Bagged Randomized Conceptual Machine Learning Method

Abstract

Formal concept analysis (FCA) is a scientific approach aiming to investigate, analyze and represent the conceptual knowledge deduced from the data in conceptual structures (lattice). Recently many researchers are counting on the potentials of FCA to resolve or contribute addressing machine learning problems. However, some of these heuristics are still far from achieving this goal. In another context, ensemble-learning methods are deemed effective in addressing the classification problem, in addition, introducing randomness to ensemble learning found effective in certain scenarios. We exploit the potentials of FCA and the notion of randomness in ensemble learning, and propose a new machine learning method based on random conceptual decomposition. We also propose a novel approach for rule optimization. We develop an effective learning algorithm that is capable of handling some of learning problem aspects, with results that are comparable to other ensemble learning algorithms.