Implementation of K-Means Clustering Algorithm with Code Description

In our experiment, we developed a program implementing the K-means clustering algorithm. The algorithm operates by identifying C cluster centroids within training samples, with each centroid representing the central point of a class. Samples are then classified into the category corresponding to their nearest centroid. In practical applications, the value of C can be determined through prior knowledge or empirical experience. The algorithm iteratively computes cluster centroids through the following steps: First, C samples are randomly selected as initial centroids. The program then calculates distances between each sample point and these C centroids, typically using Euclidean distance metrics. Each sample is assigned to the class of its closest centroid. Subsequently, for each class, the centroid is recalculated as the mean position of all assigned samples, and these updated positions become the new centroids. This iterative process continues until centroid positions stabilize (convergence criteria are met), resulting in K distinct clusters, each represented by its characteristic centroid. The K-means clustering algorithm finds extensive applications in data mining, image processing, and related fields. Implementation typically involves functions for random centroid initialization, distance computation, sample assignment, and centroid updates, with convergence monitored through maximum iteration counts or centroid movement thresholds.