Feature Engineering and Classifier Selection: A Case Study in Venusian Volcano Detection

Abstract

As machine learning has graduated from toy problems to "real world" applications, users are finding that "real world" problems require them to perform aspects of problem solving that are not currently addressed by much of the machine learning literature. Specifically, users are finding that the tasks of selecting a set of features to define a problem and obtaining a set of examples of the problem are often more important for a successful machine learning application than the selection or development of a specific classification method. In this paper we present a case study of machine learning applied to a difficult "real world" problem: detecting volcanos in SAR (synthetic aperture radar) images of Venus from the Magellan dataset. Our work demonstrates that the processes of feature selection and sample collection are critical to the production of a good classifier. We further show that the use of domain dependent knowledge can often serve to enhance the resulting ...