RWTH Process Mining Lecture 8: Heuristic Mining

Introduction

In this tutorial, we will explore the Heuristic Mining technique presented in RWTH Process Mining Lecture 8 by Prof. Wil van der Aalst. This method is significant for its ability to handle various complexities in process mining, such as concurrency, noise, and incomplete data. By understanding the two main steps involved—creating a dependency graph and constructing a causal net—we can effectively apply heuristic mining to discover processes and represent them in formats like BPMN or Petri nets.

Step 1: Create a Dependency Graph

The first step in heuristic mining involves constructing a dependency graph, which identifies the relationships between different activities in the process.

Key Actions:

• Collect Event Data: Gather event logs that record the sequence and timing of activities within your process.
• Identify Activities: List all the distinct activities present in your event data.
• Determine Dependencies: Analyze the event logs to find dependencies between activities. This can be done by:
  • Counting how many times one activity precedes another.
  • Noting any direct relationships that indicate that one activity is a prerequisite for another.

Practical Tips:

• Use data visualization tools to help illustrate these dependencies.
• Be aware of noise in your data that may misrepresent true dependencies.

Step 2: Construct a Causal Net

Once the dependency graph is established, the next step is to create a causal net, which provides a more structured representation of the process.

Key Actions:

• Define Causal Relationships: From the dependency graph, determine the causal relationships between activities. This involves:

  • Identifying which activities cause others to occur.
  • Visualizing these relationships in a structured format.

• Convert to Formal Models: Translate the causal net into formal process representations such as:

  • BPMN (Business Process Model and Notation): A flowchart method that depicts the steps in a business process.
  • Petri Nets: A mathematical modeling tool used to describe distributed systems.

Practical Tips:

• Ensure that the causal net captures all relevant activities and relationships to avoid missing critical components of the process.
• Validate the causal net by comparing it with known process behaviors or through stakeholder feedback.

Conclusion

Heuristic mining is a powerful technique for process discovery that can accommodate the complexities of real-world data. By first creating a dependency graph and then constructing a causal net, you can uncover meaningful insights into organizational processes. As a next step, consider applying these techniques to your own event data and explore how the discovered models can be used for process optimization and analysis.