Review of Dimension Reduction Methods

Purpose: This study sought to review the characteristics, strengths, weaknesses variants, applications areas and data types applied on various Dimension Reduction techniques. Methodology: The most commonly used databases employed search for papers were ScienceDirect, Scopus, Google Scholar, IEEE Xplore Mendeley. An integrative was where 341 reviewed. Results: linear techniques considered Principal Component Analysis (PCA), Linear Discriminant (LDA), Singular Value Decomposition (SVD), Latent Semantic (LSA), Locality Preserving Projections (LPP), Independent (ICA) Project Pursuit (PP). non-linear which developed work with that have complex structures Kernel (KPCA), Multi-dimensional Scaling (MDS), Isomap, Locally Embedding (LLE), Self-Organizing Map (SOM), Vector Quantization (LVQ), t-Stochastic neighbor embedding (t-SNE) Uniform Manifold Approximation Projection (UMAP). DR can further be categorized into supervised, unsupervised more recently semi-supervised learning methods. supervised versions are LDA LVQ. All other unsupervised. Supervised variants of PCA, LPP, KPCA MDS been developed. PP t-SNE also a semi version has Conclusion: application areas, explored. different