Filtering and Sampling Object-Centric Event Logs

TL;DR

This paper introduces new filtering and sampling techniques specifically designed for object-centric event logs to improve the scalability of process mining on large, complex datasets.

Contribution

It presents novel methods tailored for object-centric logs, addressing the complexity of multiple factors like events and object types, which previous techniques did not handle.

Findings

Enhanced scalability of process mining with object-centric logs

Effective filtering reduces log size while preserving behavior

Sampling techniques maintain key process characteristics

Abstract

The scalability of process mining techniques is one of the main challenges to tackling the massive amount of event data produced every day in enterprise information systems. To this purpose, filtering and sampling techniques are proposed to keep a subset of the behavior of the original log and make the application of process mining techniques feasible. While techniques for filtering/sampling traditional event logs have been already proposed, filtering/sampling object-centric event logs is more challenging as the number of factors (events, objects, object types) to consider is significantly higher. This paper provides some techniques to filter/sample object-centric event logs.