Determining the Optimal Number of Clusters using Silhouette Score as a Data Mining Technique

Abstract

The identification of the same objects is very important in determining the similarity between different objects. Nowadays, there are several techniques that allow us to divide objects into different groups that differ from one to another. In order to have the best separation between the clusters, it is required that the optimal determination of the number of clusters of a corpus be made in advance. In our research, the Silhouette score technique was used in order to make the optimal determination of this number of clusters. The application of such a technique was done through the Python language, and a corpus of unstructured job vacancy data was used. After determining the optimal number, at the end we present these clusters and the similarity between them, this presentation will be done in the form of a graph in a suitable format.

Psychophysiological strategies for enhancing performance through imagery–skin conductance level analysis in guided vs. self-produced imagery

“…Clustering is done into two clusters (expecting that they will correspond to investigated Tiers), for all participants and for each scenario, separately for guided and self-produced parts. The quality of clusterization into two clusters has been measured by the Silhouette score 56 . That score takes values from -1 to 1, and positive values, preferably higher than 0.5, prove correct clusterization.…”