Clustering of Microlearning Units Based on Improved Flamingo Search Algorithm

The rapid growth of microlearning resources has created a need for effective clustering to help learners find personalized content. This paper proposes an improved flamingo search algorithm (FSA), optimized with dynamic reverse learning, for clustering microlearning units. To address the original algorithm’s issues of local optimization and low convergence accuracy, three strategies are introduced: a good point set for initial population generation to enhance diversity, Levy flight for better global search, and dynamic reverse learning to maintain diversity in later iterations. Experiments on 18 benchmark functions show the improved algorithm achieves higher search accuracy compared with the algorithm discussed in this paper. Further tests on clustering datasets demonstrate its effectiveness, producing lower error rates and higher accuracy, making it well-suited for microlearning unit clustering. In the Glass, Wine, and Seeds datasets, DFSA-KM outperformed the comparison algorithms across all evaluation metrics. Its accuracy improved by 15.8%, 4.5%, and 1.2% compared to the mean of the K-means method, respectively.