Market trading: LSTM-based forecasting and decision making

Abstract

Power In this paper, we address the decision problem of whether a trader should trade two assets, gold and bitcoin, daily under different circumstances, and use LSTM, neural network models, to obtain daily predicted price data for gold and bitcoin based on the knowledge of past market data. Accordingly, different planning strategies are applied to gold and bitcoin to obtain the total assets we can hold after five years. We provide evidence of the optimality of the strategy in several ways and specifically analyze the sensitivity of the strategy to transaction costs and the impact of transaction costs on the strategy and the results.