Bagging Exponential Smoothing Methods using STL Decomposition and Box-Cox Transformation

Published on 21 March 2014 in Working papers

Christoph Bergmeir1, Rob J Hyndman2, José M Benítez1 Department of Computer Science and Artificial Intelligence, University of Granada, Spain. Department of Econometrics and Business Statistics, Monash University, Australia. Abstract: Exponential smoothing is one of the most popular forecasting methods. We present a method for bootstrap aggregation (bagging) of exponential smoothing methods. The bagging uses a Box-Cox transformation followed by an STL decomposition to separate the time series into trend, seasonal part, and remainder. The remainder is then bootstrapped using a moving block bootstrap, and a new series is assembled using this bootstrapped remainder. On the bootstrapped series, an ensemble of



Automatic time series forecasting

Published on 13 February 2014 in Talks

Talk presented at the conference “New Trends on Intelligent Systems and Soft Computing 2014″, University of Granada, Spain. 13-14 February 2014.


Boosting multi-step autoregressive forecasts

Souhaib Ben Taieb and Rob J Hyndman International Conference on Machine Learning (ICML) 2014, to appear. Abstract Multi-step forecasts can be produced recursively by iterating a one-step model, or directly using a specific model for each horizon. Choosing between these two strategies is not an easy task since it involves a trade-off between bias and estimation variance over the forecast horizon. Using a nonlinear machine learning model makes the tradeoff even more difficult. To address this issue, we propose a new forecasting strategy which boosts traditional recursive linear forecasts with a direct strategy using a boosting autoregression procedure at each



Efficient identification of the Pareto optimal set

Ingrida Steponavičė1, Rob J. Hyndman2, Kate Smith-Miles1 and Laura Villanova3 School of Mathematical Sciences, Monash University, Clayton, Australia Department of Econometrics & Business Statistics, Monash University, Clayton, Australia Ceramic Fuel Cells Limited, Noble Park, Australia Learning and Intelligent OptimizatioN Conference LION 8, Gainesville, Florida – USA, Feb 16-21, 2014 Abstract. In this paper, we focus on expensive multiobjective optimization problems and propose a method to predict an approximation of the Pareto optimal set using classification of sampled decision vectors as dominated or nondominated. The performance of our method, called EPIC, is demonstrated on a set of benchmark problems used in



Two-dimensional smoothing of mortality rates

Alexander Dokumentov and Rob J Hyndman Abstract: We propose three new practical methods of smoothing mortality rates (the procedure known in demography as graduation) over two dimensions: age and time. The first method uses bivariate thin plate splines. The second uses a similar procedure but with lasso-type regularization. The third method also uses bivariate lasso-type regularization, but allows for both period and cohort effects. Thus the mortality rates are modelled as the sum of four components: a smooth bivariate function of age and time, smooth one-dimensional cohort effects, smooth one-dimensional period effects and random errors. Cross validation is used to compare these new methods of graduation with