A hybrid Q-learning sine-cosine-based strategy for addressing the combinatorial test suite minimization problem.

Area of Science:

• Artificial Intelligence
• Optimization Algorithms
• Software Engineering

Background:

• The sine-cosine algorithm (SCA) is a population-based meta-heuristic algorithm effective for optimization.
• However, SCA is prone to local minima/maxima due to fixed switching probabilities and bounded function magnitudes.
• This limits its performance in complex search spaces.

Purpose of the Study:

• To introduce a hybrid Q-learning sine-cosine algorithm (QLSCA) to overcome SCA's limitations.
• To enhance the search process by incorporating Q-learning for dynamic operation selection.
• To improve solution diversity and escape local optima using Lévy flight and crossover.

Main Methods:

• Developed the Q-learning sine-cosine algorithm (QLSCA) by integrating Q-learning with SCA.
• Replaced fixed switching probability with a Q-learning mechanism for adaptive operation selection.
• Incorporated Lévy flight motion and crossover operations to enhance exploration and exploitation.
• Applied QLSCA to the combinatorial test suite minimization problem.

Main Results:

• QLSCA demonstrated statistically superior performance in reducing test suite size compared to SCA, PSTG, APSO, and CS at the 95% confidence level.
• No significant performance difference was observed between QLSCA and DPSO at the 95% confidence level.
• QLSCA showed statistically significant outperformance over DPSO in specific configurations at the 90% confidence level.

Conclusions:

• The proposed QLSCA effectively addresses the local minima/maxima vulnerability of the original SCA.
• QLSCA offers a robust and adaptive approach for combinatorial test suite minimization.
• The hybrid strategy shows significant potential for improving optimization performance in various applications.