New Version of Fast Automatic Determination of Cluster Numbers for High Dimensional Big Data

The main issue in clustering algorithms is how to efficiently define number of clusters automatically. Considering both the quality of clustering and efficiency of clustering algorithm during determination of number of clusters can be a trade off that was our main purpose to overcome with. Successfully, In our approach the best number of clusters for a large data set of high dimensional data automatically would be determined with respect to clustering quality and efficiency. We carried out experimental studies on our five previous data sets [43]. and 4 new larger ones by which we found that our procedure has the flexibility of choosing different criteria to determine the optimal K under each of them. The procedure takes the advantages of Bisecting K-Means algorithm, and indicated higher efficiency compared with Ray & Turi method when it comes to find the best number of clusters.