A Hybrid-SFLA-Based Hyper-Parameter Tuning Algorithm in SVR for Exchange Rate Prediction

This paper proposes a hybrid machine learning-based approach to forecast the exchange rate between the Indian Rupee and the US dollar. The prediction of the exchange rate is carried out by support vector regression (SVR) and a hybrid-shuffled frog leaping algorithm (HSFLA). SVR predicts the exchange rate, whereas the hybrid-SFLA is used to tune the hyper-parameters. Hybrid-SFLA is an improvement over SFLA where random movement of particles is carried out using levy flight distribution. The proposed work has been compared with state-of-the-art hybrid prediction models that have applied different meta-heuristic algorithms. To measure predictive efficiency, root mean square error, mean absolute percentage error, Theil’s U, and average relative variance have been used, and the proposed HSFLA outperformed the other methods.