High-Dimensional Diagonal Classifiers

Zinzendoff, Okwonu Friday

Ўзбекистон Миллий университети хабарлари · 2026-yil

Annotatsiya

The advancement in technology has necessitated the collection of massive high-dimensionaldata in bioinformatics, medical sciences, business, climatology, and magnetic resonance images. These collec-tions have given rise to high-dimensional (p) and small sample(n) (HDSS) problems in group classifications.The classification coefficient of the HDSS data cannot be constructed due to the ill-conditioning or singu-larity of the covariance matrix. The curse of dimensionality suffered by the classical methods has renderedthe formulation of the coefficient impossible as such different dimensionality reduction methods have beenproposed to avert the curse of dimensionality and ill conditioning. Dimensionality reduction methods suchas the principal component analysis have been applied as a prelude to reducing the dimension of the HDSSdata to enhance constructing the coefficients of any classification methods. Some of these dimension reduc-tion methods with infused constants as a benchmark selection, may increase or decrease computational timeand improve classification accuracy at the cost of information loss. In this paper, we proposed two methodsthat can solve the ill-conditioning and singularity, or dimensionality problems associated to HDSS data di-rectly. These methods are diagonal FLCM (DFLCM) and point extraction method (PEM). We investigatedthe robustness of these methods based on the optimal probability benchmark (OPB). The computationaltime (CPU time) of all the methods is also investigated. We applied three real data sets to determine thecomparative performance, and the classification results revealed that the proposed DFLCM is very robustmore than the NaГЇve Bayes classifier in terms of the probability of correct classification and CPU time.Therefore, this study affirmed that the OPB is suitable to determine the robustness of HDSS classificationproblems.

Maqola ma’lumotlari
MualliflarZinzendoff, Okwonu Friday
JurnalЎзбекистон Миллий университети хабарлари
NashriyotЎзбекистон Миллий университети
ISSN2181-7324, 3093-8996
Nashr sanasi2026-10-07
Jild2
Son2.3
Betlar169-177
TilIngliz
Ilmiy sohaFizika-matematika
Indeksga qo‘shilgan2026-10-08

Maqolaning asl sahifasi

Kalit so‘zlar

High-dimensional data, Classification, Probability, CPU time, Robustness.

Ilmiy soha

Fizika-matematika

Iqtibos keltirish

APAZinzendoff, Okwonu Friday (2026). High-Dimensional Diagonal Classifiers. Ўзбекистон Миллий университети хабарлари, 2(2.3), 169-177. https://journals.nuu.uz/index.php/actanuuz/article/view/15295
GOST (OAK)Zinzendoff, Okwonu Friday High-Dimensional Diagonal Classifiers // Ўзбекистон Миллий университети хабарлари. – 2026. – T. 2, № 2.3. – P. 169-177.
MLAZinzendoff, Okwonu Friday. “High-Dimensional Diagonal Classifiers.” Ўзбекистон Миллий университети хабарлари, 2(2.3), 169-177, 2026. https://journals.nuu.uz/index.php/actanuuz/article/view/15295
ChicagoZinzendoff, Okwonu Friday. “High-Dimensional Diagonal Classifiers.” Ўзбекистон Миллий университети хабарлари, 2(2.3), 169-177 (2026). https://journals.nuu.uz/index.php/actanuuz/article/view/15295
HarvardZinzendoff, Okwonu Friday (2026) ‘High-Dimensional Diagonal Classifiers’, Ўзбекистон Миллий университети хабарлари, 2(2.3), 169-177. https://journals.nuu.uz/index.php/actanuuz/article/view/15295
BibTeX
@article{zinzendoffokwonufriday2026,
  title = {High-Dimensional Diagonal Classifiers},
  author = {Zinzendoff, Okwonu Friday},
  journal = {Ўзбекистон Миллий университети хабарлари},
  year = {2026},
  volume = {2},
  number = {2.3},
  pages = {169-177},
  url = {https://journals.nuu.uz/index.php/actanuuz/article/view/15295}
}

Boshqa maqola uchun iqtibos tayyorlash

Ўзбекистон Миллий университети хабарлари jurnalidan boshqa maqolalar

Ўзбекистон Миллий университети хабарлари — barcha maqolalar · 2026-yil arxivi

Fizika-matematika sohasidagi boshqa maqolalar

Fizika-matematika sohasidagi OAK jurnallari