WO2022037102A1

WO2022037102A1 - Multi-party collaboration-oriented method for service value conflict detection and resolution of stakeholders

Info

Publication number: WO2022037102A1
Application number: PCT/CN2021/089365
Authority: WO
Inventors: 涂志莹; 李敏; 王忠杰; 徐晓飞; 徐汉川
Original assignee: 哈尔滨工业大学
Priority date: 2020-08-18
Filing date: 2021-04-23
Publication date: 2022-02-24
Also published as: CN111898098B; CN111898098A

Abstract

A multi-party collaboration-oriented method for service value conflict detection and resolution of stakeholders, the method comprising a conflict detection phase and a conflict resolution phase. Interval numbers of three elements are used to define indexes, conflicts are detected and quantified by means of a difference between an actual value and an expected value of each index, the level of a conflict is calculated according to the level at which a conflict node is located, the number of parent nodes, the number of child nodes, the distance from the conflict node to a local target, the distance from the conflict node to a top-level target, and a conflict propagation path, a conflict resolution sequence is determined, and the conflicts are resolved in sequence as much as possible, or the solutions for conflict minimization are obtained. These solutions can point out problems in the existing multi-party collaboration service solutions, provide reference for multi-domain multi-participant service convergence, and increase the efficiency of service convergence and improve the quality thereof.

Description

A Multi-party Collaboration-Oriented Stakeholder Service Value Conflict Discovery and Resolution Method

technical field

The invention belongs to the technical field of multi-stakeholder service fusion in software engineering, in particular to the technical field of service conflict management and optimization, and relates to a multi-party collaboration-oriented stakeholder service value conflict discovery and resolution method.

Background technique

Multi-party collaboration service means that different service providers share service resources across fields, organizations, platforms, and systems according to a specific collaboration mode, and splicing business processes to achieve interconnection and interoperability, so as to achieve richer service content and smoother service processes. , improve the ability of autonomous management and automatic coordination of services, so that the service value generated by each link in the service process can be smoothly circulated and fully transformed in the entire service Internet, and even create new value-added, forming a complete and reliable service network. In the multi-party collaborative service scenario, there are complex indicator dependencies between multiple participants, and between each participant and the global collaboration service. Because of the existence of these dependencies, the realization of the local goals of each participant is not only related to their own service indicators, but also to the collaborators and the value of the global indicators; at the same time, the realization of the global goals of the collaborative service also depends on multiple participants indicator value. Therefore, when multi-field and multi-participants try to establish a cooperative relationship to achieve service integration, they have to verify: ① whether their own goals are compromised by participating in multi-party collaboration; ② whether their own service indicators can meet the requirements of the collaborators; ③ whether their own service indicators can Whether it meets the overall requirements of collaborative services; ④ Whether its own goals can be improved by participating in multi-party collaboration. Among the above four points, ②③ is the admission principle, that is, the cooperation service must meet these two points before it can be executed according to the original goal; ① is the withdrawal principle, that is, when the participant's profit loss is too great, the participant has to withdraw; ④ It is the principle of change, that is, when the interests of the participants may be better, they should undertake more obligations to adjust indicators to eliminate conflicts and optimize the overall goal. However, in the process of actual service integration, the above ①②③ may not be satisfied at the same time, that is, a conflict occurs. The conflict caused by the poor value of the dependency index can be resolved by optimizing the value of the dependency index, while the conflict caused by the mutual exclusion of targets may not be effectively resolved, but a solution that relatively minimizes the conflict can be found. The difficulty of service integration can be quantified through the discovery of service conflicts, and service integration solutions with practicability and global superiority can be found through conflict resolution.

