Dr. Massimiliano de Leoni

Office +31 (0)40 2478430 
email m.d.leoni [at] tue.nl

Current Research Interests

Process Mining and Planning

A number of papers in Process Mining describe techniques that are ad-hoc implementations of planning techniques for a specific case. Unfortunately, in the era of big data, these ad-hoc implementations do not scale sufficiently compared with robust, well-established planning systems. Because of this, dr. de Leoni is currently researching on how instances of these ad-hoc problems can be translated into planning problems, which can be solved through off-the-shelf planners. The following publication shows how instances of the conformance checking problem can be represented as classical planning problems in PDDL for which planners can find a correct solution in a finite amount of time. If conformance checking problems are converted into planning problems, one can seamlessly update to the recent versions of the best performing automated planners, with evident advantages in term of versatility and customization:

Decision Support

It aims to provide run-time support during the business-process executions concerning how to proceed the execution of process instances. Dr. de Leoni aims to obtain effective results by combining data- and process-mining techniques on historical data. This enables decisions support that is based on real “facts”. Key publication:

Multi-perspective Process Mining

 In the past, the scarce amount of information stored in event logs only allowed for techniques that focus on the control-flow, i.e. the activities allowed for performance and the ordering with which they can occur. These techniques do not take into account other perspectives, such as the perspectives on resource, time and data.Therefore, they ignore other important factors, namely the resource constraints (who is allowed to execute what), the data aspects (in which conditions certain activities and entire branches are allowed or disallowed) and the time constraints (e.g., deadlines or maximum/minimum duration of activities). Leaving these aspects aside make the analysis focus on an underspecified (and underfitting) process model. The research of dr. de Leoni in this direction is towards discovering multi-perspective process models (i.e., including data-driven decisions and, also, time and resource constraints). When multi-perspective process models are provided, dr. de Leoni has developed techniques to check the conformance of these models against the event logs to pinpoint where deviations occur, how severe they are and which are their root-cause: Key publications:

Visual Analytics and Process Mining

Dealing with large event logs and big data is not only an opportunity; it also represents a challenge. When dealing with extremely large amount of data, process analysts need to be guided where to focus their attention so as to properly tune the process-mining techniques. The guidance can only be provided if process mining techniques are combined with human judgment and creativity to find interesting and relevant patterns. Dr. de Leoni contributed to the proposal of a framework that integrates techniques for Visual Analytics and for process mining to dynamically visualize event data on intuitive models. See, e.g.,:

Root-cause analysis

Process-mining techniques is also about answering questions like “What do the cases that are late have in common?", “What characterizes the workers that skip this check activity?", and “Do people work faster if they have more work?", etc. Such questions can be answered by combining process mining with data-mining techniques. Dr. de Leoni led the definition of a framework where, out of the set of features associated with events, one can select one dependent and a set of independent features and find a correlation between the selected dependent characteristic and the select set of independent characteristic. If a desired characteristic is not readily available in the event log, it may be possible to derive it from those directly available.  By letting the choice be customizable, one can answer a potentially infinite number of questions, such as the questions above or many other questions for which several authors have proposed ad-hoc techniques. Readers are referred to the following key publication for further information:

Process Cubes and Analytical Workflows

In the domain of large event logs, a typical aspect to deal with is the heterogeneity of the behavior recorded. If there are multiple classes of cases exhibiting markedly different behaviors, then the overall process will be too complex to interpret. Moreover, it will be impossible to see di differences in performance and conformance for the different process variants. The different process variations should be analyzed separately and compared to each other from different perspectives to obtain meaningful insights about the different behavior embedded in the process. When the number of process-cube cells to be analyze is large, their analysis can be automated by combining process cubes and analytical workflows. The following paper shows the application of process-cube and analytical workflows to the problem of generating more than 1000 reports that relate the actual performance of students of Eindhoven University of Technology to their studying behavior:

A framework for applying Analytical Workflows (a.k.a. Scientific Workflows in other domains) for Process Mining is described in: