Description of Software for Comparative Analysis of Clustering Methods

This paper presents the structure of software for comparative analysis of clustering methods. The entire path of the software is described. It all starts with data extraction. Then the stage of data preparation. The third stage is the stage of normalizing the values. Then comes the fourth stage – the stage of division into clusters. The software uses two clustering methods, namely k-means and agglomerative clustering. The main idea of the k-means method is that the data is randomly divided into clusters, after that the middle of the masses is iteratively recalculated for at least some cluster acquired at the last step, after that the vectors are again divided into clusters in accordance with which of the new centers came out to be closer to the selected indicators. Agglomerative clustering is a type of cluster analysis. The essence of the method is as follows. Several groups are being created. Any group contains one object at the initial stage, after it finds two more similar groups. Binds groups and this process is repeated until a whole group of similar instances is formed. The fifth stage includes quality assessment. For this, the silhouette method is used. The result of the program execution is the division of data into clusters by two algorithms, as well as obtaining an estimate of the quality of clustering.