Dimensionality reduction of hypergraph information system

Graph theory, as an important approach in data mining, can be applied to dimensionality reduction. As illustrated here, this paper proposes a new graph-theory method that reduces data dimensionality in a more effective and efficient manner than traditional methods. The proposed method, namely related family, is based on a hypergraph information system. The method not only compute all reducts of dimension set, but also adopts a heuristic algorithm to get one dimensionality reduction. The proposed heuristic algorithm can achieve more noisy-tolerable results in a low time complexity.