The Effect of Dimensionality Reduction on Machine Learning Models Performance: A Review

Abstract

The performance of Machine learning models are fully dependent on the training data. In today’s age where many companies transforming their businesses towards digitalization. A huge volume of raw data is generated and collected, extracting important and useful features from raw data is quite a big challenge, because it is computationally very expensive and time-consuming to train machine learning models on raw data and result in poor model’s performance due to irrelevant features. So, it is necessary to filter out data by reducing the size of features or dimensions and extracting important features from a huge amount of data before implementing any machine learning model on it. In this situation, Dimensionality reduction techniques come into play to eliminate unnecessary or irrelevant features from the data set. In this paper, we are going to review some of the important dimensionality reduction techniques on different machine learning models and evaluate the performance of how dimensionality reduction techniques improve the overall performance of models. After reviewing the research work of different authors, we conclude our results by comparing different implemented techniques of dimensionality reduction, the results before reducing the size of features/dimensions, and after implementing dimensionality reduction techniques and figure out key findings and limitations of reviewed papers.

Conflict of Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Funding

The  research received no  specific  grant  from  any  funding  agency  in  the  public, commercial, or not-for-profit sectors.

Data Fabrication/Falsification Statement

The author(s) declare that no data have been fabricated, falsified, or manipulated in this study.

Participant Consent

The authors confirm that Informed consent was obtained from all participants, and confidentiality was duly maintained.

Copyright and Licensing

For all articles published in the NIJEC journal, Copyright (c) of this study is with author(s).