Large Dimensional Data Reduction by Various Feature Selection Techniques: A Short Review

In a data dependent era, dimensions contain a huge number of variables both in rows and columns forming more complex data matrices and these dimensional expansions generate a large dimensional data (LDD). The dimensionality problem of LDD is a massive challenge for analytics purpose and it somehow burdens the machine learning approaches. Due to the faster growing rate in innovative Internet of Things and web-based technologies, static data becomes noisy and nonstochastic that results in data loss and instability. Therefore, the demand for complex data dimension reduction technique (DDR) is growing immensely to improve data prediction, analysis and visualization. Several computational techniques have implemented for DDR which is further segregated into two categories such as feature extraction techniques (FET) and feature selection techniques (FST). But, most of the existing FET methods focus on transforming the higher dimensional data into a lower dimensional space and unable to tackle with the dimensionality reduction problem. Hence, this paper focuses on various dynamic FST that not only reduces the dimensionality load but also catalyze the data analysis process. General Terms Machine Learning, Feature Extraction, Feature Selection, Curse of Dimensionality, and Data dimension reduction.