A case study by using Python to implement data and dimensionality reduction

Abstract

The purpose of this study is to research and explore the Data Dimension, and propose the data feature & selection of dimensionality reduction technique, in order to help users understand the impact and meaning between dimensionality reduction parameters and data dimension, thereby strengthening the use of dimension reduction algorithm. In previous studies, many scholars have proposed dimensionality reduction algorithms for various data types, such as Multi-Dimensional Scaling (MDS), Linear Discriminant Analysis (LDA), Principal Component Analysis (PCA), Facet Analysis (FA), Isometric Feature Maps (Isomap, using for manifold analysis), Local Linear Embedding (LLE), and Laplacian feature maps (Laplacian Eigenmaps). Most of these algorithms do not need to set parameters, and it has been obtained during the experiment that the selection of parameters has no visual analysis effect on the dataset in this experiment, and should be determined according to the feature of the dataset. This study is conducted by comparing the most used PCA and LDA dimensionality reduction techniques, as well as the analysis of merging other similarity methods while using MDS to process mixed data.