Electronic Thesis and Dissertation Repository


Doctor of Philosophy


Electrical and Computer Engineering


Dr. Abdallah Shami

2nd Supervisor

Dr. Hanan Lutfiyya



The continued penetration of technology in our daily lives has led to the emergence of the concept of Internet-of-Things (IoT) systems and networks. An increasing number of enterprises and businesses are adopting IoT-based initiatives expecting that it will result in higher return on investment (ROI) [1]. However, adopting such technologies poses many challenges. One challenge is improving the performance and efficiency of such systems by properly allocating the available and scarce resources [2, 3]. A second challenge is making use of the massive amount of data generated to help make smarter and more informed decisions [4]. A third challenge is protecting such devices and systems given the surge in security breaches and attacks in recent times [5].

To that end, this thesis proposes the use of various optimization modeling and machine learning techniques in three different systems; namely wireless communication systems, learning management systems (LMSs), and computer network systems. In par- ticular, the first part of the thesis posits optimization modeling techniques to improve the aggregate throughput and power efficiency of a wireless communication network. On the other hand, the second part of the thesis proposes the use of unsupervised machine learning clustering techniques to be integrated into LMSs to identify unengaged students based on their engagement with material in an e-learning environment. Lastly, the third part of the thesis suggests the use of exploratory data analytics, unsupervised machine learning clustering, and supervised machine learning classification techniques to identify malicious/suspicious domain names in a computer network setting.

The main contributions of this thesis can be divided into three broad parts. The first is developing optimal and heuristic scheduling algorithms that improve the performance of wireless systems in terms of throughput and power by combining wireless resource virtualization with device-to-device and machine-to-machine communications. The second is using unsupervised machine learning clustering and association algorithms to determine an appropriate engagement level model for blended e-learning environments and study the relationship between engagement and academic performance in such environments. The third is developing a supervised ensemble learning classifier to detect malicious/suspicious domain names that achieves high accuracy and precision.

Available for download on Tuesday, September 01, 2020