Research on Trading Strategy Based on Machine Learning

Abstract

In this paper, we choose the best trading strategy by building a suitable forecasting model. Specifically, we choose XGBoost, ELM, and long short-term memory network (LSTM) as our baseline models in constructing forecasting models. Furthermore, based on further predictions given by our model, we construct a set of ground rules. On this basis, two correction methods, disturbance correction and multi-step correction are proposed for the dilemma. Then we studied the investment period of this strategy and found that the optimal investment period is 1 day. Finally, a sensitivity analysis of the commission rates for bitcoin and gold is performed.