Machine Learning (ML) can serve as a backbone to process mining, by improving the quality of the event logs. Data obtained in the real world is error-prone, inconsistent, and sometimes missing crucial information. Machine Learning techniques help in filtering, extracting, and refining such data.
Process discovery and conformance checking are key to creating machine learning problems for processes. Without first discovering the process and aligning the event data with the model, it is impossible to define the machine learning problem properly. For example, to understand or predict a bottleneck or a compliance problem, one first needs to identify the (potential) problem and transform event data into a machine learning problem.
Classification techniques from machine learning are also used to improve the discovered process models. Sometimes the event log is cluttered and might not be effective in discovering meaningful process models. Clustering and classification can be applied to further distribute the event log, and apply process mining techniques.