Student psychology based optimization algorithm integrating differential evolution and hierarchical learning for solving data clustering problems

How to accurately and efficiently perform data clustering in complex multidimensional data analysis and processing tasks is a challenging research problem. Traditional optimization algorithms often need help with the problems of quickly falling into local optimum and insufficient global search ability when dealing with high-dimensional, multi-peaked and complex structured data. In order to solve this challenge, a student psychology based optimization algorithm (GDLSPBO) that integrates differential evolution and hierarchical learning mechanisms was proposed, aiming to improve the accuracy, stability and global optimization ability of data clustering. GDLSPBO enhances the population diversity and prevents the algorithm from falling into local optimum by introducing the differential evolution mechanism. Simultaneously, the hierarchical learning strategy improves the algorithm’s search efficiency and local optimization ability. The experiments are validated on the CEC-BC-2017 benchmark functions. Several real datasets and the results show that GDLSPBO achieves an F-measure of 0.9595 and an Adjusted Rand coefficient of 0.8578 on the Cancer dataset, and the clustering accuracy on the Iris dataset reaches 93.33%, which is significantly better than that of other classical optimization algorithms. This indicates that GDLSPBO has a more substantial clustering effect and higher solution accuracy in solving complex data clustering problems. The experimental results verify that the global search ability and optimization accuracy of GDLSPBO on multidimensional complex data sets have been significantly improved, demonstrating its broad applicability and robustness in practical data clustering applications.