CN110119399B - Business process optimization method based on machine learning - Google Patents
Business process optimization method based on machine learning Download PDFInfo
- Publication number
- CN110119399B CN110119399B CN201910425690.8A CN201910425690A CN110119399B CN 110119399 B CN110119399 B CN 110119399B CN 201910425690 A CN201910425690 A CN 201910425690A CN 110119399 B CN110119399 B CN 110119399B
- Authority
- CN
- China
- Prior art keywords
- service
- web service
- manufacturing
- manufacturing process
- web
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/22—Indexing; Data structures therefor; Storage structures
- G06F16/2228—Indexing structures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0633—Workflow analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/10—Office automation; Time management
- G06Q10/103—Workflow collaboration or project management
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The invention provides a business process optimization method based on machine learning, which comprises the following steps: the method comprises the steps of storing a manufacturing business Web service description and an instance in advance, responding to a user Web service request, generating abstract Web service based on the Web service description, searching a concrete Web service instance in the Web service instance, and replacing the abstract Web service with the concrete Web service instance to form a workflow combination scheme. The invention provides a business process optimization method based on machine learning, which is easy to determine a global optimal solution, saves the calculated amount, is beneficial to providing richer workflow service combination schemes, and can obtain a manufacturing process model from limited knowledge data through training, so that new data are predicted and estimated, a better workflow service result is obtained, and the method has good generalization capability and is more suitable for a dynamically variable Web service combination environment.
Description
Technical Field
The invention relates to a workflow, in particular to a business process optimization method based on machine learning.
Background
The Web service combination discovers and assembles a plurality of services into a value-added integral service according to a certain rule, and is an effective way for the industry to construct a business process in a networked manufacturing environment and realize the allocation of manufacturing resources according to requirements. The method comprises the steps of establishing a service-oriented framework system by packaging the Web service of an enterprise application system, integrating the application systems among enterprises in a Web service mode, realizing cross-enterprise service combination and cooperation, and realizing the automation of a business process through a cross-enterprise workflow system. However, the service instances that make up the composite service may change dynamically during the execution of the composite service, which makes it difficult for the service instances to be determined at the design or compilation stage, and its QoS attributes may also change over time. Based on this, the prior art proposes a workflow service composition method based on MDP. However, these conventional methods lack generalization capability, and learning results are not accurate enough and are greatly influenced by noise. And requires an environmental model of the threshold state transition probability and the reward value function. Which is often not achievable in a practical environment.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a business process optimization method based on machine learning, which comprises the following steps:
the manufacturing business Web service descriptions and instances are pre-stored,
in response to a user Web service request, generating an abstract Web service based on the Web service description,
and searching a concrete Web service instance in the Web service instance, and replacing the abstract Web service with the concrete Web service instance to form a workflow combination scheme.
Preferably, the searching for the specific Web service instance in the Web service instance further includes:
matching the user request description with the input and output parameter set of each business process description in the business process description library according to the semantic similarity of the I/O concept; if the matching is successful, performing service binding on each service template in the obtained service flow description according to QoS and user preference constraint to obtain a result of executable workflow combination service; otherwise, the service Agent locally performs dynamic construction of the business process description.
Preferably, the searching for the specific Web service instance in the Web service instance specifically includes:
and searching the Web service to be selected which is adaptive to the manufacturing process state parameter set of the manufacturing service request and adaptive to the manufacturing process execution result parameter set of the manufacturing service request.
Preferably, further comprising: respectively establishing indexes for the manufacturing process state parameters and the manufacturing process execution result parameters of the services, respectively using the manufacturing process state parameter indexes and the manufacturing process execution result parameter indexes to carry out adaptation search, and generating two sets of Web services to be selected after adaptation: the method comprises the steps of finding a set meeting the requirements of manufacturing process state parameters after the manufacturing process state parameters are adapted, and finding a set meeting the requirements of manufacturing process execution result parameters after the manufacturing process execution result parameters are adapted;
and obtaining the intersection of the two parameters to obtain the Web service to be selected which simultaneously meets the requirements of the state parameters of the manufacturing process and the execution result parameters of the manufacturing process.
Compared with the prior art, the invention has the following advantages:
the invention provides a business process optimization method based on machine learning, which is easy to determine a global optimal solution, saves the calculated amount, is beneficial to providing richer workflow service combination schemes, and can obtain a manufacturing process model from limited knowledge data through training, so that new data are predicted and estimated, a better workflow service result is obtained, and the method has good generalization capability and is more suitable for a dynamically variable Web service combination environment.
