Time Series Classification

• Grant number: 10/19375-6
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Start date: March 01, 2011
• End date: May 31, 2011
• Field of knowledge: Physical Sciences and Mathematics - Probability and Statistics - Statistics
• Principal Investigator: Pedro Alberto Morettin
• Grantee: Marcio Valk
• Host Institution: Instituto de Matemática e Estatística (IME). Universidade de São Paulo (USP). São Paulo , SP, Brazil
• Associated research grant: 08/51097-6 - Time Series, Dependence Analysis and Applications, AP.TEM

Abstract

Valk (2011) proposes clustering and classificationprocedures for time series based on quasi U-statistics. Commonly employed distance functions, such as the periodogram and the sampling autocorrelation, are used as kernel for the generalized U-statistics. The class contains a large group of metrics and models. The asymptotic normality holds under very weak regularity conditions. This project aims the generalization of the aforementioned results in the following resume: (i) enlarge the class of kernel functions; (ii) include models with outliers; and (iii) study the theoretical properties of the tests under a more flexible dependence structure. There goals follow the results obtained under PhD Project FAPESP 2007/02767-6, and are inserted on the items 2 and 4 of Projeto Temático FAPESP 2008/51097-6.