Dynamic Decision Making with Applications in Healthcare and Airline Revenue Management

The goal of my Ph.D. thesis is to develop data-driven sequential decision-making algorithms that are applicable in airline revenue management and healthcare. Since the liberalisation of the airline market in the 1970s, airline companies have aimed to design better revenue management systems. An airline revenue management system computes the optimal fare to charge for a given customer. Traditional pricing methodologies used stochastic revenue management systems to optimise their revenue. They assume a parametric forecasting model for the demand without accounting for competition, and capacity constraints. With the help of data from an aviation partner, we have access to bookings, searches, and competition pricing data.

We wish to use the available data to derive an effective initial policy to price tickets. Further, the existing data is used to create a simulator to improve the performance of the derived initial policy. We plan to benchmark the proposed algorithms against state-of-the-art methodologies from approaches that are currently employed in practice. Subsequently, we wish to run a pilot program to validate the performance of these methods in a real-world setting.
The proposed algorithms utilise the searches and competition data which would be a novel addition to the revenue management literature. Online deployment of the algorithms to drive profit would also be a novel addition to the literature. The project would fall under the operations research theme under EPSRC's strategies. We collaborate with SKY Airlines to collect data and deploy the algorithms in an online fashion.

For the development of AI tools in healthcare, doctors must trust the treatment prescribed by an algorithm. Traditional methods either use a black box method for prediction, assume a parametric model for transitions, or don't account for ambiguity. Here we aim to develop interpretable dynamic treatment regimes to sequentially decide the medicine to prescribe to a patient when there is ambiguity present. The goals of this project would be to find a notion of interpretability, develop algorithms that are interpretable, and check the performance of real-world offline data. We wish to also theoretically calculate the cost of interpretability and find conditions when this goes to zero, to enable the right prescription of medication in the healthcare setting. This would belong in the healthcare, and operations research theme under EPSRC guidelines. We are collaborating with Sourush Saghafian from Harvard Kennedy School for this project.

The primary CDT impact will be training 75 PhD graduates as the next generation of leaders in statistics and statistical machine learning. These graduates will lead in industry, government, health care, and academic research. They will bridge the gap between academia and industry, resulting in significant knowledge transfer to both established and start-up companies. Because this cohort will also learn to mentor other researchers, the CDT will ultimately address a UK-wide skills gap. The students will also be crucial in keeping the UK at the forefront of methodological research in statistics and machine learning.
After graduating, students will act as multipliers, educating others in advanced methodology throughout their career. There are a range of further impacts:
- The CDT has a large number of high calibre external partners in government, health care, industry and science. These partnerships will catalyse immediate knowledge transfer, bringing cutting edge methodology to a large number of areas. Knowledge transfer will also be achieved through internships/placements of our students with users of statistics and machine learning.
- Our Women in Mathematics and Statistics summer programme is aimed at students who could go on to apply for a PhD. This programme will inspire the next generation of statisticians and also provide excellent leadership training for the CDT students.
- The students will develop new methodology and theory in the domains of statistics and statistical machine learning. It will be relevant research, addressing the key questions behind real world problems. The research will be published in the best possible statistics journals and machine learning conferences and will be made available online. To maximize reproducibility and replicability, source code and replication files will be made available as open source software or, when relevant to an industrial collaboration, held as a patent or software copyright.