Drawings
Fig. 1 is a flowchart of a method for optimizing a business process based on machine learning according to an embodiment of the present invention.
Detailed Description
A detailed description of one or more embodiments of the invention is provided below along with accompanying figures that illustrate the principles of the invention. The invention is described in connection with such embodiments, but the invention is not limited to any embodiment. The scope of the invention is limited only by the claims and the invention encompasses numerous alternatives, modifications and equivalents. Numerous specific details are set forth in the following description in order to provide a thorough understanding of the invention. These details are provided for the purpose of example and the invention may be practiced according to the claims without some or all of these specific details.
One aspect of the invention provides a business process optimization method based on machine learning. Fig. 1 is a flowchart of a method for optimizing a business process based on machine learning according to an embodiment of the present invention.
The workflow optimization system comprises a service request receiving unit, a service combination generating unit, an executing unit, an adapting unit and a service storage unit: the service storage unit stores specific manufacturing business Web service description and examples; the service request receiving unit receives a user Web service request, converts the user service request into semantic level manufacturing flow description which can be identified by a machine, and then submits the semantic level manufacturing flow description to the service combination generating unit; the service combination generating unit is combined with the Web service description in the service storage unit, generates an abstract workflow combination scheme based on an artificial intelligence (or machine learning/neural network) algorithm and submits the abstract workflow combination scheme to the execution unit; the service adaptation unit searches a suitable Web service instance index in the service storage unit and returns the Web service instance index to the execution unit, and the execution unit replaces abstract Web service with the Web service instance obtained in the service adaptation unit and executes the workflow.
When a user Web service request req is received, whether a Web service s to be selected can meet the input and output requirements of the user Web service request req is searched in a service storage unit, and if the Web service s to be selected is found, the user can use the existing service to complete own functions. And only when the manufacturing service request is matched with the manufacturing process state parameter set and the manufacturing process execution result parameter set of the Web service to be selected, the service request and the Web service to be selected can be considered to be matched. For service requests req and composite services(s) 1 ，s 2 ，s 3 ，…s n ) The adaptation needs to satisfy two constraints:
①
②
wherein r is in Set of manufacturing process state parameters, r, representing service request req out Set of manufacturing flow execution result parameters, s, representing service request req i in Representing Web services s to be selected i S manufacturing process state parameter set of i out Denotes s i Is performed on the substrateAnd (5) setting fruit parameters.
In another alternative embodiment, the Web service request of the user is processed by the service Agent, the service Agent receives the service request description of the user, and searches the manufacturing process description capable of meeting the service request in the service process description library. The searching process is to match the user request description with the input and output parameter set of each business process description in the business process description library according to the semantic similarity of the I/O concept. If the matching is successful, performing service binding on each service template in the obtained service flow description according to constraints such as QoS (quality of service), user preference and the like to obtain an executable workflow combination service result; otherwise, the service Agent locally carries out dynamic construction of the business process description. If the business process description for realizing the required specific business target can be generated, the business process description is registered in a business process description library, and the semantic similarity matching is executed again. If the business process description for realizing the required specific business target is not generated, the service Agent performs dynamic construction of the global business process description through cooperation with the global workflow Agent; if the business process description for realizing the specific needed business target is successfully constructed, the business process description is registered in a business process description library, then the semantic similarity matching step is executed again, and if not, the result of failed combination is returned to the user.
Since there are usually a large number of manufacturing services in the service storage unit, it takes a long time to find a suitable service therefrom. The invention adopts a mode of establishing indexes for the service parameters to accelerate the searching process. The index structure is implemented by a hash table. And respectively establishing indexes for the manufacturing process state parameters and the manufacturing process execution result parameters of the service, respectively using the manufacturing process state parameter indexes and the manufacturing process execution result parameter indexes to carry out adaptation search, and generating two sets of Web services to be selected after adaptation. One is the set meeting the requirement of the manufacturing process state parameter found after the manufacturing process state parameter is adapted, and the other is the set meeting the requirement of the manufacturing process execution result parameter found after the manufacturing process execution result parameter is adapted. And finally, obtaining the intersection of the two parameters, so as to obtain the Web service to be selected which simultaneously meets the requirements of the state parameters of the manufacturing process and the execution result parameters of the manufacturing process.
