Markus Löning

Research interests

Markus’ research is on supervised learning with time-series/panel data, i.e. observations on multiple independent individuals (e.g. customers, patients or machines) collected repeatedly over time. In particular, his goals are to

1. Create a practical, consistent and statistically solid workflow for modelling and evaluating supervised learning strategies with time-series/panel data,
2. Design and implement an open-source Python toolbox that allows to put the workflow into practice,
3. Develop probabilistic supervised learning methods based on point process models for panel data containing sequences of events with exact timestamps rather than regular time-series.

He is inspired by software development projects like scikit-learn, a popular machine learning toolbox in Python which not only makes supervised learning methods widely available but also easily understandable through an intuitive and consistent API design.

Moreover, his research involves a number of real-world datasets through ongoing industry collaborations, including the loyalty-card data from a UK high-street retailer (via the Consumer Data Research Centre), the fitness-training data from a supplier of cloud-connected gym equipment, and the biochemical data from a pharmaceutical company.