Classification Preservation Using Assorted Dimensionality Reduction Techniques

• Author(s): Usman A. Baba ; Augustine S. Nsang
• Paper ID: 1704964
• Page: 245-257
• Published Date: 28-08-2023
• Published In: Iconic Research And Engineering Journals
• Publisher: IRE Journals
• e-ISSN: 2456-8880
• Volume/Issue: Volume 7 Issue 2 August-2023

Abstract

In this paper, we implement the perceptron classification algorithm and apply it to three two-class datasets which include the student, weather and ionosphere datasets. Then the k-Nearest Neighbors classification algorithm is also applied to the same two-class datasets. Each dataset is then reduced using fourteen different dimensionality reduction techniques. The perceptron and k-nearest neighbor classification algorithms are then applied to each reduced set and the performances of the dimensionality reduction techniques in preserving the classification of a dataset by the k-nearest neighbors and perceptron classification algorithm are compared. The extent to which the classification of a dataset is preserved by a given dimensionality reduction technique is evaluated using the rand index and confusion matrices.

Keywords

Classification, Confusion Matrix, Dimensionality Reduction, Eager Learner, k-Nearest Neighbors, Lazy Learner, Perceptron, Rand Index

Citations

IRE Journals:
Usman A. Baba , Augustine S. Nsang "Classification Preservation Using Assorted Dimensionality Reduction Techniques" Iconic Research And Engineering Journals Volume 7 Issue 2 2023 Page 245-257

IEEE:
Usman A. Baba , Augustine S. Nsang "Classification Preservation Using Assorted Dimensionality Reduction Techniques" Iconic Research And Engineering Journals, 7(2)