Data Collection, Preparation, Labeling, and Input Coding

doi:10.1117/3.633187.ch4

Book: Artificial Neural Networks: An Introduction

Author(s): Kevin L. Priddy, Paul E. Keller

Published: 2005

https://doi.org/10.1117/3.633187.ch4

Author Affiliations +

Abstract

Since neural networks are data driven, the adage "garbage in, garbage out" is highly relevant to the task of building a neural network. Proper collection, preparation, labeling, and coding of the data can make the difference between a successful and unsuccessful experience with neural networks. While the process of collecting data seems simple, the network designer should put some thought into the data-collection process. The designer needs to decide what he wants the neural network to do and what data requirements are needed to train the network. Will it be a classifier, an estimator (modeler), or a self-organizer (clusterer)? The designer needs to determine how and from where to obtain the data and what types of data to collect. He must also determine what the neural network will output in response to the data used as the network input. The steps in a typical data-collection plan are described next.