Research activities

I have built a research programme around the concept of personalisation in digital services. My aim is to build an end-to-end chain for the self-adaptive (i.e. data-driven) management of the dynamics of future socio-cyber infrastructures and services. This involves overseeing the operation of complex socio-cyber systems whilst adhering to a pre-established plan and adapting that plan to unforeseen events. This therefore involves being able to analyse the existing situation, model it to express constraints, and construct reasoning to make decisions in order to continue activities whilst ensuring the quality and traceability of the decisions taken.

Since 2016, my work has focused on process mining (W. Van der Aalst). This involves a set of methods designed to extract knowledge about business processes from the traces generated by information systems. I have focused in particular on trace clustering, which involves grouping together traces that share similar characteristics. I have developed an approach based on constructing a vector that characterises a trace, which is then used by the clustering algorithms. The construction of the vector is based on identifying the longest similar sub-sequences between traces and their relative positions. We can thus, by analysing users’ actual execution traces (logs of users' journeys), extract knowledge about users’ journeys and characterise these journeys.

My aim is to build a recommendation system that derives part of its justification and usefulness from taking into account each learner’s individual learning path within a system of activities that is more or less open or constrained. Process mining techniques have reached maturity in the field of information systems, where business processes are clearly defined. In contrast, socio-cyber-physical systems are often characterised as complex systems with few constraints for the user. Building on previous work on trace clustering, it is possible to classify the various trajectories identified within selected activity systems, in order to subsequently determine, in an automated manner, which typical trajectory corresponds to a user’s journey and development within a digital environment lacking clearly defined business processes. It is thus possible to situate their behaviour within the temporality and causality of a trajectory, thereby enabling, within the context of developing a hybrid intelligent system, the human expert to understand and visualise the significant elements of the user’s trajectory.

Research areas

The main area of application to which my work has contributed is learning. Indeed, the advent of Information and Communication Technologies (ICT) in education presents a real opportunity for the dissemination of knowledge to all. Indeed, personalised learning has become a key factor in learners’ success. Generally, research projects tackle the issue of providing personalised training programmes and learning pathways, supporting students in their educational and career development plans. However, few tools are capable of providing relevant indicators to recommend personalised pathways or remedial measures for students experiencing difficulties.

To broaden the scope of my work, I have explored a second area of application: digital libraries. In this context, which has been extensively studied in the social sciences, we have applied our knowledge extraction method to identify typical user journeys within digital libraries. In particular, we have studied the browsing patterns of users of Gallica, the web portal of the National Library of France.

In addition to this research into the self-adaptive management of digital services, I am interested in decentralised approaches to knowledge sharing and data protection using blockchain mechanisms.

Process Mining

Process Mining focuses on the analysis of processes by the development of a set of tools and techniques aimed at extracting process-related knowledge from event log. An event log corresponds to a set of process instances (i.e. traces) following a business process which is a set of coordinated tasks that delivers a specific service or a product. Each recorded event refers to an activity and is related to a particular trace. An event can have a time stamp and a performer (i.e. a person or a device executing or initiating an activity). Process Mining deals with several different activities, whose final aim is to extract knowledge from available log files.

I use process mining algorithms to mine self-defined business processes. I focus on Trace Clustering that aims to regroup traces with similar characteristics

Main contributions

Control of Interative Narrative Unfolding

The control of a system consists in organizing, upstream, the actions to be put in place to achieve an objective, by optimizing a possible criterion; then to take the necessary decisions, during the execution, to carry out the actions planned according to the possibilities of the system. Achieving these two goals requires, at a strategical level, to define the objectives to be achieved over a long time horizon and, if necessary, changing the system. Finally, an analysis of the system must be carried out to detect possible deadlock situations.

My research objective is to propose a solution to adapt the execution of the execution of an activity (or a process) while remaining within the framework defined by the designer of the application (the expert of the domain) but taking into account inflections given by the user (through his actions). My work aims at defining an architecture for the adaptation, the definition of a formal model and the validation of a scenario in the field of the serious games.

My research, now, focuses on defining a recommender system to help the user in its customer journey

Main contributions

Course on Process Mining