The Role of Domain Knowledge in Feature Selection for Machine Learning

Abstract

As a branch of artificial intelligence, machine learning focuses on using data and algorithms in order to make computers accomplish tasks without explicit programming. Feature selection is an essential phase in the machine learning process that has a huge impact on the performance of predictive models. Even though automated feature selection techniques are commonly used, incorporation of domain knowledge, an understanding and expertise in a specific area or field, in the process can bring about great improvement. This paper discusses the role domain knowledge plays in feature selection by showing how experts’ perspectives help in identifying meaningful features that enhance model accuracy and applicability. It also comprehensively discusses ways of acquiring domain knowledge and incorporating it in feature selection. The paper also discusses the challenges that accompany incorporating domain knowledge in feature selection.