Modeling And Enhancing Crude Oil Price Forecasting Using Enhanced Set Algebra-Based Heuristic Algorithm-Based Extreme Learning Machine

Abstract

Abstract This study has two main aspects. Firstly, we combined the Nelder-Mead Simplex Algorithm with the Set Algebra-Based Heuristic Algorithm (SAHA) in order to improve SAHA's capacity to do local searches. This integration resulted in a hybrid learning approach known as ESAHA. After that, we use the Enhanced Simulated Annealing with Hybrid Algorithm (ESAHA) to improve six benchmark functions so that we can see how well ESAHA works. Furthermore, we utilize ESAHA hybrid learning to enhance the weights and biases of an Extreme Learning Machine (ELM), resulting in the creation of a hybrid model referred to as ESAHA-ELM. We utilize the ESAHA-ELM model to predict the final price of crude oil datasets. In addition, we employ the SAHA, BMO, PSO, and GA algorithms to train the ELM and generate four alternative models for the purpose of comparison in the forecasting job. In order to examine the predictive accuracy of each model, we utilize the MAPE and MSE error metrics. Additionally, we implement the Prediction of Change in Direction (POCID) statistical test to determine if there are any significant differences between the models. The experimental investigation shows that the ESAHA-ELM model has statistical relevance in accurately capturing the inherent volatility of financial time series. In addition, it surpasses other models such as SAHA-ELM, MBO-ELM, PSO-ELM, and GA-ELM.