Wherein, the method comprises respectively indexing the state parameters of the manufacturing process and the execution result parameters of the manufacturing process, and using s 1 ，s 2 .., representing a candidate Web service in a service storage unit. i.e. i 1 ，i 2 ，i 3 ...i n Represents a manufacturing process state parameter, o 1 ，o 2 ，o 3 ...o n Representing the parameters of the execution result of the manufacturing flow. For a service request req, the set of manufacturing process state parameters is r in ＝{r 1 in ，r 2 in ，r 3 in …, the parameter set of the execution result of the manufacturing process is r out ＝{r 1 out ，r 2 out ，r 3 out … }. Before the index is established, the subset of the parameter set needs to be divided to obtain key values of a plurality of indexes.
Wherein r is i p_in I-th subset of divisions, r, representing a manufacturing process state parameter i p_out I-th division subset representing a parameter of a result of execution of the manufacturing process, N =2 n 1,n is the number of manufacturing process state parameters. In the service searching process, a tree traversal algorithm is adopted, a new service is added into the combined service set each time when a node is established towards the next layer, a manufacturing service is deleted from the service storage unit until the output of the last whole tree can meet the requirement of r out . Then selects which single manufacturing service can output r out Which manufacturing services are capable of generating the manufacturing flow state for these services, roll back all the way up. The service found finally is the necessary service for the whole business process.
Step 1: initializing a manufacturing flow execution result set Q and a manufacturing flow Web service set S, and putting S into the tail of Q. Selecting a combination from the head of line of QA service s for collecting the manufacturing process state parameters, the manufacturing process execution result parameters and the state parameter set r of the service in s in Put into set R;
step 2: taking a subset R of R 1 ，R 2 ，R 3 ...R n . Wherein R is 1 Is the complete set of R, R n Parameter 1, R i The middle parameter decreases along with the increase of the value of i (i is more than or equal to 1 and less than or equal to n); selecting R in turn i The manufacturing process state parameter and R are found from the manufacturing process Web service set through indexing i The same Web services to be selected are aggregated to form an aggregation Rs ad ；
And step 3: get Rs in turn ad Manufacturing service s in ad By s ad With the manufacturing process state parameter and R i Carrying out semantic matching on the parameters; if s is ad Match was successful and s ad If a parameter different from R is generated, the step 4 is entered;
and 4, step 4: will s is ad All elements of the sum S are saved into a new combined service set S new And delete s ad (ii) a If S is new Mid-service manufacturing flow execution result parameter and r in Constituent parameter sets R new Is equal to or contains r out If yes, go to step 5, otherwise, go to step S new Putting the queue tail of Q, and entering the step 3;
and 5: initializing an empty parameter set R para Is marked with S new The sum of the parameters of the results of the MES manufacturing process execution and r out Manufacturing services with intersection and putting the manufacturing flow state parameters of these services into R para ；
Step 6: sign S new Intermediate manufacturing process execution result parameter and R para Intersected service, clearing R para And put the newly marked service manufacturing process state parameter into R para (ii) a Repeating the marking step up to S new Intermediate manufacturing process execution result parameter and R para Without intersection, output S new The marked service;
and 7: repeating steps 2-6 until Rs is collected ad And all elements in R and Q are empty.
The method finds the service which can use the divided parameters and generates parameters different from the R parameter by selecting the subset of the state parameter set of the service request, thereby expanding the type of the execution result parameters of the combined service workflow until R can be satisfied out The method of rolling back from bottom to top is adopted to eliminate redundant services.
