Data Clustering and Validation For Traffic Management Support System

Abstract

Clustering is used to classify or group the sets of data in unsupervised way. Different clustering algorithms including partition k-mean and hierarchical k mean with Elbow method, Davies boulder index, Dunn index, Silhouette coefficient are implemented to find out the optimal classification on the features of traffic management support system. Thereafter, the validity of the optimal classes are analyzed by WSS (within sum of square) errors. Dunn index seems better than others. Correlation method is used to find the better clustering algorithm between K-mean and hierarchical algorithms. According to correlation partition k-means performs better. Thus, partition k mean with Dunn index is used to determine the optimum number of cluster for five dimensional road weights.