On January 3, 2022,Kenza AMZIL will defend her thesis work conducted at the Lispen laboratory, entitled "Contribution to causality analysis through machine learning for decision support, in a supervisory context for Industry 4.0", on the Aix-en-Provence campus at 2:00 p.m.
Summary
With the advent of Industry 4.0, the accompanying acceleration of processes, and the proliferation of data, the challenge for decision-making processes evolving in such a context is to ensure rapid and reliable decision-making.
Key performance indicators (KPIs) are closely linked to decision-making. They are both triggers and drivers. This convinces us that in order to improve decision-making processes, particular attention should be paid to KPIs. When a KPI reveals an abnormal situation, understanding the origin of this deviation is essential in order to find solutions and select one from among several.
In this thesis, we focus on this understanding, in particular on identifying the causal links between a KPI of interest and the manipulable contextual variables, as well as quantifying these causal links. To this end, we propose a causality-oriented decision support system that performs three functions: identifying contextual variables causally linked to a KPI in the form of a causal structure; ranking these variables according to their respective strengths of association with the KPI of interest; and enabling KPI prediction for proactive purposes. The first function aims to provide a better understanding of KPI deviations. It is implemented using a Bayesian causal network learning algorithm. The second function allows for better selection of the best solution and is implemented using a calculation that we propose to perform on the final weights of a neural network with good predictive power for the KPI of interest. The third function, which enables proactive decision-making, is made possible by this same neural network. The method was validated using two benchmark datasets and then compared to other techniques with the same objectives.
Keywords
Causality, Bayesian network learning, cause prioritization, neural networks, decision support, machine learning
Jury
Mr. Vincent CHEUTET, University Professor, DISP-lab, INSA Lyon - Chair
Mr. Marc ZOLGHADRI, University Professor, LISMMA, ISAE-Supméca - Rapporteur
Ms. Elise VAREILLES, Senior Lecturer, ISAE-Supaéro - Rapporteur
Mr. Lionel ROUCOULES, University Professor, LISPEN, ENSAM Aix en Provence - Examiner
Ms. Esma YAHIA, Senior Lecturer, LISPEN, ENSAM Aix en Provence - Examiner
Ms. Nathalie KLEMENT, Senior Lecturer, LISPEN, ENSAM Lille - Examiner