Existing researches on service conflict discovery and resolution focus on formal conflict checking for business execution timing and multi-objective constraint solving on service non-functional attribute dependencies. The latter is similar to the problem of the present invention. Most of them use the swarm intelligence algorithm to continuously search for optimization in random solution sets based on the principle of multi-objective optimization, trying to find relatively good non-inferior solution sets. These solutions are at least in a certain target. This solution is not necessarily able to meet the purpose of multi-party service fusion and multi-party conflict resolution, and the resolution efficiency is poor. Another disadvantage of these methods is that the conflict of multiple indicators does not occur independently, but often leads to the problem of conflict propagation due to the correlation of indicators. Existing studies have shown that the resolution order must be considered in the process of conflict resolution. The determination of , should consider the conflict size, the difficulty of conflict resolution, the criticality of conflict nodes, and the benefit evaluation after conflict resolution, all of which are not considered in traditional crowd intelligence algorithms. In addition, the traditional method considers that the value of the indicator is an exact real number, but when service modelers design non-functional attributes, it is difficult for the value of the indicator to be accurate to a specific value, and the fuzzy interval number is more It can reflect objective facts. Secondly, the non-functional attributes of services are not constant, and their values may fluctuate in different ranges under different execution environments, so there are different probabilities for certain ranges.

SUMMARY OF THE INVENTION

Aiming at the shortcomings of directly transforming conflict resolution problems into multi-objective constraint problems in most research work, the present invention provides a multi-party collaboration-oriented method for discovering and resolving conflicts of stakeholder service value.

The purpose of this invention is to realize through the following technical solutions:

A multi-party collaboration-oriented stakeholder service value conflict discovery and resolution method, comprising the following steps:

Step 1: Use the ternary interval number to represent the value range and probability distribution of each evaluation index, where:

The number of ternary intervals refers to, for each evaluation index A can be used

where: a ^- indicates the lower bound of the indicator, a ⁺ indicates the upper bound of the indicator, a ^* indicates the most likely value of the indicator, q indicates the superior direction of the indicator, q=1 indicates that the indicator is a profit indicator, the larger the indicator value, the more likely it is Well, q=0 indicates that the index is a cost index, and the smaller the index value, the better;

The probability distribution refers to the indicator

There is a specific probability distribution function F(x) and a probability density function f(x) between [a ^- , a ⁺ ], and satisfy:

_fmax (x)=f(x=a ^* );

The present invention adopts triangular distribution function, namely:

If a ^* > (a ^- +a ⁺ )/2, the indicator value distribution is right-skewed, otherwise it is left-skewed;

Step 2: Build complex metric dependencies between multi-stakeholder services, where:

The indicator dependency relationship includes the dependency relationship between each participant's local service target and its own basic indicators, its own target indicators, the cooperating party's basic indicators and the cooperating party's target indicators, as well as the cooperation service global service target and other global target indicators, several Participant's local service target indicators and dependencies between several participant's basic indicators;

Step 3: Calculate the actual value range of the target layer indicator according to the indicator dependency network and the value range of each service basic indicator, where:

The target layer indicators refer to the global target indicators of the collaborative service and the local target indicators of each participant, and there are multiple non-unique targets;

Said each service basic indicator refers to an indicator that does not depend on the value of any other indicator in the indicator set of each participant;

The actual value range refers to calculating the actual value of the upper-level indicator that has a direct or indirect dependency on it based on the given basic indicators of each service, according to the indicator dependency relationship, and the actual value is also a ternary interval number. ;

Step 4: Calculate the similarity and relative advantage between the actual value and the expected value of the indicator, determine whether there is a conflict and calculate the conflict size according to the conflict tolerance, where:

The similarity calculation formula is as follows:

Suppose the actual value of indicator A is

The expected value is [B] _q = (b ^- ,b ^* ,b ⁺ ),

The probability distribution function of [B] q is F(x), and the probability density function is f(x); the probability distribution function of [B] _q is G(y), and the probability density function is g(y), then the similarity between these two intervals is Calculated as follows:

in,

The relative advantage refers to the advantage of the actual value relative to the expected value. If q=1, then the relative advantage is the probability that the actual value is greater than the ideal value, that is, P(A>B); if q=0, then the relative advantage It is the probability that the actual value is less than the ideal value, that is, P(A<B), and P(A>B)+P(A<B)=1;

If q=1, the relative advantage is calculated as follows:

If q=0, the relative advantage is calculated as follows:

The formula for calculating the conflict size is as follows:

Step 5: Comprehensively consider the level of the conflicting node, the number of parent nodes and child nodes, the distance from the local target and the top-level target, and the conflict level of the conflict propagation path calculation node, where:

The formula for calculating the conflict level is as follows:

