Streamflow prediction in a river using twin support vector regression and extreme learning machine
Jagdish Mallick, Safalya Mohanty, Abinash Sahoo · 2024
Monthly streamflow forecasting can produce significant knowledge for hydrogeological applications which includes optimization of water resources allocation, sustainable design of urban and rural water management systems, water usage, water quality and pricing assessment, and irrigation and agriculture operations. Motivation to explore and develop proficient prediction models is a continuing effort for hydrogeological applications. This study explores the potential of twin support vector regression (TSVR) method, for monthly forecasting in Rushikulya River, Odisha, India. A comparative analysis is done for evaluating prediction performance of TSVR in comparison to extreme learning machine (ELM). Forecasting metrics such as correlation coefficient (R), root mean squared error (RMSE), and Nash-Sutcliffe efficiency (ENS) are applied for assessing effectiveness TSVR model outperformed ELM model across all statistical indices. In quantifiable terms, preeminence of TSVR over ELM model was demonstrated by ENS = 0.9783 and 0.9501, R = 0.9902 and 0.9757, and RMSE = 2.356 and 6.2213, respectively.