For a better understanding of the above method, it is illustrated by the following examples. Assuming a set of manufacturing process state parameters r for a service request in = { a, b, c }, manufacturing process execution result parameter r out = x, y, z. When the adaptation is performed through the algorithm, the parameter set needs to be divided first. The element in the parameter set R is R in . R is divided, and the subset of R comprises { a, b, c }, { a, b }, { b, c }, { a }, { b }, and { c }. When the index table is inquired, the subset is selected from at least a plurality of subsets to be inquired according to the number of elements in the parameter set. Therefore, the query is initially performed using { a, b, c }. And if the Web service to be selected is not found through the index table, selecting the next subset to query, such as { a, b }. If a service is found and the service generates new parameters, the service is added to the service set. For example, find the Web service S to be selected by { a, b } 1 。S 1 Is performed by the manufacturing process 1 out = d, e, the element of which is not in the parameter set R = { a, b, c }, at which time S will be present 1 Add to the result set while R = { a, b, c, d, e }. Since the parameter set R does not include the service request manufacturing process execution result parameter set R out And if all parameters are needed, continuing to query and dividing R again. When R = { a, b, c, d, e }. The subset of the R partition is { a, b, c, d, e }, { a, b, c, d }, { a, b, c, e }, when R contains R in The forward search ends with all elements in (1), at which point the result set is { S } 1 ，S 2 ，S 3 ，S 4 ，S 5 }. Some of these services may become redundant, requiring a bottom-up rollback removal for the redundant services. When rolling back the result set, firstly marking the manufacturing process containing the service requestAnd executing the Web service to be selected of the elements in the result parameter set. However, the Web service manufacturing process state parameters to be selected are not all from the service request manufacturing process state parameters, and some Web service manufacturing process execution result parameter sets are needed as the manufacturing process state parameters. Thus continuing to roll back, the annotation can result in S 2 、S 4 、S 5 Web service to be selected S for manufacturing process state parameters 1 . At this time, S 1 、S 2 、S 4 、S 5 All of the manufacturing flow state parameters from the state parameter set of the service request, and all of the manufacturing flow execution result parameters of the service request can be generated. Without marked Web service S to be selected 3 It is deleted as a redundant candidate Web service.
After completing the adaptation of the service request to the candidate Web service, preferably, the service combination generating unit further performs target optimized combination of the manufacturing services with the QoS attribute optimization as a target for the subset of the candidate Web service. If there are I manufacturing sub-businesses in the workflow, each sub-business corresponds to J service to be selected, expressed using the following objective function:
in the formula: t, P, a represent time, cost and performance of the Web service, respectively. k is a radical of ij Is a decision variable, takes the value 0 or 1,k ij The j to-be-selected manufacturing service s representing the ith sub-service when the number is 1 ij Is selected, k ij When 0, the jth manufacturing service s to be selected of the ith sub-service is represented ij Is not selected.
The QoS attributes of a Web service composition are determined by the composition structure of the sub-services in the manufacturing business process and the QoS index of the selected manufacturing service. In the business process, the logic relationship of the sub-businesses determines the combination structure of the Web service, and the logic relationship comprises sequence, parallelism and selection. When d sub-services are in a sequential or parallel structure from the c sub-service, each sub-service is respectively bound with a specific service, and workflow constraints of the sequential and parallel structure are expressed by the following formula;
when f sub-services are a selection structure from the e sub-service, only one service is allowed to bind to a specific service, which is expressed by the following formula.
c+d≤I；e+f≤I：
In an actual manufacturing business, business requirements usually have specific requirements for workflow global services, i.e. global constraints on combined services, and may also have specific requirements for a specific business, i.e. local constraints on individual services, so that both local constraints and global constraints exist:
in the formula: cons (T) i )、cons(P i )、cons(A i ) Respectively representing requests for service req i Local time constraints ofCons (T), cons (P), cons (a) represent global time constraints, cost constraints and performance constraints on the business process, respectively. The global constraint and the local constraint are simultaneously used as the constraint conditions of the optimization problem, so that the problem of re-planning of the combined service can be effectively avoided, and the convergence of the algorithm is improved.
The manufacturing service provider provides a plurality of Web services which are related, so that the matrix r needs to be established iji’j’ To indicate different services s ij And s i’j’ The association between them, i.e. two manufacturing services with association need to be selected simultaneously:
the optimization of the Web service combination is the selection of manufacturing service instances in each service set to be selected, and a group of vectors consisting of one service instance selected from each service set to be selected is a group of solutions; the invention is based on an incremental learning algorithm, selects a service instance by using probability distribution aiming at a Web service combination optimization model, and comprises the following steps:
step 1: respectively and randomly selecting 1 example from the service examples of each service set to be selected according to uniform distribution, repeating the selection for N times to obtain N initial entities, forming the N initial entities into an initial cluster, and recording an iteration coefficient g as 1.
Step 2: substituting the selected service instance vector into a fitness evaluation function:
maxF(s)＝ω 1 ·min T+ω 2 ·min P+ω 3 ·max A
and calculating the fitness of each point in the cluster of the g-th round.
And 3, step 3: and selecting h service instance vectors with highest fitness in the g round cluster to form an optimized cluster, wherein h is less than or equal to N.
And 4, step 4: a discrete probability distribution function of the optimized cluster is estimated. I.e. the probability of each service instance is determined only by the number of service instances in the candidate service set in which it is located. And (3) taking the optimized cluster selected in the step (3) as a sample, and estimating the specific probability distribution P:
suppose that the number of occurrences of service with sequence number s in h optimization entities is
Then->
I.e. the service instance probability for the corresponding ID. The probability distribution function is then smoothed.
