Two-dimensional visualization of classification and regression problems. Automatic prediction of behavior from sensory data in autism spectrum disorder

Abstract

This thesis formulates methods to perform classification and regression by projecting high-dimensional patterns in two dimensions. These methods create a 2D classification or regression map to visualize the data as a political (for classification) or temperature (for regression) map, where each pixel in the map has an associated prediction. The thesis also uses 26 machine learning models for the automatic prediction of behavior in the treatment of autism spectrum disorder using sensory processing information. Behavior and sensory data are extracted from their respective questionnaires. Out of 11 behavior outcomes, the prediction of externalizing problems is very reliable and accurate enough in other 7 outcomes.