Enhanced extreme learning machine via competitive learning SSA (CL-SSA) for load capacity factor prediction.

Area of Science:

• Computational Intelligence
• Machine Learning
• Environmental Science

Background:

• Extreme Learning Machine (ELM) offers fast training but suffers from sensitivity to initialization and suboptimal weight optimization, impacting accuracy.
• Existing optimization methods for ELM may not fully address these limitations, necessitating improved algorithms for enhanced performance.

Purpose of the Study:

• To develop an enhanced Competitive Learning Salp Swarm Algorithm (CLSSA) to optimize Extreme Learning Machine (ELM) weights and biases.
• To improve the precision, convergence speed, and prediction accuracy of ELM for complex problems like load capacity factor prediction.

Main Methods:

• Integration of Salp Swarm Algorithm (SSA) with Competitive Swarm Optimization (CSO) to create the CLSSA optimizer.
• Evaluation of CLSSA using CEC 2015 benchmark functions against other optimization methods.
• Assessment of the CLSSA-enhanced ELM (ELM-CLSSA) for load capacity factor prediction.

Main Results:

• CLSSA demonstrated superior optimization capabilities, outperforming other methods in 86% of CEC 2015 benchmark functions.
• The ELM-CLSSA framework achieved a 97% accuracy rate in predicting load capacity factor, significantly outperforming traditional ELM and other approaches.
• Feature analysis identified key predictors for load capacity factor, including coal energy, economic growth, technological innovation, and biomass.

Conclusions:

• The proposed CLSSA is an effective optimizer, enhancing ELM's performance by improving weight and bias optimization.
• ELM-CLSSA provides a highly accurate and reliable framework for complex predictions, crucial for environmental sustainability initiatives.
• The framework offers valuable insights for policymakers and scientists in promoting ecological conservation and addressing climate change.