Towards IoT Context Data Analysis and Classification

Abstract

The capability of generating and acquiring data has never been as pronounced as today since the advent of Information Technology in the early 19th century. Big Data is a growing trend where data processing has grown so much that it is beyond the ability of commonly used software tools to capture, manage, and analyse this data. The most fundamental challenge for the Big Data applications is to explore the large volumes of data and extract useful knowledge. Another recent development furthering the accumulation of data is the Internet of Things - wherein the amount of “things” housing different sensors has multiplied. The Internet of Things (IoT) is built on sensors and actuators, which send out large quantities of data about their state and the environment. Data processing techniques can be utilized to extract relevant insights from this treasure of data and information that is generated in IoT environments. This research investigates how sensor data from user devices can be used to determine context information about a user. As far as the classification of daily human activities (such as walking, eating, driving and sleeping) is concerned, several studies have proposed traditional computer vision methods. Unlike traditional computer vision methods, which pose mobility limitations, computational complexity, and privacy issues, our approach leverages mobile sensor data for real-time activity recognition. Three classification algorithms which are the K-Nearest Neighbours (KNN), Decision Trees, and Support Vector Machines (SVM) were evaluated based on their accuracy in classifying human activities. Findings revealed that the KNN was superior to other algorithms in providing a promising approach for context-aware systems in IoT environments. This study highlights the potential of Machine Learning and sensor data in enhancing user-centric applications through improved context-awareness.