A hybrid filter and wrapper feature selection approach for detecting contamination in drinking water management system

Feature selection is an important task in predictive models which helps to identify the irrelevant features in the high - dimensional dataset. in this case of water contamination detection dataset,the standard wrapper algorithm alone cannot be applied because of the complexity. to overcome this computational complexity problem and making it lighter,filter-wrapper based algorithm has been proposed. in this work,reducing the feature space is a significant component of water contamination. the main findings are as follows: (1) the main goal is speeding up the feature selection process,so the proposed filter -based feature pre-selection is applied and guarantees that useful data are improbable to be detached in the initial stage which discussed briefly in this paper. (2) the resulting features are again filtered by using the genetic algorithm coded with support vector machine method,where it facilitates to nutshell the subset of features with high accuracy and decreases the expense. experimental results show that the proposed methods trim down redundant features effectively and achieved better classification accuracy. © school of engineering,taylor’s university.