PhD in Novel Time Series Machine Learning Methodology for High-Dimensional Data
This PhD project at the University of Strathclyde's Department of Mathematics & Statistics focuses on developing novel time series machine learning (TSML) methodologies for high-dimensional data, with a particular emphasis on imputation of missing values and robust forecasting. The research addresses critical challenges in handling high-dimensional and discrete-valued time series, which are prevalent in finance, healthcare, and environmental science.
The project is structured around two main objectives: (i) designing advanced imputation techniques for missing data in high-dimensional time series, and (ii) creating machine learning architectures for accurate and robust forecasting. The imputation component leverages network structures and temporal dependencies, introducing frameworks such as Markov regime-switching networks, state-space models, and self-exciting spatio-temporal models. These approaches aim to handle both continuous and discrete data distributions, overcoming limitations of existing methods that are typically restricted to low-dimensional or continuous-valued cases.
For forecasting, the project will develop and extend deep learning models, including temporal convolutional networks and transformer-based architectures, tailored for high-dimensional and discrete-valued time series. The research also incorporates factor models and hybrid probabilistic-statistical approaches, with a focus on dynamic uncertainty quantification through Bayesian inference and quantile regression. These innovations are expected to enhance the accuracy, robustness, and interpretability of time series predictions.
Applications of the developed methodologies span financial forecasting and risk modeling, public health monitoring, and environmental trend analysis. The project will utilize real-world datasets from public sources, industry partners, and healthcare collaborators to validate the proposed models in practical scenarios such as anomaly detection and decision support.
Expected outcomes include publications in leading journals and conferences, open-source AI models for time series imputation and forecasting, and deployment-ready prototypes for selected applications. The project offers an excellent opportunity for candidates interested in cutting-edge research at the intersection of statistics, machine learning, and high-dimensional data analysis.
Applicants should have a strong background in mathematics, statistics, or computer science, with relevant experience in machine learning, statistical modeling, or data analysis. Proficiency in programming (e.g., Python or R) and familiarity with time series analysis are highly desirable. English language proficiency is required as per university guidelines. The application deadline is June 5, 2026.
For more information and to apply, visit the project page on FindAPhD.
