A Framework for Semantic Annotation and Mapping of Sensor Data Streams Based on Multiple Linear Regression

Abstract

In IoT, multitudes of sensors are streaming massive data which are hard to interpret meaningful information due to the presence of noise, outliers and missing value in sensor-observed data. In addition to this, heterogeneous sensors or devices in smart environment show great variations in formats, domains, and types, which stances challenges for machines to process and recognize. These challenges lead the interoperability issues in IoT. To overcome the above-mentioned issues, this work initially performs the preprocessing (i.e., removal of outlier, missing data completion) using the F-statistical tests and multiple linear regression models. Secondly, this research work proposes an Extended Sensor Markup Language for annotation of sensor-observed data and semantic mapping method to map the sensor data with standard Semantic Sensor Network (SSN) ontology for semantic interoperability.