Among them, CS _node represents the conflict size of the node, Dis _in (node) represents the farthest distance from the node to the local top-level target node, Dis _out (node) represents the shortest distance from the node to the external top-level target node, and Ind(node) represents the The in-degree of the node, Outd(node) represents the out-degree of the node, k represents the number of conflicting nodes in the local scope that are ancestors of the current node, and l represents the number of conflicting nodes that are the ancestors of the current node in the global scope;

Step 6: According to the order of the local non-target node, the global non-target node, the local target node and the global target node, take the node conflict level from large to small as the optimization order, and resolve the conflict of each node in turn;

Step 7: When resolving the conflict of each node, the goal is to minimize the conflict level of the node, and the basic index directly or indirectly dependent on the node is used as the variable, and the particle swarm optimization algorithm is used until the result converges;

Step 8: According to the optimal conflict level of the node, determine the subsequent adjustable range of the basic index it depends on, where:

The optimal conflict level of the node is the conflict level of the node after the algorithm converges in step 7. In order to avoid the subsequent conflict resolution of other nodes will have a negative impact on the resolved conflict results, it is necessary to restrict the subsequent adjustment range of the optimized basic indicators. , the principle of the present invention is that the smaller the optimal conflict level is, the smaller the subsequent adjustment range of the relevant dependent basic index is. Assuming that the optimal conflict level is CL _min and the original conflict level is CL _org , then the subsequent adjustment range of the relevant index is not greater than

Step 9: Recalculate the conflict level of each node, repeat steps 5 to 8, until there are no conflicting nodes or all non-basic indicators are optimized once to end the cycle;

Step 10: If there are no conflicting nodes, then under the condition that no new conflicts are guaranteed, optimize the global objective to obtain a better conflict-free solution, where:

The conflict-free solution refers to obtaining several group solutions with higher relative advantages of the global target under the premise that each index conflict is less than the conflict tolerance, because the global target is not unique, so the solution is not unique;

Step 11: If there are still conflicting nodes, output the minimized conflict resolution, where:

The conflict-minimizing solution refers to a solution in which the conflict is not completely resolved, but the conflict-minimizing solution has been achieved relative to the initial index setting.

Compared with the prior art, the present invention has the following advantages:

The invention uses the number of ternary intervals to define indicators, finds and quantifies conflicts through the difference between the actual value and the expected value of each indicator, and quantifies conflicts according to the level of the conflicting node, the number of parent nodes and child nodes, the distance from the local target and the top-level target, and The conflict propagation path calculates the conflict level and establishes the conflict resolution order, and resolves the conflict as much as possible or obtains a solution that minimizes the conflict in order. These solutions can point out the problems existing in the existing multi-party collaborative service solutions, and provide references for multi-domain and multi-participant service integration, so as to improve the efficiency and quality of service integration.

Description of drawings

Fig. 1 is the multi-party collaboration-oriented stakeholder service value conflict discovery and implementation flow chart of the present invention;

2 is a schematic diagram of a multi-participant index dependency relationship of the present invention;

Fig. 3 is the pseudo-code diagram of the conflict resolution algorithm by priority of the present invention;

FIG. 4 is a pseudo-code diagram of the conflict-free global multi-objective optimization algorithm of the present invention.

detailed description

The technical solutions of the present invention will be further described below in conjunction with the accompanying drawings, but are not limited thereto. Any modification or equivalent replacement of the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention shall be included in the present invention. within the scope of protection.

The present invention provides a multi-party collaboration-oriented service value conflict discovery and resolution method for stakeholders. The method includes a conflict discovery stage and a conflict resolution stage. The conflict level of the nodes is used as the basis for the conflict resolution order to resolve the conflicts one by one, and in the process of conflict resolution, the implementability, fairness and overall superiority of the resolution scheme are guaranteed.

As shown in Figure 1, the specific implementation steps are as follows:

1. Conflict discovery stage:

Step S1: Each participant and the multi-party collaborative service globally define the service evaluation index of the specific domain it is concerned about. The content of each index includes the expected value of the index (the number of ternary intervals), the superior direction of the index, whether it is a target node, and the index Information such as value boundaries. Each evaluation index A has an expected value represented by a ternary interval number, that is,

where: a ^- represents the lower bound of the indicator, a ⁺ represents the upper bound of the indicator, and a ^* represents the most likely value of the indicator, which can be the mode, median or mean. q represents the superior direction of the indicator. If it is a benefit-type indicator, the larger the value of the indicator, the better, then q=1; if it is a cost-type indicator, the smaller the indicator value, the better, then q=0.

