Prominent Feature Selection for Sequential Input by Using High Dimensional Biomedical Data set

The main aim of feature selection algorithms is to select prominent (optimal) features that are not irrelevant and redundant. Reducing the number of features by keeping classification accuracy the same is one of the critical challenges in Machine Learning. High dimensional data contains thousand of features with the existence of the redundant and irrelevant features which negatively affect the generalization capability of the system. This paper designs the improved genetic-based feature selection (IGA) for Online Sequential—Extreme Learning Machine (IGA-OSELM) algorithm with additive or radial basis function (RBF). Experimental results are calculated for the Extreme Learning Machine (ELM), OSELM, IGA-ELM, and IGA-OSELM. With the result, it is inferred that IGA-OSELM maintains the classification accuracy by minimizing 58.50% features. The proposed algorithm is compared with the other popular existing sequential learning algorithms as the benchmark problem.