Hybrid feature selection model for accurate wind speed forecasting from numerical weather prediction dataset

Abstract

The accurate weather prediction in a localized area is obtained as a challenging role and the exact determination of Wind Speed Forecasting (WSF) is suitable for better power generation and production process. However, the Numerical Weather Prediction (NWP) is a significant model to gather information on surface flows and determine the raw data. But it contains issues such as instability, noise, and irregularity that provide a complex situation in the classification process. So Speedy Correlation-based Filtering-Enhanced Adaptive Sparrow Search Wrapper Algorithm (SCEASS-WA) is proposed to assess the problems of redundancy and correlation. Moreover, the Hybrid Feature Selection Method (HFSM) model is employed in the proposed method to control the noise from the extracted features. In SCEASS, the filter model Speedy Correlation Based Filtering (SCBF) based on Symmetrical Uncertainty (SU) is mainly presented to assess the noise. Also, the Wrapper Method (WM) is determined to improve the adaptation of the Sparrow Search Algorithm (SSA) to update the position for better outcomes. Furthermore, the filter methods are varied frequently by obtaining Damping Harmonic Oscillation Theory (DHOT) that discovered the superior feature subset. Also, the high dimensional datasets are employed that are compared with existing methods to show the superiority of the model. The experimentation results revealed that the proposed method attained a better prediction of weather and enhanced the accuracy forecast.