Each indicator has a specific probability distribution function F(x) and probability density function f(x) that satisfy:

f _max (x)=f(x=a ^* ).

The present invention adopts triangular distribution function, namely:

If a ^* >(a ^- +a ⁺ )/2, the distribution of the index values is right-skewed, otherwise, it is left-skewed.

Step S2: Excavate the indicator dependencies within each participant, among multiple collaborators, and between participants and the whole world, and build an indicator dependency network. The indicator dependency includes the name of the dependency indicator and the dependency formula, and the obtained indicator The dependency network is shown in Figure 2. The incoming edge of a node represents the index of its dependence, and the outgoing edge of the node represents the index of its influence.

The above two steps can obtain the global evaluation index set of each participant and multi-party collaboration service, as shown in the following code, each index includes the expected value range exp_data, the dependency index set rel_nodes, and the dependency function rel_function (which is a lambda expression. , which can be directly parsed and solved), the effective value range of the indicator border (referring to the actual maximum adjustable range of the indicator), the indicator type is_topgoal (2: global goal; 1: local goal; 0: non-leaf non-target index; -1 : leaf node).

Parse the above indicator definition information, instantiate a directed graph DiGraph with the help of the networkx graph computing package, add nodes according to the indicator name, add edges according to rel_nodes, and add other attribute information for each node.

Step S3: extracting the basic indicators of each participant, the basic indicators refer to indicators that do not depend on any other indicators, that is, indicators that rely on an in-degree of 1 in the network, hereinafter referred to as leaf nodes. Base metric if rel_nodes is empty, or is_topgoal is -1.

Step S4: According to the value of the basic indicator, calculate the actual value imp_data of the upper-level indicator in turn according to the dependency relationship formula from bottom to top. The actual value of the leaf node is the expected value. Add all non-leaf nodes to the unknown_nodes node set with unknown actual value, traverse each node in unknown_nodes, if each indicator in its rel_nodes has imp_data, calculate the imp_data of the node, and delete it from the unknown_nodes set. Repeat the above steps until unknown_nodes is empty.

Step S5: Set the conflict tolerance, calculate the conflict size according to the similarity between the actual value and the expected value of the index and the relative advantage, the greater the similarity, the smaller the conflict size, the greater the relative advantage, and the smaller the conflict. When the conflict is less than the tolerance, the conflict size is set to 0.

Suppose the actual value of indicator A is

The expected value is [B] _q = (b ^- , b ^* , b ⁺ ).

The probability distribution function of q is F(x), and the probability density function is f(x); the probability distribution function of [B] _q is G(y), and the probability density function is g(y).

Then the similarity calculation formula of these two intervals is as follows:

in,

Relative advantage refers to the advantage that the actual value has over the expected value. If q=1, then the relative advantage is the probability that the actual value is greater than the ideal value, that is, P(A>B); if q=0, then the relative advantage is the actual value. The probability that the value is less than the ideal value, that is, P(A<B). And P(A>B)+P(A<B)=1.

Taking q=1 as an example, the formula for calculating relative advantage is as follows:

Taking q=0 as an example, the formula for calculating relative advantage is as follows:

The formula for calculating the conflict size is as follows:

Among them, S _{a, b} is the similarity between the actual value and the expected value of the index, Adv _{a, b} is the relative advantage of the actual value of the index relative to the expected value, e is a natural constant, e=2.718 .

Second, the conflict resolution stage:

If the total number of conflicting nodes is greater than 0, the conflict needs to be resolved.

Step S6: Calculate the conflict level of the conflicting nodes. The calculation of the conflict level needs to comprehensively consider the level of the conflicting node, the number of parent nodes and child nodes, the distance from the local target and the top-level target, and the conflict propagation path.

The formula for calculating the conflict level is as follows:

Among them, CS _node represents the conflict size of the node, Dis _in (node) represents the farthest distance from the node to the local top-level target node, Dis _out (node) represents the shortest distance from the node to the external top-level target node, and Ind(node) represents the The in-degree of a node, Outd(node) represents the out-degree of the node, k represents the number of conflicting nodes in the local scope that are ancestors of the current node, and l represents the number of conflicting nodes that are the ancestors of the current node in the global scope. Considering the impact of conflict propagation is generally exponential, so taking e as the base indicates the property of conflict propagation.

