Machine Learning in Automated Asset Management Processes 4.1

Abstract

The traditional (human driven) process of Asset Management has become automatized by algorithmic decision trading with so called Robo Advisors (RAs). With an increasing amount of publicly available financial data, the foundation for applying machine learning (ML) algorithms has been paved. We examine the question in which process steps of automated investment advice ML algorithms could be applied and investigate which implementations have already been placed on the market. As the following study shows, (surprisingly) ML is globally still under its development phase in Robo Advisory. German and Swiss FinTech companies thereby contribute about a third to the ML solutions in our sample. The most promising technique is the usage of Text Mining for sentiment analyses, which can be used for monitoring and rebalancing purposes or future performance forecasting. Furthermore, Text Mining algorithms can be helpful for reducing information asymmetries. Embedded into early warning systems, the derived sentiment scores can be used for hedging against future price losses. This approach would be inevitably linked to an increased access of highly sensible data. Furthermore, we try to provide an explanation for the lack of acceptance of the application of ML in RA distributions. Possible reasons for this can be found in the current MiFID II regulations, which are not specified for ML. Based on these insights, we formulate first recommendations for both the provider of RA solutions as well as for the regulator.