Hierarchical Classification of Transposable Elements with a Weighted Genetic Algorithm

Most of the related works in Machine Learning (ML) are concerned with Flat Classification, in which an instance is often associated with one class within a small set of classes. However, in some cases, instances have to be assigned to many classes simultaneously, and these classes are arranged in a hierarchical structure. This problem, called Hierarchical Classification (HC), has received special attention in some fields, such as Bioinformatics. In this context, a topic that has gained attention is the classification of Transposable Elements (TEs), which are DNA fragments capable of moving inside the genome of their hosts. In this paper, we propose a novel hierarchical method based on Genetic Algorithms (GAs) that generates HC rules and classifies TEs in many hierarchical levels of its taxonomy. The proposed method is called Hierarchical Classification with a Weighted Genetic Algorithm (HC-WGA), and is based on a Weighted Sum approach to deal with the accuracy-interpretability trade-off, which is a common and still relevant problem in both ML and Bioinformatics. To the best of our knowledge, this is the first HC method to use such an approach. Experiments with two popular TEs datasets showed that our method achieves competitive results with most of the state-of-the-art HC methods, with the advantage of presenting an interpretable model.