Weighted Cross-Validation Evolving ArtificialNeural Networks to Forecast Time Series

摘要

Accurate time series forecasting is a key tool to support decision mak-ing and for planning our day to-day activities. In recent years, several works in theliterature have adopted evolving artificial neural networks (EANN) for forecastingapplications. EANNs are particularly appealing due to their ability to model anunspecified non-linear relationship between time series variables. In this work, anovel approach for EANN forecasting systems is proposed, where a weightedcross-validation is used to build an ensemble of neural networks. Several experi-ments were held, using a set of six real-world time series (from different domains)and comparing both the weighted and standard cross-validation variants. Overall,the weighted cross-validation provided the best forecasting results.