Discovering process models for the analysis of application failures under uncertainty of event logs

摘要: Abstract Computer applications, such as servers, databases and middleware, ubiquitously emit execution traces stored in log files. The use of logs for the analysis application failures is known since early days computers. Field data studies have shown that are fraught with uncertainty, i.e., missing or noisy events logs. A body research has dealt successfully uncertainty event process mining from business management community, specifically by discovering models. literature value across several domains, but yet there no study quantifies possible improvements using models, impact context failures. This work addresses detecting First, models discovered logs; then conformance checking used to detect deviations We contribute knowledge engineering a systematic measurement failure detection capability spite events, its accuracy respect obtained Analysis done dataset 55,462 three independent real-life applications. obtain mixed answer depending on under test; our measurements provide insights into analysis.