A Hybridized ELM-Elitism-Based Self-Adaptive Multi-Population Jaya Model for Currency Forecasting

Abstract

Forecasting currency market plays an important role basically in all aspects of international financial management. Based on the claimed poor performance and less efficiency of the popular forecasting methods, an Extreme Learning Machine (ELM)—elitism-based self-adaptive multi-population Jaya model—is designed with the possibilities of getting maximum prediction accuracy. The model has evaluated by using the exchange rate data of USD to INR and USD to EURO as in most of the countries USD is used as the base currency. The experimented model has explored its prediction ability over six-time horizon, which includes the 6 days between 1 day and 1 month in advance. For a standardized comparison, two broadly accepted prediction models such as standard ELM and Back-Propagation Neural Network (BPNN) has been recommended for this research work. Along with these two standard models, a Jaya optimized technique applying on both ELM and BPNN is also considered for experimentation. The simulated result depicted the outstanding performance of the proposed ELM-elitism-based self-adaptive multi-population Jaya model over ELM, BPNN, ELM-Jaya, and BPNN-Jaya.