Design and research of online learning system based on improved differential evolution K-Means clustering algorithm

Abstract

This study focuses on designing an online learning system using an enhanced differential evolution K-Means clustering algorithm. With the growing popularity of online learning, the efficient distribution of personalized learning resources has become a key research area. By addressing the limitations of traditional methods and integrating advanced optimization techniques, the proposed system offers superior performance in terms of clustering accuracy, convergence speed, and adaptability. The detailed analysis of optimization algorithms’ categories and recent developments further positions this research within the broader context of educational technology advancements. The clustering accuracy of the improved algorithm reaches 37.13% when processing the Iris dataset. This relatively low accuracy is partly attributed to the Iris dataset’s inherent complexity, including class overlap and subtle feature differences between clusters. It also indicates potential for further optimization, such as refining the distance measurement method for overlapping classes. The experimental evaluation shows that on the Iris and Wine datasets from the UCI Machine Learning Repository, the clustering accuracy of the improved algorithm reaches 37.13% and 41.22%, respectively. Compared with the traditional K-Means algorithm, its convergence speed is increased by 4.5 times, where convergence speed is measured by both the number of iterations required to reach stable clustering results and the computational time.