Step S7: According to the four levels of non-leaf non-target local indicators, non-leaf non-target global indicators, participant local target indicators, and collaborative service global target indicators, each layer sorts the nodes according to the node conflict level from large to small. The first conflicting node that has not been optimized is taken as the node that resolves the conflict first.

Step S8: Determine the basic indicator set Dep on which the node depends, and traverse all basic indicators. If there is at least one path from the node to the node to be optimized, it is added to the indicator set Dep.

Step S9: Taking the above index set Dep as a variable, and aiming at minimizing the conflict level of the node, the particle swarm optimization algorithm is used until the result converges, and the algorithm pseudocode is shown in Figure 3.

Step S10: Calculate the subsequent adjustable range of the dependent basic index according to the optimized conflict level. The smaller the conflict level is, the smaller the subsequent adjustable range of the index in Dep will be, so as to ensure that no new conflicts will be introduced in the future. The conflict resolution is performed repeatedly. Assuming that the optimal conflict level is CL _min and the original conflict level is CL _org , the subsequent adjustment range of related indicators is not greater than

Step S11: Determine whether each conflicting node has been optimized at least once, if so, but the total number of conflicting nodes is not equal to 0, indicating that the conflict cannot be completely eliminated, output the solution that minimizes the conflict, and evaluate the participants in the solution and Collaborative Services Global Goal Relative Advantage. If not, then go back to step S6 to continue to resolve the conflict. If the total number of conflicting nodes is equal to 0, there is no need to resolve the conflict, but try to optimize the global objective, which in the present invention refers to improving the relative advantage of the global objective.

Step S12 : Under the condition that the two admission principles are not satisfied while the exit principle is not satisfied, a decomposition-based differential evolution algorithm is used to optimize the global objective of the cooperative service. The algorithm pseudocode is shown in FIG. 4 . The two admission principles refer to whether each participant's own service index can meet the requirements of the collaborator; and whether each participant's own service index can meet the global requirements of the collaboration service. The exit principle refers to whether the local goals of the participants are compromised by participating in multi-party cooperation.

Step S13: Adjust the value range of each participant's basic index, determine whether the current solution satisfies the change principle, and calculate the change willingness. The principle of change refers to whether the local goals of the participants can be improved by participating in multi-party cooperation. If the change principle is satisfied, then the index changes and change willingness=1; otherwise, the index changes but 0<change willingness<1.

The formula for calculating willingness to change is as follows:

Among them, q1,q2,...,qn refer to the participant's n local targets,

is the raw comparative advantage of the target,

It refers to the relative advantage after optimization.

Step S14: Whether the number of iterations is reached, if not, return to step S12; if yes, output a conflict-free solution, and evaluate the satisfaction of each participant in each solution, and the satisfaction is obtained according to the willingness to change the indicators.

The formula for calculating satisfaction is as follows:

Among them, m refers to the number of all indicators that the participant pays attention to, and S _i refers to the similarity between the adjusted indicators and the initial value.

Matters not addressed in the present invention are known in the art.

Claims

A multi-party collaboration-oriented service value conflict discovery and resolution method for stakeholders, characterized in that the method comprises the following steps:

Step 1: Use the ternary interval number to represent the value range and probability distribution of each evaluation index;

Step 2: Build complex indicator dependencies between multi-stakeholder services;

Step 3: Calculate the actual value range of the target layer indicator according to the indicator dependency network and the value range of each service basic indicator;

Step 4: Calculate the similarity and relative advantage between the actual value and the expected value of the indicator, determine whether there is a conflict and calculate the conflict size according to the conflict tolerance;

Step 5: Comprehensively consider the level of the conflicting node, the number of parent nodes and child nodes, the distance from the local target and the top-level target, and the conflict level of the conflict propagation path calculation node;

Step 6: According to the order of the local non-target node, the global non-target node, the local target node and the global target node, take the node conflict level from large to small as the optimization order, and resolve the conflict of each node in turn;

