Laura sebu icstcc_merging

ICTSCC 2015 - 19th International Conference on System Theory, Control and Computing

1

Merging business processes for a common workflow in an organizational

collaborative scenario

Maria Laura SEBUHoria CIOCÂRLIE

Computer and Software Engineering DepartmentPolitehnica University of Timisoara


2

Business contextHigh competition on the market for better products and services → organizations start collaborations

• Successful collaboration– Products and operational– Internal tools– Market opportunities– Audience

• Failed collaboration– Credibility – Customer satisfaction– Bad results in portofolio– Wasted resources

• Root cause - differences in approaches, methods, processes and tools used

• Half of outsourcing projects are doomed to fail !


3

Business context• Process-based solution to support multi organizational

collaboration• Identify if a collaboration is feasible (higher compatibility →

reduced costs and less effort for setup)– compute the similarity between business processes

• Same domain, standardized processes OR different domains• Different level of information for business process representation• Considering a modular design

• Collaborative solution based on merging business processes• Validation of the solution (in synch with business objectives)• Monitor the execution• Increase the transparency and structuring of cooperation


4

Business process management

• Business result = output of repetitive actions → process definition– Intra-organizational ↓ good compatibility– Inter-organizational = ∑ intra-organizational business

process• Process mining - extract process definition – Data mining on mining event process logs– Enhance– Correct– Improve


5

Business Process Merge• Assumption– Collaborations = ∑ input from all parties →

common acceptance of a process → positive context → successful outcome

• Preprocessing– Business process representations• Event driven Process Chains (EPC)• Business Process Modeling Notation (BPMN)

– Abstract: reduce to directed graph format


6

Business Process Merge - Preprocessing

• BPMN: a function that maps nodes to types– Activities (BPMN) → Node (Directed graph) Activity:Activity

Type:Activity Name– Events (BPMN) → Node (Directed graph) Event:Event Type:Event Name

• Start Event• End event• Intermediate event

– Gateway (BPMN) → Node (Directed Graph) Gateway:Gateway Type:Condition • Data based exclusive gateway• Inclusive gateway• Parallel gateway• Event-based gateway


7

Business Process Merge - Preprocessing

Translate Activity in Node Translate Gateway in Node


8

Business process merge - Algorithm• Defined over a pair of directed graphs• Initialize common graph with the set of maximum common

regions– Maximum common region = maximum connected sub-graph composed

only of matched nodes and substituted edges• Compute added nodes (nodes present in one or more of the input

graphs and not included in one of the maximum common sub-graphs)

• Compute added edges and assign weights– Weight = number of graphs which contain the edge

• Apply transitive reduction = extract the graph with as few edges as possible that has the same reachability level as the original graph


9

Business process merge - Algorithm

• Match nodes – semantic similarity (SEMILAR API)– Wordnet Lesk Tanim (Greedy and Optimum)• relatedness of two words is equal to the overlaps

counted in the dictionary definition– Wordnet Lin (Greedy and Optimum)• ratio between the amount of information describing

the common parts and the information needed to describe in particular what strings represent


10

Business Process Merge - AlgorithmFunction MergeBusinessProcess (Graph G1, Graph G2)BeginSet{Graph} MCR ← MaximumCommonRegion(G1, G2)Graph CG ← (NCG = {n € MCR}; ECG = {e € ECR})Set{Node} AddedNodes = ComputeAddedNodes(G1) ComputeAddedNodes(G2)∪Set{Node} AddedEdges = ComputeAddedEdges(G1) ComputeAddedEdges(G2)∪Foreach (Node node : AddedNodes)

If {node} ∩ NCG = ØNCG = NCG {node}∪

Foreach (Edge edge : AddedEdges)If {edge} ∩ ECG = ØECG = ECG {edge}∪

Else w(edge)++

Graph TR ← ApplyTransitiveReduction(CG)Return TREnd


11

Business process merge - Algorithm

Function ApplyTransitiveReduction (Graph MCG)BeginForeach (Edge edge : GetEdges(MCG))

List<DirectedPath> dirPaths = getListDirectedPath(edge.getNodeStart(), edge.getNodeEnd());Foreach (DirectedPath path:directedPaths)

If (weight(path)>weight(edge))deleteEdge(MCG, edge);break;

End;


12

Business Process Merge – Post-processing

• Convert merged directed graph into a business process– Map node to business process

element by analyzing the label• Calculate similarity with input

business process– Graph comparison algorithms

• Name of the activities composing a process (labels)

• Topology of the process models

• Validate– Structural: conformance checking

techniques on the event log dataset (fitness, precision, structure)

– Functional: compliance of the process model with the process requirements


13

Business Process Merge Algorithm – Case Study


14



15



16


Removed Edge Path Coverage[C, F] [ [C, E], [E, F]] and [ [C, P], [P, F] ]

[D, G] [ [D, C], [C, E] [E, G] ]

[P, R] [ [P, F], [F, M], [M, R] ]

[A, B] [ [A, X], [X, B] ]


17

Related WorkOur approach Other solutions

Semantic evaluation of label content→ merge business processes from different domains →flexibility

Labels are evaluated in a syntactic context

Nodes are compared only with nodes of the same type (activities, control)

Nodes are compared with any nodes by computing a context similarityRisky !

Transitive reduction of common graph Merge control elements (consecutive splits and joins)Trivial connectors are removed

Maximum Common Region (MCR)Compute added nodes and edgesTransitive Reduction

Construct process model such as the sum of distance between each process and the generic is minimal

Fully automatedBPMN exemplification

NOT fully automatedBackward tracebility


18

Conclusions

• Input: set of business process models• Output: business process collaborative solution• Implementation is– Directed graph based– Independent of process model representation

(directed graph)• Business process = ∑ business processes easy to

set up


19

Thank you for your attention !