Parallel Ants Colony Optimization Algorithm for Dimensionality Reduction of Scientific Documents

Abstract

Dimensionality reduction is crucial in Machine Learning, to obtain main characteristics. The method of selecting characteristics that we will use is a multivariate filter, where we will jointly evaluate the relevance between the characteristics; using unsupervised learning. For which we will use information from Institute of Education Sciences, and application of TF-IDF to obtain the weights of each word in each document. To perform the dimensionality reduction, the PUFSACO (Parallelization Unsupervised future selection based on Ant Colony Optimization) algorithm will be applied, due to the large amount of information that will be processed. The output of PUFSACO will be the input of the classification algorithm. The present work proposes to parallelize the UFSACO algorithm (Unsupervised future selection based on Ant Colony Optimization). Being the basis of PUFSACO, comparing the computational time to validate the improvement of the proposed algorithm, the results show that applying parallelization improves 117% than the original algorithm.