LSTMGA-QPSBG: An LSTM and Greedy Algorithm-based Quantitative Portfolio Strategy for Bitcoin and Gold

Abstract

Quantitative trading plays a pivotal role in financial markets. Over the past decade, quantitative trading has made remarkable improvements. Due to instability and nonlinearity in financial markets, it is still challenging to formulate high-return trading strategies to address the problem of long-term time series forecasting in financial markets. To tackle this issue, we propose an LSTM and Greedy Algorithm-based quantitative portfolio strategy in this work. First, an LSTM-based price prediction model is presented to forecast the closing price on the final trading day. Subsequently, a greedy algorithm is employed to identify the optimal daily trading strategy to pursue the overall optimal solution and achieve maximal profits. The experimental results show that the maximum value of the VaR percentage is about 9.1%, proving that the proposed strategy is feasible and effective.