Research on Improved k-Means Clustering Algorithm Based on Density-Weighted Canopy

This paper proposes an improved K-means clustering algorithm based on density-weighted Canopy to address the efficiency bottlenecks and clustering accuracy issues commonly encountered by traditional K-means algorithms when handling large-scale datasets. Traditional K-means relies on random initialization of centroids, making it susceptible to outliers and prone to getting stuck in local optima, particularly in high-dimensional data. To overcome these challenges, the Canopy algorithm is first applied to preprocess the dataset, assigning weights to data points according to their local density distribution. This helps in selecting more representative cluster centers. Subsequently, the density-weighted K-means algorithm is used for iterative optimization, enhancing the accuracy of centroid updates and mitigating the impact of outliers on the clustering performance. Experimental results demonstrate that, compared to the traditional K-means algorithm, the proposed density-weighted K-means algorithm significantly improves both clustering quality and computational efficiency, especially when handling high-dimensional and large-scale datasets.