|
Activity Number:
|
598
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 6, 2009 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Quality and Productivity
|
| Abstract - #305847 |
|
Title:
|
Adaptive Break Detection and Combining: Application to Electricity Load Forecast
|
|
Author(s):
|
Yannig Goude*+
|
|
Companies:
|
EDF
|
|
Address:
|
, , ,
|
|
Keywords:
|
Online learning ; break detection and combining ; application to electricity load forecast
|
|
Abstract:
|
Online learning for a sequence of observations associated to a set of M individual forecasts is investigated when M can vary with time. This setting occurs in industrial applications when operational models face constant developments and new methods are introduced to improve industrial processes. Additionally, relative performances of individual forecasts according to the realizations can fluctuate with time due to changes in the data generating process inducing breaks in the combining process. We develop online combining algorithms based on break detection and exponential re-weighting that achieve good forecasting performances in this framework. These algorithms are applied to electricity load forecasts and compared to individual performances.
|
- The address information is for the authors that have a + after their name.
- Authors who are presenting talks have a * after their name.
Back to the full JSM 2009 program |
