Contextual Ontology-Based Feature Selection for Teachers

Abstract

The context of teacher is indescribable without considering the multiple overlapping contextual situations. Teacher Context Ontology (TCO) presents a unified representation of data of these contexts. This ontology provides a relatively high number of features to consider for each context. These features result in a computational overhead during data processing in context-aware recommender systems. Therefore, the most relevant features must be favored over others without losing any potential ones using a feature selection approach. The existing approaches provide struggling results with high number of contextual features. In this paper, a new contextual ontology-based feature selection approach is introduced. This approach finds similar contexts for each insertion of new teacher using the ontology representation. Also, it selects relevant features from multiple contexts of a teacher according to their corresponding importance using a variance-based selection approach. This approach is novel in terms of representation, selection, and deriving implicit relationships for features in the multiple contexts of a teacher.