My research is primarily focused on the area of software engineering, and in particular the development and evaluation of techniques to support the construction and evolution of more reliable and robust software systems. A common theme in much of this work is the application of machine learning to software engineering problems; for example, to automatically generate program test data, predict software project costs, perform intrusion detection, identify the root location of faults within systems, and automatically detect software system failures. The latter of these in particular makes extensive use of a range of both semi-supervised and unsupervised (clustering) machine learning algorithms to detect anomalous entries in large very high-dimensional and complex data sets. More recently I have also been turning my attention to the converse problem of testing AI systems.
My expertise and interests in machine learning extend outside the software engineering domain and I have employed clustering and classification algorithms in a variety of other contexts such as the automatic identification of potential road accident blackspots from crowdsourced smartphone sensor data, and the detection of objects within images.
I also have extensive experience of using machine learning in a variety of industrial projects such as forecasting customer buyer behaviour, predicting building energy performance, and modelling interventions to combat sedentary behaviour.
Over my career I have taught a lare range of classes, from 1st year undergraduate to postgraduate, mainly on areas related to programming, software engineering, software design, data analytics and machine learning.
My main current teaching responsibilities are:
I am also responsible for overseeing the MEng final year group project:
- Software engineering
- Machine Learning
- ESEM 2011: 5th International Symposium on Empirical Software Engineering and Measurement
- Member of programme committee
- Testing: Academic and Industrial Conference - Practice And Research Techniques (TAICPART)
- Information and Software Technology (Journal)
- Editorial board member
- Software Testing, Verification and Reliability (Journal)
- Editorial board member
More professional activities
- Providing confidence to encourage active travel through the application of AI
- Dunlop, Mark (Principal Investigator) Roper, Marc (Co-investigator)
- 01-Jan-2020 - 28-Jan-2021
- KTP - Maru Syngro
- Barlow, Euan (Principal Investigator) Revie, Matthew (Co-investigator) Roper, Marc (Co-investigator)
- 15-Jan-2018 - 14-Jan-2020
- Optimising Industrial Service Workforces using Mobile App Data
- Roper, Marc (Principal Investigator) Minisci, Edmondo (Co-investigator) Riccardi, Annalisa (Co-investigator)
- 01-Jan-2018 - 01-Jan-2019
- Development of a cross platform, personalised digital intervention to reduce sedentary behaviour and improve physical and mental well-being at work
- Terzis, Sotirios (Principal Investigator) Kirk, Alison (Principal Investigator) Roper, Marc (Co-investigator) Wallace, William (Co-investigator) Gibson, Ann-Marie (Co-investigator) Cogan, Nicola (Co-investigator) Janssen, Xanne (Co-investigator)
- 01-Jan-2018 - 31-Jan-2018
- Predicting user responses to wellness prompts
- Terzis, Sotirios (Principal Investigator) Roper, Marc (Co-investigator) Lennon, Marilyn (Co-investigator) Wallace, William (Co-investigator)
- 18-Jan-2017 - 21-Jan-2017
- Quantitative non-destructive nanoscale characterisation of advanced materials
- Hourahine, Ben (Principal Investigator) Edwards, Paul (Co-investigator) Roper, Marc (Co-investigator) Trager-Cowan, Carol (Co-investigator) Gunasekar, Naresh (Research Co-investigator)
- "To satisfy the performance requirements for near term developments in electronic and optoelectronic devices will require pioneering materials growth, device fabrication and advances in characterisation techniques. The imminent arrival of devices a few atoms thick that are based on lighter materials such as graphene or boron nitride and also advanced silicon and diamond nano-structures. These devices pose new challenges to the currently available techniques for producing and understanding the resulting devices and how they fail. Optimising the performance of such devices will require a detailed understanding of extended structural defects and their influence on the properties of technologically relevant materials. These defects include threading dislocations and grain boundaries, and are often electrically active and so are strongly detrimental to the efficiency and lifetimes of nano-scale devices (a single badly-behaved defect can cause catastrophic device failure). These defects are especially problematic for devices such as silicon solar cells, advanced ultraviolet light emitting diodes, and advanced silicon carbide and gallium nitride based high power devices (used for efficient switching of large electrical currents or for high power microwave telecoms). For graphene and similar modern 2D materials, grain boundaries have significant impact on their properties as they easily span the whole size of devices.
Resolving all of these problems requires new characterisation techniques for imaging of extended defects which are simultaneously rapid to use, are non-destructive and are structurally definitive on the nanoscale. Electron channelling contrast imaging (ECCI) is an effective structural characterisation tool which allows rapid non-destructive visualisation of extended crystal defects in the scanning electron microscope. However ECCI is usually applied as a qualitative method of investigating nano-scale materials, has limitations on the smallest size features that it can resolve, and suffers from difficulties in interpreting the resulting images. This limits this technique's ability to work out the nature of defects in these advanced materials.
We will make use of new developments in energy resolving electron detectors, new advances in the modelling of electron beams with solids and the knowledge and experience of our research team and partners, to obtain a 6 fold improvement in the spatial resolution of the ECCI technique. This new energy-filtered way of making ECCI measurements will radically improve the quality of the information that can be obtained with this technique. We will couple our new capabilities to accurately measure and interpret images of defects to other advanced characterisation techniques. This will enable ECCI to be adopted as the technique of choice for non-destructive quantitative structural characterisation of defects in a wide range of important materials and provide a new technique to analyse the role of extended defects in electronic device failure."
- 01-Jan-2017 - 30-Jan-2021
Computer and Information Sciences
View University of Strathclyde in a larger map