SLSTM-Based Covariance Prediction and Its Application in Hierarchical Risk Parity Asset Allocation

In this paper, we integrate the selective state space model (SSSM) with the long short term memory (LSTM) neural network to construct a selective long short term memory (SLSTM) neural network model. We propose a novel dynamic covariance matrix prediction method and apply it to improve the hierarchical risk parity (HRP) asset allocation strategy.