Nonlinear Time Series Analyses in Industrial Environments and Limitations for Highly Sparse Data

This source preferred by Emili Balaguer-Ballester

Authors: Balaguer-Ballester, E.

Volume: 18

Publisher: Springer

This data was imported from Scopus:

Authors: Balaguer-Ballester, E.

Volume: 18

Pages: 51-60

ISBN: 9783642342189

DOI: 10.1007/978-3-642-34219-6_6

This work presents case studies of effective knowledge transfer in projects that focused on using nonlinear time series analyses in varied industrial settings. Applications, characterized by intricate dynamical processes, ranged from e-commerce to predicting services request in support centres. A common property of these time series is that they were originated by nonlinear and potentially high-dimensional systems in weakly stationary environments. Therefore, large amount of data was typically required for providing useful forecasts and thus a successful transfer of knowledge. However, in certain scenarios, classifications or predictions have to be inferred from time windows containing only few relevant patterns. To address this challenge, we suggest here the combined use of statistical learning and time series reconstruction algorithms in industrial domains where datasets are severely limited. These ideas could entail a successful transfer of knowledge in projects were more traditional data mining approaches may fail. © Springer-Verlag Berlin Heidelberg 2013.

The data on this page was last updated at 05:12 on February 26, 2020.