Evaluating a New Genetic Algorithm for Automated Machine Learning in Positive-Unlabelled Learning.

Positive-Unlabelled (PU) learning is a growing area of machine learning that aims to learn classifiers from data consisting of a set of labelled positive instances and a set of unlabelled instances, where the latter can be either positive or negative instances, but their label is unknown. There are many PU-learning algorithms, so an exhaustive search to find the best algorithm for a given dataset is computationally unfeasible. We recently proposed GA-Auto-PU, the first Genetic Algorithm-based Automated Machine Learning system for PU learning, and reported its preliminary results. This work presents an improved version of this system with an extended search space to include spy-based techniques, and provides an extensive evaluation of the new and previous versions of this system.