- 102 Views
- Uploaded on
- Presentation posted in: General

Multi-phase Process Mining: Building Instance Graphs

Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author.While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server.

- - - - - - - - - - - - - - - - - - - - - - - - - - E N D - - - - - - - - - - - - - - - - - - - - - - - - - -

Multi-phase Process Mining:Building Instance Graphs

Jason Ree

3/31/11

School of Technology Management

UNIST

- Instance Graph
- Corresponds to a specific class of Petri nets known as marked graphs, T-systems or partially ordered runs
- An abstraction of the instance onto instance Event-driven Process Chains (EPCs)
- Instance EPC
- Describes the control-flow of a case (i.e. a single process instance)
- Provides a graphical representation describing the causal relations

- Instance EPC

- In other words,
- Process Instance of length n: σ = t1t2 … tn∈ T+ , where ti are tasks
- W ∈ T+ N denotes a bag (multiset of process instances)
- W(σ): the number of times a process instance of the form σ appears in the log

T+

σ

T

- Instance Net: a model of one instance
- Since events that appear multiple times in a process instance have to be duplicated in an instance net, we define an instance domain to be used as a basis for generating instance nets
- In an instance net, the instance σ is extended with some ordering relation to reflect some causal relation.

- Also
- Since the set of entries is given as a log, and an instance mapping can be inferred for each instance based on textual properties, only the ordering relation based on the given log needs to be defined.

- In other words,
- An instance net is defined only as a set of entries from the log and an ordering on that set
- Instance nets require
- 1. sequence of events σ∈ T+as they appear in a specific instance
- 2. ordering on the domain of σ is required

Instance Net (σ, )

- Example
- Causal ordering inferred on T
- S wA
- S w B

T = {S, A, B}

- Example
- where case 1 = σ1 and case 2 = σ2
- σ1 = SAB and Dσ1 = {1,2,3}
- Using the causal relation the relation is inferred such that 1 2 and 1 3

Instance Net (σ, )

- In other words,
- An instance graph is a graph where each node represents one log entry of a specific instance and can be used as a basis to generate models in a particular language
- Also it is a graph that typically describes an execution path of some process model, as well as causal relations between tasks

- Note:
- An instance EPC doesn’t contain any connectors other than AND-split and AND-joins connectors
- There is exactly one initial event and one final event

- Open Log data using ProM

- Select Multi-Phase Macro Plugin
- Mining > Multi-phase Macro Plugin

- Configure Options as needed for analysis

- Visualization and Analysis of Event Log Data

- 1. Process Log
- 2. Finding Causal Relations
- {SA, AB, AC, AD, AE, BF, DH, EH, FG, CG, HG, GT}

- 3. Creation of Instance Graph
- 1) Instance ordering from Causal Relations
- 0 1, 1 2, 2 3, 3 4, 4 8, 8 9, 2 5, 5 8, 2 6, 6 7, 7 8, 8 9, 9 10

- 2) Drawing Instance Graph from Instance Ordering

- 1) Instance ordering from Causal Relations

- 4. Conversion of Instance Graph into Instance EPC

- Questions?