Semantic-driven dimension reduction for wireless internet of things

Abstract

In recent years, mobile communication and artificial intelligence technologies have been widely used in the construction of wireless networks, bringing about a dramatic increase in data size. Existing wireless networks usually consist of a large number of nodes, with the potential risk of the curse of dimensionality. High dimensionality plays a negative role in learning effectiveness and efficiency, which should have been studied in depth but is neglected in existing wireless network research. In order to generate effective semantic-driven efficiency, this paper focuses on semantic-driven dimensionality reduction for wireless Internet of Things. Specifically, this paper introduces a series of feature selection techniques centered on Mahalanobis distance for dimensionality reduction, which helps to select discriminative features by measuring the effectiveness of semantic preferences and semantic-driven efficiency through Mahalanobis distance. Experiments on a set of wireless sensor data and various high-dimensional microarray data validate the superior performance of the proposed method.