Institutt for datavitenskap og analyse

Handelshøyskolen BI

Building on recent advances in Natural Language Processing and modeling of sequences, we study how a multimodal Transformer-based deep learning architecture can be used for measurement and structural narrative attribution in macroeconomics. The framework we propose combines (news) text and (macroeconomic) time series information using cross-attention mechanisms, easily incorporates differences in data frequencies and reporting delays, and can be used together with Reinforcement Learning to produce structurally coherent summaries of high-frequency news flows. Applied and tested on both simulated and real-world data out-of-sample, the results we obtain are encouraging.

Bernoulli 32(1) s. 24-48 Doi: https://doi.org/10.3150/25-bej1902

This article surveys key aspects of ambit stochastics and remembers Ole E. Barndorff-Nielsen’s important contributions to the foundation and advancement of this new research field over the last two decades. It also highlights some of the emerging trends in ambit stochastics.

Journal of Commodity Markets 41 Doi: https://doi.org/10.1016/j.jcomm.2026.100541

This study examines short-to mid-term point forecasting of daily electricity prices, with particular emphasis on the role of renewable energy sources. We use data from the market zone corresponding to the Northern region of Italy, applying both time series and machine learning methodologies. The forecasts are evaluated for two individual years, 2019 and 2024. In 2019, traditional energy variables such as electricity load, natural gas prices, and imports, were the primary drivers of forecast accuracy. During this period, adding renewable energy production data offered negligible benefits, with solar and wind contributing only marginally. By contrast, in 2024, market volatility increased greatly due to geopolitical conflicts and increased renewable energy integration. Under these conditions, while solar and wind still added limited value, hydropower improved forecast accuracy substantially. The results suggest that the role of renewable energy sources in electricity price forecasting is growing. However, their predictive power is influenced by their market share and by their variability and predictability.

Stochastic Processes and their Applications 196 Doi: https://doi.org/10.1016/j.spa.2026.104922

There exist multiple regression applications in engineering, industry and medicine where the outcomes follow an adaptive experimental design in which the next measurement depends on the previous observations, so that the observations are not conditionally independent given the covariates. In this paper we present new asymptotic theory for such processes of dependent regression data. This includes stand-alone results for adaptive covariate processes, which, in particular, enable us to develop new likelihood theory for inference on the regression model parameters. Existing results asserting asymptotic normality of the maximum likelihood estimator require regularity conditions involving the second or third derivatives of the log-likelihood. Here we instead extend the theory of differentiability in quadratic mean (DQM) to the setting of adaptive designs, which requires strictly fewer regularity assumptions than the classical theory. In doing so, we discover a new DQM assumption, which we call summable differentiability in quadratic mean (S-DQM). As applications, we first verify asymptotic normality for two classical adaptive designs, namely the Bruceton 'up-and-down' design and the Robbins–Monro design. Next, we consider a more complicated problem, namely a Markovian version of the Langlie design.

Proceedings of the 2025 IEEE International Parallel and Distributed Processing Symposium Workshops (IPDPSW), 3–7 June 2025, Milan, Italy s. 241-250 Doi: https://doi.org/10.1109/IPDPSW66978.2025.00047

Large-scale classification of social media content is a crucial technique for finding, studying, and analyzing misinformation in online social networks. Based on a manually labeled dataset of COVID-19 related conspiracy tweets, we train an NLP classifier and test methods for performing inference at scale using both GPUs as well as the Graphcore IPU AI accelerator.We apply our methods on a large dataset of about 2.5 billion tweets, demonstrating that using our methods, large scale inference is possible using affordable research infrastructures. Furthermore, we find that the IPU, due to its tile-centric design, is especially suited for such inference tasks.As a result, we obtain the AICO dataset of around 18 million tweets related to COVID-19 conspiracy theories that were posted between January 2020 and November 2021, which we make available for other researchers interested in studying the topic further under https://huggingface.co/datasets/Jlangguth/AICO.