Skip to Main content Skip to Navigation

Vers une nouvelle forme d'accompagnement des processus dans les systèmes interactifs : apport de la fouille de processus et de la recommandation

Abstract : An information system is a socio-technical system comprising Business Processes and related data. With the development and democratization of IT tools, stored information is getting more important and distributed. The same is true for processes that are becoming increasingly complex and sensitive for organizations. In order to obtain a service, we had to compose business processes to collect information, transform it and reinject it. The objective of this thesis is to explore the problematic of process control in order to give options for the fabrication of a companion that would guide the user when discovering a process. We focused on computer science aspects. In particular, we are studying the possibility of defining a recommendation methodology based on extracted processes and implementing the corresponding software architecture. The works presented are at the interface between several domains. We chose a research approach based on an iterative cycle. After analyzing the field of process mining and recommendation, we concluded that we needed to strengthen our approach to information gathering. This led us to carry out studies on trace-based systems. We then sought to validate the continuity of our approach on a simple case study. It is about personalizing the course of a student during his training. We have set up a demonstrator which, based on the collection of information from previous promotions, extracts knowledge about the students' courses and makes recommendations on the consequences of the course for a particular student. This study allowed us to set up our end-to-end recommendation process and to propose a first sketch of our architecture. We then looked for a more ambitious case study for which no business process was predefined by an expert. We wanted to see if it is possible to identify behaviors and / or strategies of users using a system. We have placed ourselves in a learning context where the learner is involved in a simulation of a micro-world. This case study allowed us to show how to adapt our methodology and how to take contextual data into account. This case study gave rise to an experiment where two groups used our simulator. The first without recommendation, which allowed us to build a set of execution traces that were used to extract the necessary knowledge on our business processes. The second group benefited from our recommendation system. We observed that in the latter group the performance criterion was improved because the trial / error phenomena are considerably reduced. The experience gained during this thesis pushes us to direct our work toward helping to personalize learning trajectory. In particular, with the definition of a class, taking into account the learner's profile, both in terms of knowledge acquired and learning strategies, leads to the creation of a learning path, and therefore a selection of training blocks, which must be personalized. The methodology we have proposed is a brick to build such an ecosystem.
Document type :
Complete list of metadatas

Cited literature [135 references]  Display  Hide  Download
Contributor : Abes Star :  Contact
Submitted on : Friday, March 27, 2020 - 5:54:09 PM
Last modification on : Wednesday, October 14, 2020 - 3:55:20 AM
Long-term archiving on: : Sunday, June 28, 2020 - 3:12:14 PM


Version validated by the jury (STAR)


  • HAL Id : tel-02522177, version 1



Joffrey Leblay. Vers une nouvelle forme d'accompagnement des processus dans les systèmes interactifs : apport de la fouille de processus et de la recommandation. Interface homme-machine [cs.HC]. Université de La Rochelle, 2019. Français. ⟨NNT : 2019LAROS021⟩. ⟨tel-02522177⟩



Record views


Files downloads