Nature-Inspired Feature Selection Algorithms: A Study

In this digital era, the amount of data generated by various functions has increased dramatically with each row and column; this has a negative impact on analytics and will increase the liability of computer algorithms that are used for pattern recognition. Dimensionality reduction (DR) techniques may be used to address the issue of dimensionality. It will be addressed by using two methods: feature extraction (FE) and feature selection (FS). This article focuses on the study of feature selection algorithms, which includes static data. However, with the advent of Web-based applications and IoT, the data are generated with dynamic features and inflate at a rapid rate, thus it is prone to possess noisy data, which further limits the algorithm's efficiency. The scalability of the FS strategies is endangered as the size of the data collection increases. As a result, the existing DR methods do not address the issues with dynamic data. The utilization of FS methods not only reduces the load of the data, but it also avoids the issues associated with overfitting.