Early Classification of Time Series: Cost-based multiclass Algorithms - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2021

Early Classification of Time Series: Cost-based multiclass Algorithms

Résumé

Early classification of time series assigns each time series to one of a set of pre-defined classes using as few measurements as possible while preserving a high accuracy. This implies solving online the trade-off between the earliness and the prediction accuracy. This has been formalized in previous work where a cost-based framework taking into account both the cost of misclassification and the cost of delaying the decision has been proposed. The best resulting method, called ECONOMY-γ, is unfortunately so far limited to binary classification problems. This paper presents a set of six new methods that extend the ECONOMY-γ method in order to solve multiclass classification problems. Extensive experiments on 33 datasets allowed us to compare the performance of the six proposed approaches to the state-ofthe-art one. The results show that: (i) all proposed methods perform significantly better than the state of the art one; (ii) the best way to extend ECONOMY-γ to multiclass problems is to use a confidence score, either the Gini index or the maximum probability.
Fichier principal
Vignette du fichier
2021_dsaa_paper.pdf (1.01 Mo) Télécharger le fichier
Origine : Fichiers produits par l'(les) auteur(s)

Dates et versions

hal-03650228 , version 1 (24-04-2022)

Identifiants

Citer

Paul-Emile Zafar, Youssef Achenchabe, Alexis Bondu, Antoine Cornuéjols, Vincent Lemaire. Early Classification of Time Series: Cost-based multiclass Algorithms. 2021 IEEE 8th International Conference on Data Science and Advanced Analytics (DSAA), Oct 2021, Porto, Portugal. pp.1-10, ⟨10.1109/DSAA53316.2021.9564134⟩. ⟨hal-03650228⟩
28 Consultations
80 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More