Suppose that the maximum response time of all service instances in each candidate service set is maxQoS respectively rt i I =1,2, …, n, then the argument of each service instance in each candidate service set is estimated to be:
Augment s ＝(ω 1 ·min T+ω 2 ·min P+ω 3 ·max A)/maxQoS rt i 。
for service instance with sequence number 0, its probability P 0 =0, the smoothed candidate service set i (i =1,2, …, n), the probability of each instance is:
the probability distribution function is updated according to the following rules:
P″(s|i)＝λP’(s|i)+(1-λ)P s where λ represents the learning ratio.
And 5: sampling is performed according to the estimated probability distribution to generate a new entity. That is, h service instances are generated by adopting a roulette algorithm according to the probability distribution P' in a matrix formed by the service instances of each service set to be selected as part of the entity in the next round.
Step 6: and (5) forming a new round of cluster by the h entities with the highest fitness in the g round of cluster and the h entities generated in the step 5, and increasing the iteration coefficient by 1.
And 7: if g reaches the set iteration times, the algorithm is ended, and the optimal entity in the g-wheel cluster is used as an optimization result; otherwise the algorithm goes to step 2 to continue execution.
Through the multi-objective algorithm, the Web service combination problem in the workflow is optimized, the complexity of calculation is reduced, and the accuracy of the optimization result is improved.
Probability distribution P for each manufacturing service s s The availability q of the corresponding web service may also be recorded av (P s ) A cost q pr (P s ) Reliability q re (P s ) And a time delay q de (P s ) Multidimensional attribute as an index for measuring the performance of the business process, thereby constructing a quality of service parameter evaluation matrix M = [ q ] of the business process i，j ](1 ≦ i ≦ 4,1 ≦ j ≦ r), expressed as:
wherein, each row in the matrix represents the same QoS attribute of different service processes, and each column in the matrix corresponds to the available service process set P r One service of (2) performs the flow.
Filtering out service execution flows which do not meet the limitation of a certain condition of a user, wherein if the time delay cannot exceed the time length required by the user; and then, performing secondary screening, filtering by comparing the relative distance between the service providing node and the service request node on the workflow, finally determining the performance of the combined service execution flow according to the score, and returning the optimal service flow as the response of the user request to the user.
The workflow optimization system preferably monitors and analyzes the change information of the running context of the manufacturing service Web service combination in the process of executing the manufacturing service Web service combination, and adjusts the Web service combination according to the change information of the context. Correspondingly, the service combination generating unit further comprises three modules, namely a combiner, a context sensor and a modifier.
The combiner completes the analysis of the Web service combination description file to obtain a corresponding abstract service combination sequence, finds a Web service set meeting the functional requirements, combines and selects the Web services, and sends the Web services to the execution unit for execution. The context sensor monitors the corresponding service context information and the change of the user context information, analyzes the influence of the change of the context information on the Web service combination execution process and triggers a workflow adjustment event. The corrector adjusts the Web service combination in different conditions of the operation so as to adapt to the dynamic change of the workflow environment.
After receiving the Web service sets from the registry, the Web service sets are written into the alternative library in sequence. The context monitor reads the previous context information from the context information base, compares the previous context information with the currently monitored and collected context information, and writes new context information into the context information base if the context information changes. And judging whether the change information still meets the user requirements and relevant constraints. If the change of the context information has no influence on the combined execution of the Web services, the currently executed Web service can still meet the user requirements and related constraints. Otherwise, sending out a message of suspending execution to the execution unit, transmitting the analysis result and the corresponding strategy given by the user to the corrector, and repeating the manufacturing service combination and selection by using the Web service combination method to obtain the Web service combination sequence with the optimal service combination quality.
For the service quality of a manufacturing service provider, aiming at the characteristic that the QoS attribute in the network combined service can dynamically change at any time, the QoS attribute of the Web service is further analyzed and described, and is modeled into an attribute set containing reliability and coupling degree.
The reliability is the reliability guarantee capability that the determined single service instance can provide by itself and is marked as sigma i The influence of a single point on the whole workflow obtained by the dimension combination of the business process is as follows:
The coupling degree is defined in the participation of the business processWhen the service instance is constructed, the compatibility degree of the service instances i and j determined based on the component development information is recorded as mu ij ∈[0，1]The combination mode on the reliable business process is as follows:
wherein: w is more than or equal to 1 as a control parameter.
