Important Features Selection Method for Temporal and Information Aware Clustering of Internet of Things Stream Data

Abstract

The Internet of Things (IoT) is essential because it creates new services that allow multiple disparate devices to be seamlessly connected over the Internet. The Internet of Things is growing every second in several industries, including smart cities, smart homes, smart factories and oil and gas.  The high dimensionality of data has made machine learning and data mining more challenging. This study presents a comprehensive overview of advanced dimensionality reduction techniques that aid in the selection of essential features for machine learning and IoT-based data analytics. The proposed method for clustering of IoT data uses Finite Differential method (FDM), Principal component analysis (PCA) and K-mode Clustering. These techniques are based on criterion measures, and datasets, and are inspired by soft computation technology. In Internet of Things applications, the resulting clustered features can help identify and reduce uncertainty, improve data clarity, and minimize data loss. The article discusses potential avenues for future research and provides readers with information regarding the suitability of various data reduction techniques by providing an overview of Dimensionality Reduction applications across various fields.