Reverse adaptive krill herd locally weighted support vector regression for forecasting and trading exchange traded funds

Georgios Sermpinis, Charalampos Stasinakis, Arman Hassanniakalager

Research output: Contribution to journalArticle

6 Citations (Scopus)
40 Downloads (Pure)

Abstract

This study introduces a Reverse Adaptive Krill Herd-Locally Weighted Support Vector Regression (RKH-LSVR) model. The Reverse Adaptive Krill Herd (RKH) algorithm is a novel metaheuristic optimization technique inspired by the behavior of krill herds. In RKH-LSVR, the RKH optimizes the locally weighted Support Vector Regression (LSVR) parameters by balancing the search between local and global optima. The proposed model is applied to the task of forecasting and trading six ETF stocks on a daily basis over the period 2010–2015. The RKH-LSVR's efficiency is benchmarked against a set of traditional SVR structures and simple linear and non-linear models. The trading application is designed in order to validate the robustness of the algorithm under study and to provide empirical evidence in favor of or against the Adaptive Market Hypothesis (AMH). In terms of the results, the RKH-LSVR outperforms its counterparts in terms of statistical accuracy and trading efficiency, while the time varying trading performance of the models under study validates the AMH theory.
Original languageEnglish
Pages (from-to)540-558
Number of pages19
JournalEuropean Journal of Operational Research
Volume263
Issue number2
Early online date10 Jun 2017
DOIs
Publication statusPublished - 1 Dec 2017

Keywords

  • Forecasting
  • Support Vector Regression
  • Krill herd
  • Adaptive market hypothesis
  • Optimization

Cite this