Elite Voting Crossover and Mutation of Genetic Algorithm for Feature Selection

Abstract

In record classification, not all available attributes might be both useful and relevant. Hence the need for feature selection arise. Over the years, the Machine Learning community has used a number of algorithms for feature selection. One of such algorithms is Genetic Algorithm (GA). Recently, these algorithms have a fall in performance due to growth in data dimensionality and sample size. This calls for efforts to enhance the performance of this algorithm to meet the current trends. Given that the performance of GA depends on the parameter values used and genetic operators applied, it is pertinent to improve performance of GA by improving the genetic operators. To this end, this work introduced new genetic operators (crossover and mutation) of the GA to be used in Feature Selection (FS).The efficiency of proposed algorithm was measured using Extreme Learning Machine (ELM) with Cleveland dataset from the University of California Irvine Machine Learning repository. The result shows a promising improvement over performance. The proposed algorithm is particularly important in situation of time constraints (online) and low computation power availability.