fitness: ability to copy event logs
Process analysis: Measuring the consistency of event logs for a given process model. It can quantify consistency and provide intuitive diagnosis.
Process models are used to analyze processes, such as checking consistency or evaluating the performance of business process redesigns.
The consistency check can measure the "good" degree of the process model compared to the event log that records the execution of the process。
We focus on the fitness of consistency. Given a process model and a series of activities in the log showing the execution of the process instance, if the model allows the same sequence of activities (or very similar activities), the applicability of the trajectory is high (ie, good ).
The deviation of the adaptive dimension manifests itself as skipped activities or inserted activities. Skipped activities are activities that should be performed according to the model but do not occur in the log. On the contrary, inserted activities refer to activities that occur in the log, but should not occur according to the model.
In fact, the severity of skip/insert activities may depend on the characteristics of the activity, for example, certain activities can be skipped without insertion without serious problems. Important activities can cause serious problems. For example, the typical process of handling insurance claims in an insurance company is shown in Figure 1. In Petri nets. In the case of small claims, the "check documents" or "check reasons" activities are often skipped. However, the severity of skipping these activities is less than the severity of skipping basic activities such as "sending money."
precision
A method of measuring the accuracy of the process model is proposed. First, the event log is aligned with the model to give the event log. In this way, for unsuitable logs, the measurement is not sensitive, and more accurate values can be obtained.
The starting point for most business process management (BPM) activities is process models, as they provide insights into possible solutions [21]. Process models are used for analysis (for example, simulation [5]), formulation [21], redesign [18] and process improvement [28, 29].
The consistency check technology compares the process execution recorded in the form of an event log with the process model to quantify how well the model performs relative to its execution.
We focus on precision dimensions. Given an event log and a process model, precision penalizes the model to allow impossible behaviors to be observed in the log.
The "flower" model (F) can provide misleading insights because it also allows no more behavior to appear in the log. In contrast, the model (P) only allows tracking to occur in the log. Therefore, in terms of logs, the accuracy of model P is better than model F
Accuracy: The deviation between the behavior observed in the event log and the modeled behavior in the process model,
First, we align the log with the model to find the complete sequence of activities most similar to the trace in the model for each trace. Then, we use these alignments to measure the accuracy between the original log curve and the model. In this article, we introduce various possible accuracy calculation methods based on alignment.
The accuracy in [24-26] is measured based on log-based model reproduction, but the method in this section is based on alignment [9]. The advantages are manifold. First, the trace in the log does not need to be fully fitted
Estimate accuracy by facing the model and log behavior: Detect the inaccuracy between the model and the log by juxtaposing the log and the behavior allowed by the model (that is , the case where the behavior allowed by the model is more than the behavior reflected in the log ).