Projet de recherche doctoral numero :5166

Description

Date depot: 5 avril 2018
Titre: Data Series Outlier Detection
Directeur de thèse: Themis PALPANAS (LIPADE)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini

Resumé: Data series outlier detection is a problem that finds applications in a wide range of domains, and especially where identifying abnormal situations is of critical importance. Previous work is based on the development of algorithms that can only detect specific types of patterns defined by domain experts (e.g., such as linear trends, high-frequency oscillations, abrupt changes, etc.), operate on individual points instead of sequences, may not always produce answers in real-time, or require human supervision. In this work, we propose to develop novel, general, unsupervised, real-time, sequence-based algorithms, in the case of very large, and continuously evolving (streaming) data series, that are able to detect outliers of arbitrary multiplicity.



Doctorant.e: Boniol Paul