?url_ver=Z39.88-2004&rft_val_fmt=info%3Aofi%2Ffmt%3Akev%3Amtx%3Adc&rft.title=Combining+Neural+Network+Forecasts+on+Wavelet-Transformed+Time+Series&rft.creator=Aussem%2C+Alex&rft.creator=Murtagh%2C+Fionn&rft.subject=Artificial+Intelligence&rft.subject=Dynamical+Systems&rft.subject=Machine+Learning&rft.subject=Neural+Nets&rft.subject=Speech&rft.subject=Statistical+Models&rft.description=We+discuss+a+simple+strategy+aimed+at+improving+neural+network+prediction+accuracy%2C+based+on+the+combination+of+predictions+at+varying+resolution+levels+of+the+domain+under+investigation+(here%3A+time+series).+First%2C+a+wavelet+transform+is+used+to+decompose+the+time+series+into+varying+scales+of+temporal+resolution.+The+latter+provide+a+sensible+decomposition+of+the+data+so+that+the+underlying+temporal+structures+of+the+original+time+series+become+more+tractable.+Then%2C+a+Dynamical+Recurrent+Neural+Network+(DRNN)+is+trained+on+each+resolution+scale+with+the+temporal-recurrent+backpropagation+(TRBP)+algorithm.+By+virtue+of+its+internal+dynamic%2C+this+general+class+of+dynamic+connectionist+network+approximates+the+underlying+law+governing+each+resolution+level+by+a+system+of+nonlinear+difference+equations.+The+individual+wavelet+scale+forecasts+are+afterwards+recombined+to+form+the+current+estimate.+The+predictive+ability+of+this+strategy+is+assessed+with+the+sunspot+series.&rft.date=1997&rft.type=Journal+(Paginated)&rft.type=PeerReviewed&rft.format=application%2Fpostscript&rft.identifier=http%3A%2F%2Fcogprints.org%2F551%2F2%2FConnectScience.ps&rft.identifier=++Aussem%2C+Alex+and+Murtagh%2C+Fionn++(1997)+Combining+Neural+Network+Forecasts+on+Wavelet-Transformed+Time+Series.++%5BJournal+(Paginated)%5D+++++&rft.relation=http%3A%2F%2Fcogprints.org%2F551%2F