Enhancing K-Means Clustering with Data-Driven Initialization and Adaptive Distance Measures

K-means clustering is a widely used unsupervised learning algorithm for partitioning data into distinct clusters. However, the performance of k-means heavily depends on the initial cluster centroid positions and the distance metric used. This paper proposes a novel approach to enhance the effectiveness of k-means clustering by incorporating data-driven centroid initialization and adaptive distance measures. The proposed method utilizes a density-based technique to identify potential cluster centers from the data, providing a more informed initialization compared to random or heuristic-based approaches. Additionally, the algorithm dynamically adjusts the distance metric based on the local data distribution, allowing for more accurate cluster assignments, especially in the presence of non-spherical or irregularly shaped clusters. Extensive experiments on various real-world datasets demonstrate the superiority of the proposed method over traditional k-means and other state-of-the-art clustering algorithms in terms of clustering accuracy and stability.