Fully Associative Ensemble Learning for Hierarchical Multi-Label Classification

In Hierarchical Multi-label Classification (HMC), rich hierarchical information is used to improve classification performance. Global approaches learn a single model for the whole class hierarchy [3, 6]. Local approaches introduce hierarchical information to the local prediction results of all the local classifiers to obtain the global prediction results for all the nodes [2, 5]. In this paper, we propose a novel local HMC framework, Fully Associative Ensemble Learning (FAEL). Specifically, a multi-variable regression model is built to minimize the empirical loss between the global predictions of all the training samples and their corresponding true label observations. Let X and Y represent local prediction matrix and label observation matrix, respectively. We define W = {wi j} as a weight matrix, where wi j represents the weight of the ith label’s local prediction to the jth label’s global prediction. In the basic model, the objective function is: