Clustering Faster and Better with Projected Data

Abstract

The K-means clustering algorithm can take a lot of time to converge, especially for large datasets in high dimension and a large number of clusters. By applying several enhancements it is possible to improve the performance without significantly changing the quality of the clustering. In this paper we first find a good clustering in a reduced-dimension version of the dataset, before fine-tuning the clustering in the original dimension. This saves time because accelerated K-means algorithms are fastest in low dimension, and the initial low-dimensional clustering bring us close to a good solution for the original data. We use random projection to reduce the dimension, as it is fast and maintains the cluster properties we want to preserve. In our experiments, we see that this approach significantly reduces the time needed for clustering a dataset and in most cases produces better results.