined as in same entity subclass, for finishing certain function of application system, by the combination of several entities according to certain order;

2nd, set up the internetware dependability evaluation system according to the network configuration software characteristic, the credibility of system is divided into the credible and integrated credibility of entity, and wherein: the entity credibility is defined as the credibility that each entity of forming network configuration software is assessed as independent member; Integrated credibility is defined as the credibility that two or more group of entities are fitted together the formed functional unit in back;

2.1st, comprise in the credible index of the entity of internetware system: priori index and evaluation index; Wherein:

The priori index definition is the initial assessment to the entity credibility, comprises traditional credible indexes of functional, reliability, ease for use, efficient, maintainability, portability and reusability, can adopt classic method to calculate and obtain;

Evaluation index is defined as by to assessing the index of obtaining with credible closely bound up software development methodology, software administration, the software arrangements factor of software, is used for the priori index is revised, so that credible tolerance is more accurate, objective;

2.2nd, comprise in the integrated credible index of internetware system: static index and dynamic indicator; Wherein:

Static index definition is from the system architecture aspect, the structure dependence of assessment inter-entity and the index of semantic dependence, index when also promptly designing; Described system architecture comprises inter-entity interfaces consistance and protocol compatibility;

Security when dynamic indicator is defined as from network configuration software deployment and operation, the index that stable aspect is assessed, index when also promptly moving;

3rd, calculate credible index based on Bayesian network

But the conditional probability of telecommunications services can be provided when credibility is regarded as system and satisfied credible indexes, calculate credible index, comprise following steps based on Bayesian network:

Step 1, the credible calculating of basic entity

Calculate single basic entity credibility by following formula:

$P\left(D\right|M)=\frac{P\left(M_e\right|D)P\left(M_p\right)}{P\left(M_e\right)}$

P in the formula (D|M) is the entity credibility, P (M _p)={ P (M _Pi) | i=1,2 ..., n} is the priori index, P (M _e)={ P (M _Ei) | i=1,2 ..., n} is an evaluation index;

Step 2, the credible calculating of upper layer entity

Calculate the upper layer entity credibility that relies on basic entity by following formula:

$P\left(D\right|M)=\frac{P\left(M_e\right|D)P\left(M_p\right)}{P\left(M_e\right)}\·P\left(D\right|D_{\mathrm{Dp}})$

P in the formula (D|M) is the entity credibility, P (D|D _Dp) by the credibility of the basic entity of dependence current entity when credible, P (M _p)={ P (M _Pi) | i=1,2 ..., n} is the priori index, P (M _e)={ P (M _Ei) | i=1,2 ..., n} is an evaluation index;

Step 3, iterative process: each upper layer entity repeating step 2 is carried out iterative computation, the credible lower course of solid that relies on of upper layer entity, up to the credibility that draws top layer entity, the last node layer that on the level of the superiors' entity, increases again, by comprehensive top layer entity credibility, and obtain the credibility of total system by the integrated credible index corrected Calculation of inter-entity;

Step 4, get optimum solution: when system structure mode can be divided into many dependences trees, every dependence tree is all set up Bayesian network, and calculate final entire system credibility, make optimal selection.

Images