Course Time Series Analysis and Forecasting (April 5, 7 and 12 2017)

Date:
05 April
Time:
00:00
Location:
Tu/e
Subscription:
From 09 March
Subscribe for this event

This spring Eindhoven University of Technology organizes, in cooperation with PAO-TM, two post-graduate courses relevant to the field of data science.

Time series occur in a wide range of disciplines, ranging from business, economics and social sciences to biomedical and engineering contexts. In analyzing time series one searches for structures and patterns to describe and explain the underlying process and to forecast, based on adequate models fitted, future values or to predict results from alternative scenarios. In the course “Time series analysis and forecasting”, apart from the “traditional” methods for trend and seasonal decomposition of time series (eg. Holt-Winter exponential smoothing models), more advanced statistical techniques available for these tasks, both in the time-domain (eg.Box-Jenkins ARMA-models) and in the frequency domain (eg. spectral and periodogram analyses) are discussed and underlying principles are explained. Furthermore attention is paid to the analysis of multivariate time series that are cross-correlated (Transfer function models and XARIMA models). The use of the representative statistical software R, is demonstrated and participants get the opportunity for hands-on experience in analyzing and forecasting time series.

Target Group

This course aims at people who have to analyse and predict time series data: data that are collected sequentially over time. The course is also suitable for teachers at universities and HBO.

Experience Level

Academic or HBO level, or equivalent level of knowledge gained by experience. Knowledge of basic statistical techniques like testing and regression modeling is assumed.

Results

After successful completion of the course participants have gained insight and experience with current approaches for time series analysis, modeling and forecasting. More specifically this holds for exponential smoothing models (Simple, Holt and Holt-Winter), for Box-Jenkins models (ARMA, ARIMA, SARIMA) and for multivariate time series models (transfer function and XARIMA-models). Furthermore, participants should be able to analyze, model and validate time series data with the representative statistical software R, independently and use the models obtained for time series forecasting and scenario analysis.

For more information