Feature selection for time series prediction – A combined filter and wrapper approach for neural networks

Abstract

Modelling artificial neural networks (NN) for accurate time series prediction poses multiple challenges, in particular specifying the network architecture in accordance with the underlying structure of the time series. The data generating processes may exhibit a variety of stochastic or deterministic time series patterns of single or multiple seasonality, trends and cycles, overlaid with pulses, level shifts and structural breaks, all depending on the discrete time frequency in which it is observed. For heterogeneous datasets of time series, such as the 2008 ESTSP competition, a universal methodology is required for automatic network specification across varying data patterns and time frequencies. We propose a fully data driven forecasting methodology that combines filter and wrapper approaches for feature evaluation, construction and transformation. The methodology identifies time series patterns, creates and transforms explanatory variables and specifies multilayer perceptrons for heterogeneous sets of time series without expert intervention. Examples of the valid and reliable performance in comparison to established benchmark methods are shown for a set of synthetic time series and for the ESTSP"08 competition dataset, where the proposed methodology obtained second place.

The impact of special days in call arrivals forecasting: A neural network approach to modelling special days

“…To select the hidden nodes we employ the methodology described by Crone and Kourentzes (2010) as extended by Kourentzes and Crone (2010), which is based on regression diagnostics to identify the relevant autoregressive inputs for the ANN. The two seasonal cycles of the call arrivals data are coded using pairs of trigonometric dummy variables:…”

A Feature Selection Method for Large-Scale Network Traffic Classification Based on Spark

“…The wrapper methods take the subsequent classification algorithm as the evaluation index of the feature subset; they achieve higher classification accuracy, but result in low efficiency and a large amount of computation. Thus, how to integrate the two methods to bridge the gap between the accuracy and the executing efficiency is a challenge in the research of feature selection [3].…”

Investigation of vehicle crash modeling techniques: theory and application

“…Apart from the strictly engineering and technical applications of the regressive models, there is a huge number of works taking advantage of this methodology but devoted to the other areas of research. In [28,29], one can find regressive models utilized, e.g., to forecast the stock market data-i.e., the weekly averaged exchange rates between the British pound and the US dollar in the given time period. A significant contribution was made in [30].…”

“…According to (28), for the model and crash test being analyzed in Section 3.1, we calculate the transition damping coefficient:…”