Adaptive dimensional search-based orthogonal experimentation SSA (ADOX-SSA) for training RBF neural network and optimal feature selection

This article suggests a hybrid salp swarm algorithm (SSA), adaptive dimensional search-based orthogonal experimentation SSA (ADOX-SSA), to improve SSA’s optimization skills. ADOX-SSA includes an adaptive dimensional search (ADS) mechanism and a newly proposed operator named ADS-based quantization with orthogonal experimentation (ADOX) as add-on tools in SSA. Applying ADS while updating search agents’ positions attains a tuned exploration–exploitation balance, making the convergence faster. Alongside, the utilization of ADOX for each search agent intensifies exploitation because of an ADS-based quantization operation and, later on, an orthogonal experimentation process in ADOX that diversifies the population, enhancing exploration. FirstAXSA’s optimization skills are evaluated by solving CEC2014 and CEC2022 benchmark problems. As per the results computed, ADOX-SSA has more robust skills in exploitation and exploration than SSA, offering better solutions in 100% and 88% of all unimodal and multimodal instances solved. Its more outstanding exploration–exploitation balance and complexity-handling skills than SSA are further shown by the 96% and 50% success rates on overall hybrid and composite situations, respectively. Compared to some recent algorithms, with some percentage of tied and competitive results, the method acts better than others at 100%, 91%, 92%, and 44% of all unimodal, multimodal, hybrid, and composite cases, showcasing its sophisticated optimization abilities. Furthermore, three separate tests further validate the statistical worth of ADOX-SSA’s results, each giving it the top rank among all the compared algorithms for all problem instances. Finally, by being more successful in 67% of neural network training occasions for data classification, with 22% ties and 75% feature selection scenarios, with 12% ties, ADOX-SSA validates its progressive effect in resolving real-world challenges.