Estimation of Average Latent Waiting and Service Times of Activities from Event Logs

Analysis of performance is crucial in the redesign phase of business processes where bottlenecks are identified from the average waiting and service times of activities and resources in business processes. However, such averages of waiting and service times are not readily available in most event logs that only record either the start or the completion times of events in activities. The transition times between events in such logs are the only performance features that are available. This paper proposes a novel method of estimating the average latent waiting and service times from the transition times that employs the optimization of the likelihood of the probabilistic model with expectation and maximization (EM) algorithms. Our experimental results indicated that our method could estimate the average latent waiting and service times with sufficient accuracy to enable practical applications through performance analysis.