Tuning the Model Factory
Hi! My name is Yngvild Hamre and I am a first-year industrial PhD student in statistics at the Department of Mathematical Sciences at the Norwegian University of Science and Technology (NTNU) in Trondheim. My project is a collaboration with DNB, the largest bank in Norway, where I got employed after finishing my master’s thesis at NTNU.
In DNB, I worked as a Data Scientist for two years before starting the PhD. The work was mainly related to a framework for making prediction models. The prediction models are an important part of the bank’s strategy for personalisation. Using such models, we can find the customers who are most likely to perform different actions, for example buying a product or churning. This enables DNB to be more relevant in the customer communication with advice and offers tailored to the individual, making a better customer experience and increased sales.
The possibilities in the field of modelling have increased at a tremendous rate the recent years, presenting a wide range of options when it comes to both platforms and algorithms. Thus, a Data Scientist in DNB needs to have broad overview of different technologies, as well as the ability to dig into complex algorithms. As the bank decided to support me in pursuing an industrial PhD, I am going to focus on the latter the next couple of years. The goal of the project is to test new methods for optimizing the hyperparameters of the machine learning algorithms used in the modelling framework at DNB. This will be done using design of experiments, a field in which my supervisor, John Tyssedal, has been working for many years. We eventually aim to compare Bayesian optimisation and response surface methodology with respect to hyperparameter tuning. If we succeed in finding useful methods, they will be implemented in the modelling framework at DNB and may thereby hopefully improve the performance of future models.
So why has DNB chosen to support this project? After all, one cannot guarantee results beforehand. Being a major player within technology in Norway, DNB has chosen to support several PhD projects at NTNU. It is important for DNB to have close collaboration with the academic community to keep up with the technological development and maintain a strong professional environment within the bank. Supporting research can also be considered as a way of taking social responsibility, which is another focus for DNB. Either way, I am very grateful for the opportunity to once again delve into the wonderful world of math and statistics, with the added motivation of trying to find something of interest for DNB.
Do not hesitate to contact me if you think the projects sounds interesting – I would be very happy to discuss different aspects of modelling and personalisation!