Improved the Performance of the K-Means Cluster Using the Sum of Squared Error (SSE) optimized by using the Elbow Method

Abstract

K-Means is a simple clustering algorithm that has the ability to throw large amounts of data, partition datasets into several clusters k. The algorithm is quite easy to implement and run, relatively fast and efficient. Another division of K-Means still has several weaknesses, namely in determining the number of clusters, determining the cluster center. The results of the cluster formed from the K-means method is very dependent on the initiation of the initial cluster center value provided. This causes the results of the cluster to be a solution that is locally optimal. This research was conducted to overcome the weaknesses in the K-Means algorithm, namely: improvements to the K-Means algorithm produce better clusters, namely the application of Sum Of Squared Error (SSE) to help K-Means Clustering in determining the optimum number of clusters, From this modification process, it is expected that the cluster center obtained will produce clusters, where the cluster members have a high level of similarity. Improving the performance of the K-Means cluster will be applied to determining the number of clusters using the elbow method.

A quantitative discriminant method of elbow point for the optimal number of clusters in clustering algorithm

“…This approach tends to repeatedly generate similar clusters for the dataset originating from the same data source. That is, this approach is stable for input randomization [14].…”

“…That is, the Elbow method does not always work well to determine the optimal cluster number [13]. The cluster number obtained by using the Elbow method is a subjective result because it is a visual method [14], and does not provide a measurement metric to show which elbow point is explicitly the optimum. To overcome these shortcomings of the Elbow method, a quantitative discriminant method is proposed to work out a straightforward value as the estimated potential optimal cluster number for the analyzed dataset.…”

A Quantitative Discriminant Method of Elbow Point for the Optimal Number of Clusters in Clustering Algorithm

“…This approach tends to repeatedly generate similar clusters for the dataset originating from the same data source. That is, this approach is stable for input randomization [14] . However, as mentioned above, when the elbow point is ambiguous, the Elbow method will become unreliable.…”

Estimating Macrofracture Toughness of Sandstone Based on Nanoindentation

“…The K-means algorithm includes the following steps: (1) determine the clustering number n through the elbow method [42,43]; (2) select n initial clustering centers stochastically and partition the objects to the nearest clustering center to form a cluster according to the nearest-neighbor rule;…”