PropertyValue
?:abstract
  • Undecided voters in pre-election polls, even though an increasing phenomenon and issue in electoral research, have mostly been neglected in conventional analysis so far. We argue to include this inherent form of uncertainty in a set-valued manner, in order to make the most of the valuable information, not improperly reducing voters’ response to either an spuriously precise answer or to drop outs. The resulting consideration set consists of all elements the individual is still pondering between and can be interpreted in two ways, depending on the question at hand. First, for the sake of forecasting, it can be seen as a coarse version of the yet unknown element the individual ends up choosing, using the information for so-called epistemic modeling. Second, from an so-called ontic view, it can be seen as entity of its own, representing the individual’s current position accurately and thus allowing to examine structural properties within the population. Both views provide good opportunities for machine learning. In this paper we introduce one exemplary approach based on each view, analysing structural properties using spectral clustering and forecasting using random forests, providing initial methodology for this type of complex, non-stochastic uncertainty. The theory is applied with constructed consideration sets to the most recent German federal election of 2017, using data from the German Longitudinal Election Study. The results are promising, laying the groundwork for further machine learning approaches concerning this natural type of inherent uncertainty. (xsd:string)
?:author
?:comment
  • (GLES) (xsd:string)
?:dataSource
  • GLES-Bibliography (xsd:string)
?:dateCreated
  • 11. Fassung, Dezember 2021 (xsd:gyear)
?:dateModified
  • 2020 (xsd:gyear)
?:datePublished
  • 2020 (xsd:gyear)
?:duplicate
?:fromPage
  • 1 (xsd:string)
is ?:hasPart of
?:inLanguage
  • english (xsd:string)
is ?:mainEntity of
?:name
  • Undecided voters as set-valued information – machine learning approaches under complex uncertainty (xsd:string)
?:publicationType
  • inproceedings (xsd:string)
?:reference
?:sourceCollection
  • ECML/PKDD 2020 Tutorial and Workshop on Uncertainty in Machine Learning, Virtual Conference, 14.-28.09.2020 (xsd:string)
?:sourceInfo
  • Bibsonomy (xsd:string)
  • In ECML/PKDD 2020 Tutorial and Workshop on Uncertainty in Machine Learning, Virtual Conference, 14.-28.09.2020, 1-12, 2020 (xsd:string)
?:studyGroup
  • German Longitudinal Election Study (GLES) (xsd:string)
?:tags
  • 2020 (xsd:string)
  • FDZ_Wahlen (xsd:string)
  • GLES (xsd:string)
  • GLES_input2021 (xsd:string)
  • GLES_pro (xsd:string)
  • GLES_version11 (xsd:string)
  • datfeld (xsd:string)
  • english (xsd:string)
  • inproceedings (xsd:string)
  • jak (xsd:string)
  • transfer21 (xsd:string)
  • vttrans (xsd:string)
?:toPage
  • 12 (xsd:string)
rdf:type