Multiple-output modeling for multi-step-ahead time series forecasting

Accurate prediction of time series over long future horizons is the new frontier of forecasting. Conventional approaches to long-term time series forecasting rely either on iterated one-step-ahead predictors or direct predictors. In spite of their diversity, iterated and direct techniques for multi-step-ahead forecasting share a common feature, i.e. they model from data a multiple-input single-output mapping. In previous works, the authors presented an original multiple-output approach to multi-step-ahead prediction. The goal is to improve accuracy by preserving in the forecasted sequence the stochastic properties of the training series. This is not guaranteed for instance in direct approaches where predictions for different horizons are performed independently. This paper presents a review of single-output vs. multiple-output approaches for prediction and goes a step forward with respect to the previous authors contributions by (i) extending the multiple-output approach with a query-based criterion and (ii) presenting an assessment of single-output and multiple-output methods on the NN3 competition datasets. In particular, the experimental section shows that multiple-output approaches represent a competitive choice for tackling long-term forecasting tasks.

[1]  David M. Allen,et al.  The Relationship Between Variable Selection and Data Agumentation and a Method for Prediction , 1974 .

[2]  Mauro Birattari,et al.  Lazy Learning Meets the Recursive Least Squares Algorithm , 1998, NIPS.

[3]  Ronald J. Williams,et al.  A Learning Algorithm for Continually Running Fully Recurrent Neural Networks , 1989, Neural Computation.

[4]  Michel Verleysen,et al.  Using the Delta Test for Variable Selection , 2008, ESANN.

[5]  J. D. Farmer,et al.  State space reconstruction in the presence of noise" Physica D , 1991 .

[6]  James McNames,et al.  A Nearest Trajectory Strategy for Time Series Prediction , 2000 .

[7]  Gianluca Bontempi,et al.  Local learning techniques for modeling, prediction and control , 2000 .

[8]  E. Lorenz.  Atmospheric Predictability as Revealed by Naturally Occurring Analogues , 1969 .

[9]  Jing Gao,et al.  Multistep-Ahead Time Series Prediction , 2006, PAKDD.

[10]  Amaury Lendasse,et al.  Long-term prediction of time series by combining direct and MIMO strategies , 2009, 2009 International Joint Conference on Neural Networks.

[11]  Johan A. K. Suykens,et al.  Lazy learning for iterated time-series prediction , 1998 .

[12]  Amaury Lendasse,et al.  Time series prediction using DirRec strategy , 2006, ESANN.

[13]  Kazuyuki Aihara,et al.  Prediction of Chaotic Time Series with Noise , 1995 .

[14]  M. Birattari,et al.  Lazy learning for local modelling and control design , 1999 .

[15]  M. Bayazit,et al.  The Power of Statistical Tests for Trend Detection , 2003 .

[16]  Gianluca Bontempi,et al.  Long Term Time Series Prediction with Multi-Input Multi-Output Local Learning , 2008 .

[17]  Mauro Birattari,et al.  Local Learning for Iterated Time-Series Prediction , 1999, ICML.

[18]  J. Doyne Farmer,et al.  Exploiting Chaos to Predict the Future and Reduce Noise , 1989 .

[19]  Andrew W. Moore,et al.  Locally Weighted Learning , 1997, Artificial Intelligence Review.

[20]  Mauro Birattari,et al.  Lazy learning for modeling and control design , 1997 .

[21]  Andreas S. Weigend,et al.  Time Series Prediction: Forecasting the Future and Understanding the Past , 1994 .

[22]  J. Friedman,et al.  Predicting Multivariate Responses in Multiple Linear Regression , 1997 .

[23]  Amaury Lendasse,et al.  Direct and Recursive Prediction of Time Series Using Mutual Information Selection , 2005, IWANN.

[24]  Amaury Lendasse,et al.  Methodology for long-term prediction of time series , 2007, Neurocomputing.

[25]  Jianqing Fan.  Nonlinear Time Series , 2003 .

[26]  Farmer,et al.  Predicting chaotic time series. , 1987, Physical review letters.

[27]  Jianqing Fan,et al.  Nonlinear Time Series : Nonparametric and Parametric Methods , 2005 .