Process Mining on Customer Complaint Analysis using Inductive Miner and k-Means Clustering

UWV, a Dutch agency responsible for employee insurance and labor market services, provided a data set of customer complaints collected over eight months. Managing complex process flows is essential for companies to avoid delays, rework, and waste of resources. This research combines process mining with k means to improve insights from clustering customer complaint data. Using the Inductive Miner algorithm for process mining and clustering the data with k-means based on similarity provides deeper insights and optimizes the analysis beyond what process mining alone can achieve. These insights allow UWV to customize its service strategy more effectively. The results showed that integrating process mining with k-means clustering significantly improved the precision of the process model, making it a more accurate representation of real-world processes. The model fitness improved by $6 \%$, and the precision increased by $21 \%$ after clustering. However, this approach also resulted in a $37 \%$ decrease in generalization and a $1 \%$ decrease in simplicity. It indicates that while the clustered models are more detailed and precise, they become less adaptable to unobserved variation and slightly more challenging to interpret.