2018/06/20 by Mallenahalli, Naresh, T. Hitendra Sarma, Sarma, T. Hitendra
Computer Science · #Evolutionary Algorithms and Applications #FOS: Computer and information sciences #Machine Learning and Data Classification #Metaheuristic Optimization Algorithms Research #Neural and Evolutionary Computing (cs.NE)
paper · pdf · doi:10.48550/arxiv.1806.10551
openalex publication_date 2018/06/20 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28
Feature selection is the process of identifying statistically most relevant\nfeatures to improve the predictive capabilities of the classifiers. To find the\nbest features subsets, the population based approaches like Particle Swarm\nOptimization(PSO) and genetic algorithms are being widely employed. However, it\nis a general observation that not having right set of particles in the swarm\nmay result in sub-optimal solutions, affecting the accuracies of classifiers.\nTo address this issue, we propose a novel tunable swarm size approach to\nreconfigure the particles in a standard PSO, based on the data sets, in real\ntime. The proposed algorithm is named as Tunable Particle Swarm Size\nOptimization Algorithm (TPSO). It is a wrapper based approach wherein an\nAlternating Decision Tree (ADT) classifier is used for identifying influential\nfeature subset, which is further evaluated by a new objective function which\nintegrates the Classification Accuracy (CA) with a modified F-Score, to ensure\nbetter classification accuracy over varying population sizes. Experimental\nstudies on bench mark data sets and Wilcoxon statistical test have proved the\nfact that the proposed algorithm (TPSO) is efficient in identifying optimal\nfeature subsets that improve classification accuracies of base classifiers in\ncomparison to its standalone form.\n