Electronic Thesis and Dissertation Repository


Master of Engineering Science


Electrical and Computer Engineering


Dr. Miriam Capretz


Data in all of its form is becoming a central part of our existence, it is being captured in every facets of our everyday life: social media, pictures, smartphones, wearable devices, smart building etc. One of the main drivers of this Big Data Revolution is the Internet of Things, which enables inert objects to communicate through a multitude of sensors. The data amassed fuels a thirst for information, the extraction of such knowledge is rendered possible through Data Analytics Techniques.

However, when it comes to sensor data our large-scale ability to perform analytics is highly limited by the difficulties associated with collecting sensor data labels. Current crowdsourcing platforms historically used to gather labels are unable to process sensor data due to its low level nature. The solution proposed in this thesis enables the deployment of a crowdsourcing platform for sensor data. This research presents a novel solution to acquire sensor labels by leveraging the power of crowdsourcing using gamification. The work in this thesis describes not only a framework that facilitates the capture of sensor data label through a flexible gamification architecture but also a solution that outlines the mechanics required to integrate gamification in a variety of contexts. Additionally, the framework is designed in a flexible manner to support any type of sensor data given that human can readily interact with them. Additionally, the work presented describes and supports both real time and historical data analytics through the captured data and associated labels.

This work was successfully evaluated in the context of a case study where the gamification implementation was tested for a number of electrical sensors. Real time and historical data analytics were successfully performed with the use of the framework. The robustness of the solution was evaluated though the injection of invalid data and the result showed that the framework is effectively capable of reducing the level of noise in the data labels.