Improving interaction prediction exploiting background information
Konstantinos Pliakos, Celine Vens · Lirias · 2016
During the last years, a burst of interest has been witnessed in the prediction of interactions that occur in biomedical networks. Despite the research effort made so far, accuracy and efficiency are still open problems. Here, a new prediction scheme is proposed that is based on supervised learning using Random Forest (RF) extended by Kernel Principal Component Analysis (KPCA). The obtained experimental results reaffirmed the potential of the proposed approach.