|
|
Classification analysis
Rensová, Dita ; Kalina, Jan (advisor) ; Juríček, Jozef (referee)
Xazev pracc: Kiasifikacni analy/a Autor: Dita. Rensova Kak'dra: Katedra. pravdepodobnost i a ma.tema.ticke statist iky Vedouci ba.kalarske prace: Ur.rer.nat. Ja.n Kaliua e-mail vedoucfho; kalina n-karlin.inff.cuni.c7 Al)stvakt: V toto pnici sc- zabvvame modely klasifikacm analyzy. Popiseme jednotliva klasiiikacni pravidla, a son\islosti mexi niini. Nejprve sc zaniefhne na modely line-arm a kvadraticke klasifikaee pru pripad dvou sknpin. kterr dak1 xobccninu: na, linearni a. kvadi'at.icke inodcly pro pn'pad kla,sifikae.e do vice skupin. Pole so bndeine /abyvat pra\xlepodobnosti s[>a.tnc klasifikaeo nrcit(''ho objektu do sknpiny a.mettxlami, jak Into pravdepodobnost odhad- nont. Dale se zmhn'nie o vyu/iti diskriniinacnieh skc'irii pri kiasifikaei a so- ziianiune se s modoloni logist.ieke kla.sifika.ee. Na zaver pfetlvedi^iuj ponzitf nekt.erych vy!)ranyeli modclii na, konkn'M nieli dateeh / oborn lesnictvi. Klieova slova: Lincarni klasifikace. kvadraticka klasilikacc, lo^ist.icka klasi- fikace. diskrhninacc a klasifikace Title: Cla.ssifica.t ion analysis Author; Uita Hensova Depaituiont,: Department ol Probability and Mathematical Statistics Supervisor: IJr.rer.nat. .Jan Kalina Supervisor's e-mail address: kalina:(i;karhn.mfl.cuni.c/ Abstract.: In the- present work we study methods for classification analysis....
|
|
|
Robust classification and discrimination
Rensová, Dita ; Kalina, Jan (advisor) ; Jonáš, Petr (referee)
This thesis is focused on classification methods and their robust alternatives. First, we recall the standard classification rules of linear and quadratic discrim- ination analysis. We also show some methods for estimating their probability of missclassification. Next we describe some existing robust multivariate estimators, their properties and computational algorithms. These estimators are consequently used to construct robust classification rules. Then, we describe the principal com- ponent analysis as a technique for dimension reduction. Again, we study methods for its robustification. Finally, we illustrate the usage of robust classification on both numerical simulations and real data. We also investigate the influence of the principal component analysis on classification results.
|
|
|
Classification analysis
Rensová, Dita ; Juríček, Jozef (referee) ; Kalina, Jan (advisor)
Xazev pracc: Kiasifikacni analy/a Autor: Dita. Rensova Kak'dra: Katedra. pravdepodobnost i a ma.tema.ticke statist iky Vedouci ba.kalarske prace: Ur.rer.nat. Ja.n Kaliua e-mail vedoucfho; kalina n-karlin.inff.cuni.c7 Al)stvakt: V toto pnici sc- zabvvame modely klasifikacm analyzy. Popiseme jednotliva klasiiikacni pravidla, a son\islosti mexi niini. Nejprve sc zaniefhne na modely line-arm a kvadraticke klasifikaee pru pripad dvou sknpin. kterr dak1 xobccninu: na, linearni a. kvadi'at.icke inodcly pro pn'pad kla,sifikae.e do vice skupin. Pole so bndeine /abyvat pra\xlepodobnosti s[>a.tnc klasifikaeo nrcit(''ho objektu do sknpiny a.mettxlami, jak Into pravdepodobnost odhad- nont. Dale se zmhn'nie o vyu/iti diskriniinacnieh skc'irii pri kiasifikaei a so- ziianiune se s modoloni logist.ieke kla.sifika.ee. Na zaver pfetlvedi^iuj ponzitf nekt.erych vy!)ranyeli modclii na, konkn'M nieli dateeh / oborn lesnictvi. Klieova slova: Lincarni klasifikace. kvadraticka klasilikacc, lo^ist.icka klasi- fikace. diskrhninacc a klasifikace Title: Cla.ssifica.t ion analysis Author; Uita Hensova Depaituiont,: Department ol Probability and Mathematical Statistics Supervisor: IJr.rer.nat. .Jan Kalina Supervisor's e-mail address: kalina:(i;karhn.mfl.cuni.c/ Abstract.: In the- present work we study methods for classification analysis....
|
guest ::
