Asset pricing, internet data, machine learning with natural language processing in the era of sustainable and decentralized finance with uncertainty

This project aims to deepen our understanding of financial markets through improvements in their modelling with machine learning methods. These methods will allow us to aggregate internet-derived information on sentiment and attention with other market anomalies and to analyze their non-linear interactions in order to better understand their role in asset pricing. We will use advanced language learning models for extraction of multi-dimensional sentiment-related features from financial news and social media texts and for analysis of their impact on market anomalies and market efficiency. We will use machine learning models to investigate cryptocurrency markets and to find new ways for modeling future cryptocurrency returns based on transaction data, order book data and internet derived sentiment and attention proxies. We expect that our research in the areas described above will also provide further insights into closely related areas of modeling of commodity prices, climate risk in finance, credit risk, option pricing and corporate finance issues as well.

A grant of around 245,000 euros from the Czech Science Foundation Grant (GACR).