Not logged in.
Quick Search - Contribution
Contribution Details
| Type | Master's Thesis |
| Scope | Discipline-based scholarship |
| Title | Optimal Asset Allocation with Reinforcement Learning |
| Organization Unit | |
| Authors |
|
| Supervisors |
|
| Language |
|
| Institution | University of Zurich |
| Faculty | Faculty of Business, Economics and Informatics |
| Date | 2021 |
| Abstract Text | This work examines deep reinforcement learning for optimal asset allocation in the US stock market using a custom implementation of the proximal policy optimization algorithm (PPO). Two different policy distributions, namely the Gaussian and Dirichlet distribution are tested as well as two different network architectures, namely a feed-forward neural network and the latter combined with an additional LSTM module, with the aim to provide the agent with more information about the past. The portfolio rebalancing is done under consideration of constraints such as transaction costs, budget constraints, no short-selling constraints and no fractional shares trading. The state space is constructed including technical momentum indicators, past returns, volatility and the volatility index (VIX). The best-performing strategy is PPO using a Gaussian policy with a feed-forward neural network, while the Dirichlet policy is found to enforce equal weights under all circumstances. I |
| Export |
BibTeX
|