Hybird Clustering Algorithm Based on BP Neural Network and K-means

In the era of big data, efficient and accurate clustering algorithms are indispensable for organizing and mining information. This paper focuses on the optimization of clustering algorithms, particularly by integrating the advantages of the Backpropagation (BP) neural network and the K-means algorithm to enhance clustering accuracy and stability. This paper proposed improvement strategy introduces a novel fusion method that combines the self-learning characteristic of the BP neural network with the crowd wisdom of the K-means algorithm, resulting in a hybrid model that dynamically adjusts cluster centroids. In this model, the BP network is responsible for generating and updating initial cluster centroids, while the K-means algorithm allocates samples based on these centroids. This complementary approach forms an iterative optimization process. The aim is to address the limitations of traditional K-means when dealing with complex data structures and the initialization challenge faced by the BP network in clustering tasks. The implementation section involves programming the improved algorithm using the sklearn library and testing it on the classic iris dataset. Through comparisons with the original K-means algorithm and clustering solely with the BP neural network,this demonstrates the superiority of the improved algorithm in clustering performance.