Compression and Visualization of High-Dimensionality Data Using Auto-Associative Neural Networks

Interpreting the information hidden in multidimensional data can be considered as a challenging and also a complicated task. The compression, dimension reduction and visualization of these multidimensional data provide ways to better understanding and interpretation of the problem. Usually, dimension reduction or compression is considered as the first step to data analysis and exploration. Here, the focus is given on multidimensional data reduction using a supervised artificial neural networks technique namely the Auto-Associative Neural Networks (AANN). The AANN can be considered as very powerful tool in exploratory data analysis. It has the ability to deal with linear and nonlinear correlation among variables. This technique is often referred to as nonlinear principal component analysis (NLPCA) or sometimes is also known as the bottleneck neural network due to its specific structure that consists of a combination of two networks – compression and decompression. By using this structure, AANN can reduce high dimensional data onto lower dimensional data by compressing them on its bottleneck layer that later can be used for data visualization and interpretation. In this paper, the technique of AANN is described, developed and applied on two different multidimensional datasets. The results have shown that the AANN is able to compress multidimensional data into only two nonlinear principal components at its bottleneck layer and these compressed data can provide visualization of different clusters of data. Keywords—Auto-Associative Neural Networks, Dimension Reduction, Data Clustering, Iris flowers, Olive oils