Krause, Miguel2023-01-182023-01-182022-09-19http://hdl.handle.net/2003/4120210.17877/DE290R-23047This dissertation consists of four independently written essays dealing with inference and prediction of financial data sets. The first part of this dissertation focuses on the inference element and contains two chapters that explore the question of how financial markets priced companies’ stocks during the market collapse caused by the COVID-19 pandemic in the beginning of 2020. As the COVID-19 pandemic and the subsequent economic lockdown represented one of the most impacting exogenous shocks to financial markets in recent history, it led to a huge increase in uncertainty about a firm’s future cash flows. This environment thus allowed us to examine the drivers and characteristics that may make firms more resilient to crises and help to reduce investor uncertainty. The last two essays of this dissertation move away from the inference element and deal with the prediction of financial time series data using unsupervised machine learning methods. In the finance literature so far, machine learning models are mainly used for discriminative tasks, such as point forecasts or classifications. However, in this dissertation, we show how the finance literature can be extended by using generative probabilistic models, which aim to learn the underlying distribution of the data and are able to generate realistic artificial samples. Since time series in the real world are highly stochastic, probabilistic sampling has the advantage of providing a complete distribution of possible scenarios instead of a single prediction.enGenerative modelsDeep learningCOVID-19InvestorsEfficiencyInvestor relations330doctoral thesisDeep learningCOVID-19InvestitionLeistungsfähigkeit