Step 7: When resolving the conflict of each node, the goal is to minimize the conflict level of the node, and the basic index directly or indirectly dependent on the node is used as the variable, and the particle swarm optimization algorithm is used until the result converges;

Step 8: According to the optimal conflict level of the node, determine the subsequent adjustable range of the basic index it depends on;

Step 9: Recalculate the conflict level of each node, repeat steps 5 to 8, until there are no conflicting nodes or all non-basic indicators are optimized once to end the cycle;

Step 10: If there are no conflicting nodes, then under the condition that no new conflicts are introduced, optimize the global objective to obtain a more superior conflict-free solution;

Step 11: If there are still conflicting nodes, output the minimized conflict solution.
The multi-party collaboration-oriented service value conflict discovery and resolution method for stakeholders according to claim 1, characterized in that in step 1, the number of ternary intervals refers to the number of ternary intervals used for each evaluation index A.
where: a - indicates the lower bound of the indicator, a + indicates the upper bound of the indicator, a * indicates the most likely value of the indicator, q indicates the superior direction of the indicator, q=1 indicates that the indicator is a profit indicator, and q=0 indicates that the indicator is a profit-type indicator. The indicators are cost indicators.
The multi-party collaboration-oriented stakeholder service value conflict discovery and resolution method according to claim 1, wherein in the step 1, the probability distribution refers to the index
There is a specific probability distribution function F(x) and a probability density function f(x) between [a - , a + ], and satisfy:

fmax (x)=f(x=a * );

The present invention adopts triangular distribution function, namely:
The multi-party collaboration-oriented method for discovering and resolving conflicts of stakeholder service value according to claim 1, wherein in step 2, the index dependency relationship includes each participant's local service target, its own basic index, and its own target index. , the dependencies between the basic indicators of the cooperating party and the target indicators of the cooperating party, and the dependencies between the global service target of the collaborative service and other global target indicators, the local service target indicators of several participants, and the basic indicators of several participants.
The multi-party collaboration-oriented stakeholder service value conflict discovery and resolution method according to claim 1, characterized in that in the step 4, the similarity calculation formula is as follows:

Suppose the actual value of indicator A is
The expected value is [B] q = (b - ,b * ,b + ),
The probability distribution function of [B] q is F(x), and the probability density function is f(x); the probability distribution function of [B] q is G(y), and the probability density function is g(y), then the similarity between these two intervals is Calculated as follows:

in,
The method for discovering and resolving conflicts of stakeholder service value for multi-party collaboration according to claim 1, wherein in step 4, the relative advantage refers to the advantage of the actual value relative to the expected value, and if q=1, then Relative advantage is the probability that the actual value is greater than the ideal value, namely P(A>B); if q=0, then the relative advantage is the probability that the actual value is less than the ideal value, namely P(A<B), and P (A>B)+P(A<B)=1.
The multi-party collaboration-oriented service value conflict discovery and resolution method for stakeholders according to claim 6, characterized in that when q=1, the calculation formula of relative advantage is as follows:

When q=0, the formula for calculating relative advantage is as follows:
The multi-party collaboration-oriented stakeholder service value conflict discovery and resolution method according to claim 1, is characterized in that in described step 4, the calculation formula of conflict size is as follows:

Among them, S a, b is the similarity between the actual value and the expected value of the indicator, and Adv a, b is the relative advantage of the actual value of the indicator relative to the expected value.
The multi-party collaboration-oriented stakeholder service value conflict discovery and resolution method according to claim 1, wherein in the step 5, the calculation formula of the conflict level is as follows:

Among them, CS node represents the conflict size of the node, Dis in (node) represents the farthest distance from the node to the local top-level target node, Dis out (node) represents the shortest distance from the node to the external top-level target node, and Ind(node) represents the The in-degree of a node, Outd(node) represents the out-degree of the node, k represents the number of conflicting nodes in the local scope that are ancestors of the current node, and l represents the number of conflicting nodes that are the ancestors of the current node in the global scope.
The multi-party collaboration-oriented method for discovering and resolving conflict of stakeholder service value according to claim 1, wherein in step 8, the optimal conflict level of a node is the conflict level of the node after the algorithm converges in step 7, and it is assumed that the most The superior conflict level is CL min , and the original conflict level is CL org , then the subsequent related indicators can be adjusted not more than