Then, a utility function is established to measure the reliability level of the reliable service process, and the utility function based on the QoS attributes is as follows:
F P(s1，s2，…，sn) ＝ασ P(s1，s2，…，sn) +βμP (s1，s2，…，sn)
wherein: α + β =1, and α and β are weight values of the reliability and the coupling, respectively.
In view of the high dynamic of cloud manufacturing service resources, the following process selects a workflow generation strategy of a subsequent service instance based on service capacity and routing validity, extracts service node information and the current state of a service route to jointly decide a service instance instantiation process, selects a service instance with longer routing validity and larger service capacity to complete corresponding functions, and improves the success rate of manufacturing service combinations to the maximum extent while enhancing the stability of the service flow.
The routing validity RE is determined by the minimum value of the routing validity built by the neighbouring nodes, i.e. RE = min { LET (θ) i，j ) }; wherein: theta.theta. i，j Representing the route between two adjacent nodes i, j; LET (theta) i，j ) To route theta i，j The predicted validity period of (a). A route disruption occurs when a node moves out of communication range of another node.
Service capability is for user node N i Obtaining a service node N j Service instance s of j Is expressed as:
Rb Ni (s j )＝α×Q j /rly(N i ，N j )
wherein: q j Is a service nodeN j Providing a service s j (ii) ability of; rly (N) i ，N j ) Representing a slave user node N i To service N j The number of relay times of (2); α is a constant.
The specific steps of the service instance selection algorithm based on the route prediction validity period and the service capability optimal strategy are as follows:
a) Assume that a workflow topology diagram of a service composition scheme generated based on a user service composition request contains m service instances: s = { S = 1 ，S 2 ，…，S m Predefining and generating a service set to be selected, to which each service instance with the same function belongs: SA (S) j )＝{s 1 j ，s 2 j ，…，s mj j J is more than or equal to 1 and less than or equal to m, m j ＞0，s mj j Is a specific service instance. The nodes thereof represent the service sets to be selected for completing the service.
b) In the service process topological graph, judging the current node V based on the following steps i And predecessor node V j Whether the QoS of (c) is consistent:
(1) Assume the current node V i And predecessor node V j The service sets to be selected are SA (S) respectively i )、SA(S j )；
(2) For SA (S) j ) Each service s in n j (1≤n≤m j ) And SA (S) i ) All services s in n i (1≤n≤m i ) Two by two are compared, if s n j Quality of service F(s) of output n j ) Is in accordance with s n i Submitted business requirements, will s n j The service is marked as valid for selection.
c) Extracting node service information and network routing state, calculating service instance for realizing next service at current node V i Service capability of (RbN) i (s j ) And a route prediction validity period t of the service instance i,j And sorting after weighted average. Selecting the optimal and suboptimal service instance as the service selection return value, generating the final business process, and forming a sub-graph G ' (V ', E '), the graphThe vertex V' is the service instance s instantiated in the corresponding service set to be selected n j Edge E' represents two service instances V i And predecessor node V j And simultaneously solving to obtain the manufacturing service capability and the weighted average optimal value of the route connection time.
In conclusion, the invention provides a business process optimization method based on machine learning, which is easy to determine a global optimal solution, saves the calculated amount, is beneficial to providing richer workflow service combination schemes, and can obtain a manufacturing process model from the limited knowledge data through training, so that new data can be predicted and estimated, a better workflow service result can be obtained, and the method has good generalization capability and can be more suitable for a dynamically changeable Web service combination environment.
It will be apparent to those skilled in the art that the modules or steps of the present invention described above may be implemented in a general purpose computing system, centralized on a single computing system, or distributed across a network of computing systems, and optionally implemented in program code that is executable by the computing system, such that the program code is stored in a storage system and executed by the computing system. Thus, the present invention is not limited to any specific combination of hardware and software.
It should be understood that the above-described embodiments of the present invention are merely illustrative of or explaining the principles of the invention and are not to be construed as limiting the invention. Therefore, any modification, equivalent replacement, improvement and the like made without departing from the spirit and scope of the present invention should be included in the protection scope of the present invention. Further, it is intended that the appended claims cover all such variations and modifications as fall within the scope and boundary of the appended claims, or the equivalents of such scope and boundary.
