In this thesis, we investigate different practical problems in portfolio management. We present solutions to these problems based on machine learning. News Events are classified as possible risk sources by quantifying their causal effects on Chinese stock markets based on variational inference. Furthermore, we define and investigate asset class exposures of retail funds based on the arbitrage pricing theorem, a volatility filter from signal processing and a machine learning algorithm for pattern recognition in time series data. Lastly, we give a machine learning approach on optimal portfolio allocation/management. The asset allocation process is formulated as a Markov decision process and optimal portfolios with respect to their expected return are derived based on Reinforcement learning.
«
In this thesis, we investigate different practical problems in portfolio management. We present solutions to these problems based on machine learning. News Events are classified as possible risk sources by quantifying their causal effects on Chinese stock markets based on variational inference. Furthermore, we define and investigate asset class exposures of retail funds based on the arbitrage pricing theorem, a volatility filter from signal processing and a machine learning algorithm for patte...
»