Master K-means Clustering: How the Algorithm Finds Compact Groups

1 K-means Algorithm

K-means is a simple yet classic distance‑based clustering algorithm that uses distance as a similarity metric, assuming that the closer two objects are, the more similar they are. The algorithm aims to produce compact and independent clusters as the final goal.

It iteratively searches for a partition of the data into k clusters such that the total error, measured by the sum of squared distances between samples and their cluster centroids, is minimized.

The k clusters have the following characteristics: each cluster is as compact as possible, and different clusters are as far apart as possible.

The basis of K‑means is the minimum squared error criterion. If the data is expressed mathematically, the objective is to minimize the sum of squared errors between each point and its assigned cluster centroid (the mean vector of the cluster, also called the centroid).

Directly finding the global minimum is NP‑hard, so a heuristic iterative method is used.

2 Steps

Step 1: Randomly select k cluster centroids.

Step 2: Repeat the following until convergence: For each sample i, assign it to the nearest cluster.

Step 3: For each cluster, recompute its centroid.

Convergence is reached when the distance between the newly computed centroids and the previous ones falls below a predefined threshold, indicating that centroid positions have stabilized.

Reference

ThomsonRen GitHub https://github.com/ThomsonRen/mathmodels