Claims (2)
1. A business process optimization method based on machine learning is characterized by comprising the following steps:
the manufacturing business Web service descriptions and instances are pre-stored,
in response to a user Web service request, generating an abstract Web service based on the Web service description,
searching a concrete Web service instance in the Web service instance, replacing the abstract Web service with the concrete Web service instance to form a workflow combination scheme;
the specific search of the Web service instance in the Web service instance specifically comprises:
searching for Web services to be selected which are adaptive to the manufacturing process state parameter set of the manufacturing service request and adaptive to the manufacturing process execution result parameter set of the manufacturing service request;
respectively establishing indexes for the manufacturing process state parameters and the manufacturing process execution result parameters of the services, respectively using the manufacturing process state parameter indexes and the manufacturing process execution result parameter indexes to carry out adaptation search, and generating two sets of Web services to be selected after adaptation: the method comprises the steps of finding a set meeting the requirements of manufacturing process state parameters after the manufacturing process state parameters are adapted, and finding a set meeting the requirements of manufacturing process execution result parameters after the manufacturing process execution result parameters are adapted;
obtaining the intersection of the two parameters to obtain the Web service to be selected which simultaneously meets the requirements of the state parameters of the manufacturing process and the execution result parameters of the manufacturing process;
after completing the adaptation of the service request to the candidate Web service, the method further comprises:
aiming at the subset of the Web services to be selected, optimizing the QoS attribute as a target, and performing target optimization combination on the manufacturing services; if there are I manufacturing sub-businesses in the workflow, each sub-business corresponds to J service to be selected, expressed using the following objective function:
in the formula: t, P and A respectively represent time, cost and performance of the Web service; k is a radical of ij Is a decision variable, takes the value 0 or 1,k ij The j to-be-selected manufacturing service s representing the ith sub-service when the number is 1 ij Is selected, k ij When 0, the jth manufacturing service s to be selected of the ith sub-service is represented ij Is not selected;
the QoS attribute of the Web service combination is determined by the combined structure of the sub-services in the manufacturing service flow and the QoS index of the selected manufacturing service; in the business process, the logic relationship of the sub-business determines the combination structure of the Web service, and the logic relationship comprises sequence, parallelism and selection; when d sub-services are in a sequential or parallel structure from the c sub-service, each sub-service is respectively bound with a specific service, and workflow constraints representing the sequential and parallel structures are represented as follows:
when f sub-services are a selection structure from the e-th sub-service, only one service is allowed to bind to a specific service, which is expressed as:
c+d≤I；e+f≤I：
simultaneously using the global constraint and the local constraint as constraint conditions of the optimization problem; the global constraint represents that business requirements have specific requirements for workflow global services, and the local constraint represents that a specific requirement exists for a specific business:
in the formula: cons (T) i )、cons(P i )、cons(A i ) Respectively representing requests for service req i The cons (T), cons (P), cons (a) represent the global time constraint, cost constraint, and performance constraint for the business process, respectively.
2. The method of claim 1, wherein finding a specific Web service instance in the Web service instances further comprises:
matching the user request description with the input and output parameter set of each business process description in the business process description library according to the semantic similarity of the I/O concept; if the matching is successful, performing service binding on each service template in the obtained service flow description according to QoS and user preference constraint to obtain a result of executable workflow combination service; otherwise, the service Agent locally performs dynamic construction of the business process description.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910425690.8A CN110119399B (en) | 2019-05-21 | 2019-05-21 | Business process optimization method based on machine learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910425690.8A CN110119399B (en) | 2019-05-21 | 2019-05-21 | Business process optimization method based on machine learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110119399A CN110119399A (en) | 2019-08-13 |
| CN110119399B true CN110119399B (en) | 2023-04-18 |
Family
ID=67523033
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910425690.8A Active CN110119399B (en) | 2019-05-21 | 2019-05-21 | Business process optimization method based on machine learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110119399B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113657844B (en) * | 2021-06-15 | 2024-04-05 | 中国人民解放军63920部队 | Task processing flow determining method and device |
