CIn - Centro de Informática UFPE

Eventos Relacionados

Defesa de Tese de Doutorado Nº 351: "An Adaptive Learning System for Time Series Forecasting in the Presence of Concept Drift"

O aluno Rodolfo Carneiro Cavalcante irá defender seu trabalho no dia 13 de março, às 9h, na sala D224 Início: 13/03/2017 às 09:00 Término: 13/03/2017 às 00:00 Local: Sala D224

Pós-Graduação em Ciência da Computação – UFPE
Defesa de Tese de Doutorado Nº  351

Aluno: Rodolfo Carneiro Cavalcante
Orientador: Prof. Adriano Lorena Inácio de Oliveira
Co-orientador: Prof.  Leandro Lei Minku (University of Leicester)
Título: An Adaptive Learning System for Time Series Forecasting in the Presence of Concept Drift
Data: 13/03/2017
Hora/Local:  9h – Centro de Informática – Sala D224
Banca Examinadora:

Profa. Teresa Bernarda Ludermir (UFPE / Centro de Informática)
Prof. Ricardo Bastos Cavalcante Prudêncio (UFPE / Centro de Informática)
Prof. Paulo Salgado Gomes de Mattos Neto (UFPE / Centro de Informática)
Prof. André Carlos Ponce de Leon Ferreira de Carvalho  (USP / ICMC)
Prof. Tiago Alessandro Espinola Ferreira (UFRPE / Departamento de Estatística e Informática)


A time series is a sequence of observations collected over fixed sampling intervals. Several real-world dynamic processes can be modeled as time series, such as stock price movements, exchange rates, temperature of a city, among others. One of the main problems in time series analysis is the forecasting of future values. As a special kind of data stream, a time series may present concept drifts, which are changes in the underlying data generation process from time to time. When a concept drift occurs, old observations may become irrelevant to define the current state and future behavior of the data stream. The concept drift phenomenon affects negatively time series analysis and forecasting methods based on observing past behaviors of a time series. Despite the fact that concept drift is not a new research area, the effects of concept drifts in time series are not widely studied. Some approaches proposed in the literature to handle concept drift in time series are passive methods that successive updates the learned model  to the observations that arrive from the data stream. These methods present no transparency to the user and present a potential waste of computational resources. Other approaches are active methods with first detect to adapt the learned model to a concept drift in data. These methods present an explicit drift detection, which inform the user the occurrence of changes. By using explicit detection, the learned model is updated or retrained just in the presence of drifts, which can reduce the space and computational complexity of the learning system. These methods generally are based on monitoring the residuals of a fitted model or on monitoring the raw time series observations directly. However, these two sources of information (residuals and raw observations) may not be so reliable for a concept drift detection method applied to time series. The main contribution of this work is an adaptive active learning system which is able to handle concept drift in time series. The proposed method, called FW-FEDD is based on feature extraction and feature weighting to handle concept drift in an explicit way, being trustworthy and transparent to users. FW-FEDD implements a forecasting module composed by a pool of forecasting models in which each model is specialized in a different time series concept. Several computational experiments on both artificial and real-world time series showed that the proposed method is able to improve the concept drift detection accuracy compared to methods based on monitoring raw time series observations and residual-based methods. Results also showed the superiority of FW-FEDD compared to other passive and active adaptive learning systems in terms of forecasting performance.

Palavras-chave: Adaptive learning systems. Data streams. Concept drift. Time series forecasting. Time series features.
  • © Centro de Informática UFPE - Todos os direitos reservados
    Tel +55 81 2126.8430 - Cidade Universitária - 50740-560 - Recife/PE
Plano4 Consultoria Web