Chapter 4 - Context: Separating the forest and the trees—Wavelet contextual conditioning for AI

Abstract

This chapter discusses challenges related to artificial intelligence (AI) and contextual data used for data-driven decision support and presents discrete wavelet transformation (DWT) as a viable preprocessing step for dealing with contextual variability. Current methods often do not give ample consideration to preprocessing and typically couple AI algorithms tightly to the problem domain, making them brittle in varied contexts. DWT-based decomposition trees may provide a means of preprocessing training data while providing data reduction and separating data trends from contextual fluctuations. A method for selecting a preferred decomposition level is proposed, and both the DWT preprocessing augmentation and the preferred level selection methodology are evaluated through simulation.

Publication
Academic Press