Title of article
A tabusearchwithanoscillationstrategyforthediscriminant analysis problem
Author/Authors
Arnaud Fre´ ville ، نويسنده , , Sa?¨d Hanafi ، نويسنده , , Fre´de´ ric Semet ، نويسنده , , NicolasYanev ، نويسنده ,
Issue Information
ماهنامه با شماره پیاپی سال 2010
Pages
9
From page
1688
To page
1696
Abstract
This articleproposesatabusearchapproachtosolveamathematicalprogrammingformulationofthe
linear classificationproblem,whichconsistsofdetermininganhyperplanethatseparatestwogroupsof
points aswellaspossiblein Rm. Thetabusearchapproachproposedisbasedonanon-standard
formulationusinglinearsysteminfeasibility.Thesearchspaceisthesetofbasesdefinedonthematrix
that describesthelinearsystem.Themovesareperformedbypivotingonaspecifiedrowandcolumn.
On realmachinelearningdatabases,ourapproachcomparesfavorablywithimplementationsbasedon
parametricprogrammingandirreducibleinfeasibleconstraintsets.Additionalcomputationalresultsfor
randomlygeneratedinstancesconfirmthatourmethodprovidesasuitablealternativetothemixed
integerprogrammingformulationthatissolvedbyacommercialcodewhenthenumberofattributesm
increases.
Keywords
discriminant analysis , Linear classification problem , Tabu search
Journal title
Computers and Operations Research
Serial Year
2010
Journal title
Computers and Operations Research
Record number
927775
Link To Document