Performance Analysis of Deep Autoencoder and NCA Dimensionality Reduction Techniques with KNN, ENN and SVM Classifiers

TL;DR

This study compares Deep Autoencoder and NCA for dimensionality reduction on diverse datasets, evaluating their compatibility with KNN, ENN, and SVM classifiers in MATLAB to determine optimal combinations.

Contribution

It provides a comparative analysis of Deep Autoencoder and NCA techniques with multiple classifiers across nine datasets, highlighting their effectiveness and compatibility.

Findings

Deep Autoencoder and NCA reduce dataset dimensions by 50%.

Classifier accuracy varies with different dimensionality reduction techniques.

Certain combinations outperform others in classification accuracy.

Abstract

The central aim of this paper is to implement Deep Autoencoder and Neighborhood Components Analysis (NCA) dimensionality reduction methods in Matlab and to observe the application of these algorithms on nine unlike datasets from UCI machine learning repository. These datasets are CNAE9, Movement Libras, Pima Indians diabetes, Parkinsons, Knowledge, Segmentation, Seeds, Mammographic Masses, and Ionosphere. First of all, the dimension of these datasets has been reduced to fifty percent of their original dimension by selecting and extracting the most relevant and appropriate features or attributes using Deep Autoencoder and NCA dimensionality reduction techniques. Afterward, each dataset is classified applying K-Nearest Neighbors (KNN), Extended Nearest Neighbors (ENN) and Support Vector Machine (SVM) classification algorithms. All classification algorithms are developed in the Matlab environment. In each classification, the training test data ratio is always set to ninety percent: ten percent. Upon classification, variation between accuracies is observed and analyzed to find the degree of compatibility of each dimensionality reduction technique with each classifier and to evaluate each classifier performance on each dataset.