Integrating IT Infrastructure Metrics into Business Process Mining: A Telecommunications Industry Case Study

Business Process Mining has become crucial in optimizing operational efficiency by analysing event logs to discover process bottlenecks and predict future behaviours. However, traditional techniques mainly focus on internal process attributes, often overlooking external factors that significantly impact performance. This research addresses the gap by investigating the influence of IT infrastructure performance metrics, specifically within the telecommunications industry, on business process execution. Leveraging event logs combined with infrastructure anomaly data, we employ One-Class Classification methods (OneClassSVM and Isolation Forest) to identify process deviations related to infrastructure issues such as hardware failures and network congestion. Our methodology aligns time-series event logs with concurrent infrastructure anomalies, revealing how these external events affect order-processing workflows. The OneClassSVM yields a recall of 0.81 and an F1 score of 0.70 for class 1, and a recall of 0.93 and an F1 score of 0.95 for non-class 1. The Isolation Forest attains a recall of 0.80 and an F1 score of 0.73 for class 1, while it achieves a recall of 0.94 and an F1 score of 0.96 for non-class 1. Results demonstrate the efficacy of integrating exogenous IT features, improving predictive accuracy, anomaly detection, and facilitating more robust process monitoring. Both methods in our model attain elevated recall and F1 scores in predicting both classes. While acknowledging limitations such as empirical labelling of data and class imbalance issues, this study highlights the importance of external contextual data in refining process mining techniques and contributes practical insights for predictive maintenance, root-cause analysis, and customer experience optimization within telecommunications. Future research will focus on enhancing data labelling accuracy and incorporating additional order-level features for more precise analysis.