|
Title:
|
TEMPUS: A PROTOTYPE SYSTEM FOR TIME SERIES ANALYSIS AND PREDICTION |
|
Author(s):
|
Tim Schlüter, Stefan Conrad |
|
ISBN:
|
978-972-8939-23-6 |
|
Editors:
|
António Palma dos Reis and Ajith P. Abraham |
|
Year:
|
2010 |
|
Edition:
|
Single |
|
Keywords:
|
Knowledge discovery in databases; temporal data mining; time series analysis and prediction; Hidden Markov Models; Derivative Dynamic Time Warping; river l |
|
Type:
|
Full Paper |
|
First Page:
|
12 |
|
Last Page:
|
19 |
|
Language:
|
English |
|
Cover:
|
|
|
Full Contents:
|
click to dowload 
|
|
Paper Abstract:
|
Time series occur in nearly every area of life, e.g. in medicine, astronomy, geology. Understanding their nature and predicting their future development can be of great importance for numberless applications, among others for epilepsy seizure prediction, sunspot detecting and floodwater prediction. Therefore in this paper we present TEMPUS, a prototype system for the analysis and prediction of time series. TEMPUS is capable of analyzing and predicting time series of several origins, which is demonstrated by predicting water levels in rivers. The prediction of water levels is of great interest for boat transportation, floodwater forecast and many other applications. TEMPUS analyzes time series consisting of river level measurements of numerous stations, detects correlated time series by using Derivative Dynamic Time Warping and predicts future river levels by means of a Hidden Markov Model established on base of the former detected correlated time series. We describe the foundation of TEMPUS and present an evaluation on a large dataset provided by the Federal Institute for Nature, Environment and Consumerism of North Rhine-Westphalia, Germany, which proves TEMPUS functionality |
|
|
|
|
|
|
