Comparative analysis on feature selection process with high dimensionality through intelligent IoT sensors for hyper sensing images

Abstract

Due to the high correlation between the spectral features and the noise in the spectral bands, which could have a big effect on classification performance, the classification job has usually been done at the same time as dimensionality reduction. It is known as Hughes effect or the “curse of dimensionality” in the literature when there are insufficient training examples compared to the number of spectral characteristics, which has a detrimental influence on classification performance. In this study, we compare several feature selection methods using a high-dimensional representation model through intelligent IoT sensors. Comparing it to traditional feature-selection methods in terms of classification accuracy, stability of the picked features, and computing time, it is examined using a variety of hyper spectral datasets. The findings demonstrate that the suggested method offers both high classification accuracy and reliable features with an acceptable computing time.