Presentation is loading. Please wait.

Presentation is loading. Please wait.

Decision Mining in Prom A. Rozinat and W.M.P. van der Aalst Joosung, Ko.

Similar presentations


Presentation on theme: "Decision Mining in Prom A. Rozinat and W.M.P. van der Aalst Joosung, Ko."— Presentation transcript:

1 Decision Mining in Prom A. Rozinat and W.M.P. van der Aalst Joosung, Ko

2 Outline Introduction Running Example Using Decision Trees for Analyzing Choices in Business Process Challenge for Decision Mining in Business Process Decision Mining with the ProM Framework

3 Introduction Some process mining techniques can be used for Process discovery and Conformance checking. We explore the potential of techniques in order to gain insight into the data perspective of business processes. The process mining algorithm identifies decision points The decision point is concerned with the choice between a Timeout activity and a Process answer activity. The application of data mining techniques has the potential to gain knowledge. Decision Miner offers a wide range of tools related to process mining and process analysis

4 only use the first two columns to obtain some process model. the result of applying the α- algorithm to the event log of some survey process

5 Running Example

6

7 Using Decision Trees for Analyzing Choices in Business Processes ➡ We need to identify decision points in order to analyze the choices in a business process. Identifying Decision Points in a Process Model - In a Petri net, a decision point corresponds to a place with multiple outgoing arcs. - The set of possible decisions must be described with respect to the event log. - A decision can be detected if the execution of the activity has been observed. - A process instance contains the given ‘footprint’ There was a decision for the associated alternative path in the process - It is sufficient to consider the occurrence of the first activity per alternative branch in order to classify the possible decisions.

8 Turning a Decision Point into a Learning Problem - Concept Description : the structural pattern inferred for such a classification problem. - To convert every decision point into a classification problem, whereas the classes are the different decision that can be made. - All attributes that have been written before the considered choice construct may be relevant for the routing of a case at that point.

9

10

11 Turning a Decision Point into a Learning Problem - In order to solve such a classification problem, there are various algorithms available. - Decision Tree : most popular of inductive inference algorithms, and provide a number of extensions. - There are effective methods to avoid overfitting the data. - A decision tree classifies instances by sorting them down the tree from the root to some leaf node.

12 Challenges for Decision Mining in Business Processes Two important challenges need to be addressed in real-life business processes. The first challenge relates to the quality of data, and the correct interpretation of their semantics. The second challenge relates to the correct interpretation of the control-flow semantics of a process model. Labeling Function(l) : l ∈ T ⇏ L is a partial labeling function association each activity with at most one log event. Invisible Activity : An activity t ∈ T, if t ∉ dom(l). Real-life process models may be different from not real-life process models.

13 Because invisible activities cannot observed in the log, the first activity is not always sufficient. Invisible activities need to be traced until the next visible activities have been found. Duplicate Activity : t ∈ T is duplicate activity if ∃ t ∈ T t ≠ t’ ⋏ l(t)=l(t’) Although duplicate activities have an associated log event, its occurrence cannot be used to classify the possible choices. The solution to deal with duplicate activities is to treat them in the same way as invisible activities. ′′

14 If the first transition found in such a branch is neither invisible nor duplicate, ➞ the associated log event can be directly used to characterize the corresponding decision class. If the first transition found in such a branch is either invisible or duplicate, ➞ it is necessary to trace the succeeding transitions Another obstacle can be seen in the correct interpretation of the loop semantics of a process model. Decision points contained in a loop (a) Decision point containing a loop (b) Decision points are loops (c)

15

16 Decision Mining with the ProM Framework The Decision Miner plug-in determines the decision points contained in a Petri net model.

17

18

19


Download ppt "Decision Mining in Prom A. Rozinat and W.M.P. van der Aalst Joosung, Ko."

Similar presentations


Ads by Google