1. INTRODUCTION
Feature extraction is a technique that describes a large set of data by utilizing minimal amount of resources. It builds data to be informative and non-redundant, facilitating the subsequent learning and generalization steps, and leading to better human interpretations.
When the input is suspected to be redundant and too large to be processed then it can be transformed into a minimized set of features .The selected features contain the relevant data from the input values, so that the preferred work can be done by using this method instead of the whole data.
Feature extraction is a method of constructing combinations of the variables to get the data with sufficient accuracy. It is also called dimensionality reduction and are used such as ISO map, Multifactor dimensionality reduction, Independent component analysis, Kernel Principle Component Analysis, Nonlinear dimensionality reduction, Latent semantic analysis, Partial least squares, PCA (Principal component analysis), Multi-linear Principal Component Analysis, Multi-linear subspace learning, Semi definite embedding and Auto encoder. On important area where feature extraction can be applied is image processing. There are also software packages targeting machine learning applications that focus in feature extraction.