Micro-MetaStream: Algorithm selection for time-changing data

Author(s)

Debiaso Rossi, Andre Luis ^[1] ; Soares, Carlos ^{[2, 3]} ; de Souza, Bruno Feres ^[4] ; Ponce de Leon Ferreira de Carvalho, Andre Carlos ^[5]

Affiliation

^[1] Sao Paulo State Univ UNESP, Campus Itapeva, Itapeva, SP - Brazil

^[2] Univ Porto, Fac Engn, Fraunhofer Portugal AICOS, Porto - Portugal

^[3] Univ Porto, Fac Engn, LIAAD INESC TEC, Porto - Portugal

^[4] Univ Fed Maranhao UFMA, Sao Luis, Maranhao - Brazil

^[5] Univ Sao Paulo, Inst Ciencias Matemat & Comp, Sao Carlos - Brazil

• Document type: Journal article
• Source: INFORMATION SCIENCES; v. 565, p. 262-277, JUL 2021.
• Web of Science Citations: 0

Abstract

Data stream mining needs to deal with scenarios where data distribution can change over time. As a result, different learning algorithms can be more suitable in different time periods. This paper proposes micro-MetaStream, a meta-learning based method to recommend the most suitable learning algorithm for each new example arriving in a data stream. It is an evolution of MetaStream, which recommends learning algorithms for batches of examples. By using a unitary granularity, micro-MetaStream is able to respond more efficiently to changes in data distribution than its predecessor. The meta-data combines meta-features, characteristics describing recent data, with base-level features, the original variables of the new example. In experiments on real-world regression data streams, micro-metaStream outperformed MetaStream and a baseline method at the meta-level and frequently improved the predictive performance at the base-level. (c) 2021 Elsevier Inc. All rights reserved. (AU)

• FAPESP's process: 08/11569-6 - Meta-learning applied to data streams problems
• Grantee: André Luis Debiaso Rossi
• Support Opportunities: Scholarships in Brazil - Doctorate