Research on the role of edge samples in the analysis of large data sets

In the process of dataset construction, although large sample datasets provide richer feature information, the use of small samples as a substitute for large samples has become a feasible solution due to computational resource constraints or the need to improve model efficiency. This paper proposes an algorithm optimization strategy based on edge sample selection. Specifically, the structural information between samples is first learned through a graph neural network, while the feature learning of the samples is represented by an autoencoder. Then, an adaptive mechanism is employed to divide the samples into cluster center samples and edge samples, which are clustered separately using the k-means method and fuzzy C-means clustering (FCM). A performance comparison of the graph neural network-based clustering algorithm is conducted on three datasets, and the results show that the use of edge samples can partially replace the original large sample dataset, effectively improving model computational efficiency.