- Reza Rejaie (Chair)
- Ramakrishnan Durairajan
- Jun Li
- Daniel Lowd
- Raghuveer Parthasarathy, Physics
The significant increase in the scale and complexity of networked systems, from online retail networks to computer networks, on one hand, and the progress in machine learning techniques that is supported by the rapid development of software and hardware components, on the other hand, creates a unique dynamic. There is a pull from the networked systems for automated and scalable methods to handle the challenges of management, scheduling, and monitoring of such complex systems while there is a push from the machine learning side to solve such problems. In computer networks, forecasting the future values of network data streams have a high impact on scheduling, resource management, and monitoring of the system's health status. Detection of anomalies can help to identify issues before the damage is widely spread in the network. Similarly, in online retail stores, the dynamic between the customers' dependency on online rating and sellers' motivation to modify the online rating in their favor, might not be constructive in all cases and therefore, demands scalable and automated approaches to identify misuse of the system that can affect trust and experience.
In this dissertation, we discuss the applications of machine learning techniques in two networked systems namely (i) online reviews that demonstrate the network of shoppers and sellers in an online retail store, and (ii) a campus-wide network that represents the scale and complexity of today's computer networks. In each system, we explain the challenges of using machine learning techniques, the pros and cons of using them, and approaches such as model explainability to facilitate their usage in a practical setting.