machine learning portfolio replication LASSO LARS trading strategies hidden Markov model hierarchical modelling
Abstract:
In this article, we address the problem of portfolio replication raised by researchers in finance. We develop a new machine learning algorithm L1 Regularized Rolling Regression and also make inference on trading strategies based on the daily return and cumulative return space. We also incorporate hierarchical modeling and hidden Markov model to refine our results. Synthetic portfolio are constructed to simulate the performance of this parametric learning method. Real data analysis are shown to prove its capability of handling complex and low frequency data. This new method could also be generalized to unwind other problems of similar kind.
