BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:e917eabc492f8df606cc4d203e73b0f4
DTSTAMP:20241105T002102Z
SUMMARY:Nikos Kourentzes\, PhD\, University of Skövde\, Sweden 
DESCRIPTION: \n\nAbstract: The use of multiple temporal aggregation levels 
 of a time\nseries has gained some momentum for improving forecast accuracy
  and\nidentifying the structure of a time series. In this talk\, we review
 \nthe main ideas behind its use. So far\, there are two main approaches\nt
 o do this\, the first using a forecast combination view\, namely the\nMult
 iple Temporal Aggregation Algorithm\; and the second using\nhierarchical f
 orecasting\, namely Temporal Hierarchies. While both have\ndemonstrated be
 nefits\, the latter is more flexible\, overcoming many of\nthe limitations
  of the former\, importantly allowing the use of\nforecasts from any sourc
 e. Yet\, in terms of forecast accuracy\, both\nremain competitive. We show
  that reconciling the two views leads to a\nmodified version of Temporal H
 ierarchies that substantially improves\nover the standard implementation. 
 Furthermore\, we show that the\nstandard implementation violates helpful c
 onstraints established in\nthe forecast combination literature while impos
 ing unhelpful\nconstraints inherited from hierarchical forecasting. Based 
 on the\nmodified Temporal Hierarchies\, we identify conditions that using\
 ntemporal hierarchies is admissible to standard forecasts. Finally\, we\nc
 onnect the hierarchical structure to the organizations’ context and\nsho
 w how this impacts the evaluation of forecasts.\n 
DTSTART:20211122T170000Z
DTEND:20211122T183000Z
LOCATION:Gerri C. LeBow Hall\, 3220 Market Street\, 722\, Philadelphia\, PA
  19104
END:VEVENT
END:VCALENDAR