Description
Discovers the actual process model from event logs, checks conformance to the intended flow, and shows bottlenecks and deviations.
Process mining requires event logs with a case identifier, activity name and timestamp. Discovery algorithms (alpha, heuristic miner, inductive miner) build process models such as Petri nets or BPMN from the logs. Conformance checking replays logs against an intended model to quantify fitness and locate deviations. Enhancement overlays performance: waiting times, rework loops, bottlenecks. For UX, this formalises the comparison between the designed flow and what users actually do, at scale, with quantitative fitness metrics. It is particularly suited to multi-step, multi-actor processes such as claims, onboarding with verification or B2B approval flows.
Attributes
- Discipline
- Computational & Machine Learning
- Loss type
- Unmigrated
- Stage
- Post-launch
- Data
- Quantitative
- Effort
- Weeks
Steps
- Extract an event log with case ID, activity, timestamp and optional resource and attributes.
- Clean and abstract activities to a meaningful level; filter infrequent noise variants.
- Discover process models with the inductive miner (guaranteeing sound models) and inspect variants.
- Run conformance checking against the intended design flow and compute fitness and precision.
- Analyse performance: throughput times, waits, loops and handovers between actors.
- Investigate deviations with qualitative research, then redesign and re-mine to measure change.
Sources
- van der Aalst, Process Mining: Data Science in Action (2nd ed., 2016)
- IEEE Task Force on Process Mining, Process Mining Manifesto (2011)
Also known as
- Process discovery
- Conformance checking
- Event log mining
Categories
- Analytics & Behaviour
- Computational & Machine Learning