A Multi-Level Process Mining Framework for Correlating and Clustering of Biomedical Activities using Event Logs

Cost, time and resources are major factors affecting the quality of hospitals business processes. Bio-medical processes are twisted, unstructured and based on time series making it difficult to do proper process modeling for them. On other hand, Process mining can be used to provide an accurate view of biomedical processes and their execution. Extracting process models from biomedical code sequenced data logs is a big challenge for process mining as it doesn’t provide business entities for workflow modeling. This paper explores application of process mining in biomedical domain through real-time case study of hepatitis patients. To generate event logs from big datasets, preprocessing techniques and LOG Generator tool is designed. To reduce complexity of generated process model, a multilevel process mining framework including text similarity clustering algorithm based on Levenshtein Distance is proposed for event logs to eliminate spaghetti processes. Social network models and four distinct types of sub workflow models are evaluated using specific process mining algorithms. Keywords—biomedical event data; business process modeling; Levenshtein similarity clustering; multilevel process mining; spaghetti process models