Tutorials
The role of the tutorials is to provide a platform for a more intensive scientific exchange amongst researchers interested in a particular topic and as a meeting point for the community. Tutorials complement the depth-oriented technical sessions by providing participants with broad overviews of emerging fields. A tutorial can be scheduled for 1.5 or 3 hours.
Tutorial on
Introduction to the Data Analysis of Time Series
Instructor
|
Torsten Ullrich
Fraunhofer Austria Research GmbH
Austria
|
|
|
|
Brief Bio
Torsten Ullrich is a researcher at Fraunhofer Austria Research GmbH and at the Institute of Computer Graphics and Knowledge Visualization of Graz University of Technology.
He studied mathematics at the University Karlsruhe (TH) and received his Ph.D. on “Reconstructive Geometry” in computer science from Graz University of Technology, Austria in 2011. His main research areas are visual computing in combination with numerical optimization, statistics, and data analysis. He has been the project coordinator for various research projects. Currently, he is the Deputy Head of the business area Data-Driven Design of Fraunhofer Austria Research GmbH, where he is responsible for scientific research coordination.
|
Abstract
For time series, a variety of statistical methods exist to calculate future values and to make reliable forecasts. In this context, both the trend development and the extreme values (minima and maxima) are of interest.
In this tutorial, statistical and machine learning methods for modeling will be presented and used for forecasting. A comparison of these methods shall give an overview and be a general decision support, which provides a recommendation for action for own data sets.
Keywords
introduction course; time series; data analysis; machine learning
Aims and Learning Objectives
Many introductions to the topic focus on a “hands-on” approach, which unfortunately corresponds often to a “trial and error” approach. There, programming languages and libraries are introduced, but the understanding of a statistical method is frequently reduced to the function call of a software library.
In this tutorial the methods are in the center. The knowledge about the individual methods, the modeling possibilities, and the limitations, permit in practice a purposeful method selection and a purposeful procedure. The main goals of the tutorial are
(1) to provide an overview of the methods and
(2) a recommendation of what can be used when.
Target Audience
The tutorial is intended for users of data analysis with time-dependent data. Possible applica-tions include (but are not limited to) economics (analysis of financial developments), com-puter science and engineering (time of failure analysis), medicine (evaluation of long-term studies), etc.
Prerequisite Knowledge of Audience
The tutorial assumes mathematical basics as they are part of all STEM (science, technology, engineering, and mathematics) disciplines.
Detailed Outline
The tutorial starts with elementary techniques such as simple linear regression and then in-troduces more complicated models such as autoregressive moving average (ARMA). The comparison of individual methods and their discussion form the focus of the tutorial. An out-look on current machine learning methods completes the tutorial.