Corrected Sine Cosine Algorithm for Continuous Optimization

Developer: Hsing-Chih Tsai Contact Info: [e-mail] Please find SCA_main.m and run it. The SCA variant is attached.

This study examines search biases in the sine cosine algorithm (SCA), a popular optimization method, and proposes new variants to enhance performance. Two algorithm corrections are suggested: removing the absolute value notation and the parameter r3 from the position update equation. The original SCA's bias favors functions with optima at the coordinate origin but performs poorly in shifted domains. A new variant, TSCA3, is proposed to integrate individual-wise updating and evolutionary processes. Numerical results on the CEC2005 and CEC2019 benchmark suites indicate that TSCA3 achieves lower mean function errors than the original SCA and attains near-complete dominance in pairwise statistical tests for both low- and high-dimensional cases. Compared to classical metaheuristics, TSCA3 exhibits competitive performance on high-dimensional and shifted problems. Analyses of population diversity and exploration–exploitation balance confirm that TSCA3 transitions from broad initial exploration to concentrated late-stage exploitation, enabling faster and more reliable convergence on continuous optimization tasks. TSCA3 offers fundamental improvements and integration potential, serving as a robust alternative to the original SCA. This research also highlights potential limitations in benchmark functions with global optima near the origin and emphasizes the need for diverse optimal points in continuous optimization assessments.