The framework for evaluating the information completeness of event logs.

Process mining is a sub-field of Business Process Manangement, which aims at mining process models from event logs that record the executions of real business processes. Consequently it is necessarily important to know whether or not the given event log is information complete, i.e. it contains all information required to mine the original process model. This is the so-called log completeness problem of event logs. The most difficult part of the problem is that the original process model is unknown.

The problem has been studied already and some approaches have been proposed to evaluate the information completeness of a log. See http://bpmcenter.org/wp-content/uploads/reports/2012/BPM-12-04.pdf for more detail.

The framework here is proposed to evaluate the approaches available for the log completeness probelm. Main features of the framework cover

• defining different types of information unit,
• parsing event logs files (only in mxml format till now),
• implementing approaches available in Java,
• calling these approaches and output the evaluation results into 'csv' files.

By means of the framework, researchers and/or developers would focus on the proposal of new approaches. A new approach can be very easily added to the framework by implementing the interface 'howmuch.estimator.Estimator' and then put the jar file into the directory './lib/plugins/estimators/'.

Run the script 'logcomp.sh' to get the evaluation results of the specified approaches on some specified logs. Refer to './conf/lc.conf' for mor information on configuration of the running.

Refer to https://github.com/hunter99/EstimatorDummy for more information about developing a new estimator (the implementation of a new approach).