TECHNIQUES FOR HIGH-DIMENSIONAL DATA REDUCTION

Authors

  • Dr. Syed Waqar ul Qounain Jaffry Professor of Artificial Intelligence & Director, National Centre of Artificial Intelligence, Department of Information Technology, University of the Punjab, Lahore, Pakistan.

DOI:

https://doi.org/10.71465/csb48

Keywords:

High-dimensional data, dimensionality reduction, Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), t-SNE, Auto encoders, machine learning, data visualization

Abstract

 

High-dimensional data, characterized by an increased number of features or variables, poses significant challenges in data analysis due to the curse of dimensionality, leading to overfitting, computational inefficiencies, and interpretability issues. This paper explores various techniques for reducing the dimensionality of datasets while retaining as much meaningful information as possible. These techniques are essential for improving the performance of machine learning models and for enhancing data visualization. We examine traditional approaches such as Principal Component Analysis (PCA), Linear Discriminant Analysis (LDA), and more recent methods such as t-Distributed Stochastic Neighbour Embedding (t-SNE), Auto encoders, and Deep Learning-based techniques. The advantages, disadvantages, and applications of each technique are discussed with real-world examples. Furthermore, we propose a hybrid model that combines the strengths of these techniques for optimal dimensionality reduction.

image

Downloads

Published

2020-06-30