Refining Face Recognition: Maximizing Performance with Dimensionality Reduction and Ensemble Learning in K-Nearest Neighbors
Chethan TS -, Abheesh Puthukkudy - · International Journal For Multidisciplinary Research · 2024
In the field of image classification, the K-Nearest Neighbors (KNN) algorithm is favored for its simplicity and effectiveness. However, the high dimensionality of image data often challenges KNN’s performance. This study investigates the impact of three dimensionality reduction techniques—Principal Component Analysis (PCA), Uniform Manifold Approximation and Projection (UMAP), and Linear Discriminant Analysis (LDA)—on enhancing KNN’s accuracy and efficiency in image classification, where KNN and Bagging classifier is computed without using any library but only using mathematical formulation. Additionally, the study examines the effect of ensemble learning, specifically through bagging, on KNN's performance.