Feature selection has become revenue to many research regions that manage machine learning and data mining since it allows the classifiers to be cost-efficient, time-saving, and more precise. In this chapter, the feature selection strategy is consolidating by utilizing the combined feature selection technique, specifically recursive feature elimination, chi-square, info-gain, and principal component analysis. Machine learning algorithms like logistic regression, random support vector machine, and decision trees are applied in three different datasets that are pre-processed with combined feature selection technique. Then these algorithms are ensembled using voting classifier. The improvement in accuracy of the classifiers is observed by the impact of the combined feature selection.
TopLiterature Survey
Xie and Wang (2011) introduced a new strategy called half breed highlight determination, it improves F-score value along with Sequential Forward Search (IFSFS). The research got the main F-score by separating courses of real numbers assessment and isolation among extra genuine number arrangement. The better F-score and Sequential Forward Search are merged as ideal segment during time spent component assurance, be that as it may, that improved F-score as channel methodology evaluation model, and, SFS as appraisal system covering procedure.
Aruna et al (2012) came out with a hybrid feature decision procedure explicitly IGSBFS (Information Gain and Sequential Backward Floating Search), which solidifies potential gains channels just as covers pick ideal part beginning with first rundown of abilities reliant upon a characteristic model of NB. Information Gain variable evaluator are used in the opposite assurance progressively or otherwise called as straight forward decision along with skimming forward decision (IGSBFS) (FS2). In IGSBFS, IG functions as channels to dispose of abundance features. The course of action precision of the offered, method put forward is 98.9% with 10 features.
Xie et al (2013) developed an alternate hybrid FS computation across F-score where critical attributes as per a disorder dataset. Incorporate subset age pertinent features, glancing through techniques, for instance, SBFS, ESFS, and SFFS and summarized F-score to survey meaning of every segment. These techniques merge the advantages of channels and covers to pick the absolute component subset from the principal rundown of abilities to bring together the consistent and profitable classifiers.
Maryam et al (2017) put forth a hybrid procedure incorporate assurance methodology, Chi GA (Chi-Square and Genetic Algorithm), uses great conditions with channel, covering procedures picking the ideal component determining out remarkable component. Chi-square used as channel method to wipe out abundance features, GA picks ideal part SVM used a divider and the preliminary yield proposes model-based multiclass SVM with Chi-Square.