A Study of BOS Investment Strategy Based on BP Neural Network Prediction

Abstract

As a new type of trading method, quantitative trading has been developing rapidly in recent years in developed countries such as the United States. By building BP neural network models and BOS models, we analyze the rise and fall of gold and bitcoin prices, develop the best trading strategies, and guide investors in their investment decisions. The future trading day prices of gold and bitcoin were predicted using the BP neural network model; based on the known historical data, the distribution law of the corresponding rise and fall and the distribution law of the number of consecutive rises and falls were obtained using the Apriori algorithm, respectively, to develop the BOS (buy or sell) system. The resulting return curves were found to have an annualized return of roughly 5.43% for gold and 56.86% for bitcoin, with a relatively smooth fluctuation for gold and a relatively large fluctuation for bitcoin. Then, the impact of the change in transaction cost on the BOS model and total return is investigated by varying the transaction cost several times. The model is discussed in three aspects to provide the best strategy.