Feature Selection in High Dimensional Data: A Review

Abstract

By choosing an ideal subset of the total features, feature selection in machine learning is essential to reducing the quantity of the data and increasing classifier performance. Nowadays, the size of data is increasing exponentially in fields like text classification, microarray data, bioinformatics, gene expression, information retrieval, etc. In high dimensional or big data, the learning model’s predictions are not accurate because of noisy or irrelevant features, so there is a challenge to reduce the data dimensionality. This paper introduces the concepts of feature relevance, relevant feature selection, and evaluation criteria. An overview and comparison of existing feature selection methods for various application domains are also done.