Quantitative stock portfolio optimization by multi-task learning risk and return

Yu Ma, Rui Mao, Qika Lin, Peng Wu, Erik Cambria

Abstract: Highlights•We propose a framework for stock ranking-based portfolio optimization.•We use a multi-task learning model to jointly learn stock return and volatility risk.•We fuse multiple effects of related stocks with a heterogeneous graph attention module.•Our model is evaluated by stock ranking prediction and financial backtesting.