| Texto completo | |
| Autor(es): |
Ribeiro Alves, Andre Hallwas
;
Braga Ilidio Silva, Pedro de Carvalho
;
Cerri, Ricardo
Número total de Autores: 3
|
| Tipo de documento: | Artigo Científico |
| Fonte: | 38TH ANNUAL ACM SYMPOSIUM ON APPLIED COMPUTING, SAC 2023; v. N/A, p. 8-pg., 2023-01-01. |
| Resumo | |
Information about interactions between objects can be used to solve many important problems. One of these important problems is drug-target interaction prediction, where different machine learning methods can be applied to solve the prediction task. Among them, Predictive Bi-Clustering Trees (PBCTs) stand out for being a global-based multi-label algorithm with the ability to predict all interactions simultaneously. PBCTs induce a decision tree based on the interaction matrix to produce partitions, where each leaf node corresponds to a partition of the initial matrix. To be used, it needs an interaction matrix built from a true bipartite graph containing the interactions referring to the objects. However, it has a significant disadvantage over unbalanced datasets or datasets with a high rate of unknown (unlabeled) data. In this work, we propose a semi-supervised approach to improve predictive bi-clustering trees, where the semi-supervised impurity function replaces the impurity reduction function used in tree splits. We applied our approach to predict drug-target interaction and obtained competitive results compared to the original state-of-the-art PBCT. (AU) | |
| Processo FAPESP: | 20/11611-4 - Identificação de SNPs e genes relacionados à produtividade de grão em arroz utilizando aprendizado de máquina |
| Beneficiário: | Ricardo Cerri |
| Modalidade de apoio: | Auxílio à Pesquisa - Programa eScience e Data Science - Regular |