PhD projects in Forecasting
Forecasting time series data is an integral part of Operational Research, both as an established research area in its own right and as a prerequisite to many complex and uncertain real world problems.
OR models in simulation, logistics, transport and revenue management often assume stationary and homoscedastic environments, neglecting the often time dependent properties of the dataset.
Most real world time series exhibit non-stationary properties such as trends, multiple overlying seasonality, outliers, changes of variance, structural breaks in the form of level shifts, stochastic/local time trends or shifting seasonal patterns, and causal influences through external calendar effects (e.g. Christmas, Easter) and internal corporate decision making (e.g. Marketing policies of price and promotions). The complex interaction of these effects require valid and reliable forecasting in order to uphold the assumptions of the OR models.
We welcome PhD research proposals in all theoretical aspects of forecasting, deepening our theoretical understanding of the methods and algorithms, employing contemporary algorithms developed in other disciplines to forecasting problems, extending methods in the context of a particular OR decision problem or research cluster or linking forecasting uncertainty to the OR decision to overcome the divide and conquer paradigm of disparate forecasting and decision making.
Topical areas of interest include, but are not limited to:
- Extending nonlinear, nonparametric algorithms from computational intelligence such as Neural networks and Support Vector Machines (e.g employing novel optimisation algorithms for parameterisation)
- Exploring algorithms of time series clustering or classification to applications in novel applications (e.g. pattern or anomaly detection for Green Logistics, Healthcare, Transport etc.)
- Developing customised forecasting algorithms for the time series properties and challenges in Green Logistics, Healthcare, and Transport
- Exploring the interaction of forecasting errors to complex decision making, e.g. for inventory planning and human resource scheduling.
For further information contact .