본문 바로가기
  • Home

Machine-Learning-based Optimization of ETF Split Trading Strategies with Technical Indicators

  • Journal of The Korea Society of Computer and Information
  • Abbr : JKSCI
  • 2025, 30(7), pp.53~63
  • Publisher : The Korean Society Of Computer And Information
  • Research Area : Engineering > Computer Science
  • Received : June 11, 2025
  • Accepted : June 30, 2025
  • Published : July 31, 2025

Dongwon Lee 1

1한성대학교

Accredited

ABSTRACT

This study aims to develop a machine-learning-based asset allocation strategy that dynamically adjusts investment weights using technical indicators. By employing a parametric structure grounded in Bollinger Bands and the Commodity Channel Index (CCI), the strategy offers flexible responses to market changes. Several optimization algorithms, including Differential Evolution (DE), Powell, L-BFGS-B, TNC, COBYLA, and TRUST-CONSTR, were applied to tune the strategy parameters using historical ETF data. Each optimizer was evaluated through repeated simulations, and the models were assessed based on annual realized profits, return stability, and overfitting metrics. Among them, DE consistently delivered superior performance, demonstrating strong generalization and low levels of overfitting. These outcomes are attributed to DE’s ability to escape local optima and effectively explore the search space. These results indicate that combining global optimization with technical indicators enables robust, data-driven investment strategies that are adaptable to dynamic financial environments.

Citation status

* References for papers published after 2024 are currently being built.