A Framework Utilizing Genetic Algorithm Wrapped Neural Network Model in Discovering Effective Technical Indicators for the Cryptocurrency Market

Abstract

A main challenge of cryptocurrency trading is selecting technical indicators which fits the dynamic nature of the cryptocurrency market. This research proposes a framework that integrates a genetic algorithm with a neural network to effectively explore the efficacy of traditional technical indicators in cryptocurrency. It optimizes both the selection of technical indicators and neural network parameters through tailored genetic operations such as mutation and crossover, allowing for enhanced exploration of the solution space. Through rigorous testing on historical cryptocurrency market data in two distinct periods, the proposed model demonstrates superior predictive accuracy and improved trading performance compared to traditional methods, generating a 19.33% profit in the first period and 7.13% in the second period, outperforming the buy-and-hold benchmark. The results highlight the robustness of the model, which consistently delivered positive returns across varying market conditions, including both bullish and bearish phases.