| CN113641977B (en) * | 2021-10-18 | 2022-01-18 | 成都派沃特科技股份有限公司 | Method and server for sending operation and maintenance answering information |
| CN116431246B (en) * | 2023-06-13 | 2023-08-11 | 建信金融科技有限责任公司 | Office automation software configuration method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104166702A (en) * | 2014-08-04 | 2014-11-26 | 浙江财经大学 | Service recommendation method oriented to service supply chain network |
| CN109688056A (en) * | 2018-12-07 | 2019-04-26 | 南京理工大学 | Intelligent Network Control System and method |
Family Cites Families (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7114146B2 (en) * | 2003-05-02 | 2006-09-26 | International Business Machines Corporation | System and method of dynamic service composition for business process outsourcing |
| CN1870562A (en) * | 2005-05-23 | 2006-11-29 | 国际商业机器公司 | Dynamic web service calling method, system and web service agent |
| US7707173B2 (en) * | 2005-07-15 | 2010-04-27 | International Business Machines Corporation | Selection of web services by service providers |
| WO2008015417A1 (en) * | 2006-07-31 | 2008-02-07 | British Telecommunications Public Limited Company | Automatic composition of web services based on syntactiv and semantic rules |
| CN101655943B (en) * | 2009-09-14 | 2016-12-07 | 南京中兴软件有限责任公司 | Enterprise application integrated working flow management method and system |
| CN102004767A (en) * | 2010-11-10 | 2011-04-06 | 北京航空航天大学 | Abstract service logic-based interactive semantic Web service dynamic combination method |
| KR20120066116A (en) * | 2010-12-14 | 2012-06-22 | 한국전자통신연구원 | Web service information processing method and web service compositing method and apparatus using the same |
| CN102123172B (en) * | 2011-02-25 | 2014-09-10 | 南京邮电大学 | Implementation method of Web service discovery based on neural network clustering optimization |
| CN102780580B (en) * | 2012-06-21 | 2015-02-25 | 东南大学 | Trust-based composite service optimization method |
| US9020840B2 (en) * | 2012-10-19 | 2015-04-28 | International Business Machines Corporation | System and method for custom-fitting services to consumer requirements |
| US10217053B2 (en) * | 2015-06-23 | 2019-02-26 | International Business Machines Corporation | Provisioning service requests in a computer system |
| CN105046351A (en) * | 2015-07-01 | 2015-11-11 | 内蒙古大学 | Reinforcement learning-based service combination method and system in uncertain environment |
-
2019
- 2019-05-21 CN CN201910425690.8A patent/CN110119399B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104166702A (en) * | 2014-08-04 | 2014-11-26 | 浙江财经大学 | Service recommendation method oriented to service supply chain network |
| CN109688056A (en) * | 2018-12-07 | 2019-04-26 | 南京理工大学 | Intelligent Network Control System and method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110119399A (en) | 2019-08-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109491790B (en) | Container-based industrial Internet of things edge computing resource allocation method and system | |
| CN102388595B (en) | Resource matching method and device in VPC migration | |
| CN110119399B (en) | Business process optimization method based on machine learning | |
| CN109818786B (en) | Method for optimally selecting distributed multi-resource combined path capable of sensing application of cloud data center | |
| CN114721833B (en) | Intelligent cloud coordination method and device based on platform service type | |
| CN102158417A (en) | Method and device for optimizing multi-constraint quality of service (QoS) routing selection | |
| CN113037877B (en) | Optimization method for time-space data and resource scheduling under cloud edge architecture | |
| CN106055395A (en) | Method for constraining workflow scheduling in cloud environment based on ant colony optimization algorithm through deadline | |
| CN104881581A (en) | IoT (Internet of Things) data high-efficiency analysis method | |
| CN111966495B (en) | Data processing method and device | |
| CN113098714A (en) | Low-delay network slicing method based on deep reinforcement learning | |
| CN104834751A (en) | Data analysis method based on Internet of things | |
| CN102945283A (en) | Semantic Web service combination method | |
| CN110119268B (en) | Workflow optimization method based on artificial intelligence | |
| Li et al. | Neighborhood search-based job scheduling for IoT big data real-time processing in distributed edge-cloud computing environment | |
| CN113010296B (en) | Formalized model based task analysis and resource allocation method and system | |
| CN110135747B (en) | Flow customization method based on neural network | |
| CN113596868A (en) | 5G network slice resource management mechanism based on SDN and NFV | |
| Beigrezaei et al. | Minimizing data access latency in data grids by neighborhood‐based data replication and job scheduling | |
| CN112925964A (en) | Big data acquisition method based on cloud computing service and big data acquisition service system | |
| CN115062868B (en) | Pre-polymerization type vehicle distribution path planning method and device | |
| WO2022252694A1 (en) | Neural network optimization method and apparatus | |
| Elkael et al. | Improved monte carlo tree search for virtual network embedding | |
| CN114154685A (en) | Electric energy data scheduling method in smart power grid | |
| CN112925831A (en) | Big data mining method and big data mining service system based on cloud computing service |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |