NMF-DCA: An efficient dendritic cell algorithm based on non-negative matrix factorization

Abstract

We discuss in this paper the challenge of enhancing the dendritic cell algorithm preprocessing phase. In short, to minimize data dimensionality we propose a new dendritic cell algorithm based on Non-negative Matrix Factorization. The aim of this method is to extract latent features from lower rank data transformation. The proposed method was divided in two steps. The first step is a factorization of the original data. Secondly, the new reduced space should be assigned to its respective signal category. Experimental findings show that the preprocessing step of the dendritic cell algorithm is significantly improved with respect to its execution and a higher accuracy rate. Our algorithm is also compared to other classification algorithms particularly MLP, SVM, and KNN. The comparison shows that the actual rate of current algorithms is outperformed by the proposed algorithm.