Dimension reduction techniques: Current status and perspectives

During the past decades, there has been an overflow of information in terms of data collection and storage capacities in most of the fields. As a result, researchers involved in engineering, biology, remote sensing, economics and consumer transactions are most likely to face challenges in handling data with larger datasets to carry out simulations. Higher dimensional data sets give rise to new theoretical developments, but at the same time pose many mathematical challenges. At times, all the measured variables need not be important to understand the phenomenon of interest. This calls for reducing the dimension of the data, formally termed as "Dimensionality reduction". It refers to the method of reducing the number of input variables in the data set, so as to ease the task of predictive modelling. Dimension reduction may be linear or non-linear. In this paper, we present a survey of conventional as well as modern methods available for dimension reduction, with an emphasis on Principal Component Analysis and their applications in different fields of engineering, pertaining to production, additive manufacturing, earth science and data science. This survey would cater the need of retaining significant parameters, while disregarding unnecessary information to the future researchers and guide them further to use an appropriate method for dimension reduction to simplify the complexity of the problem.Copyright (c) 2021 Elsevier Ltd. All rights reserved. Selection and peer-review under responsibility of the scientific committee of the International Conference Additive Manufacturing and Advanced Materials-AM2 2021.