Analyzing Of Clustering Algorithms for Achieving High Evaluation Metrics

Abstract

In the real data world, there are various clustering algorithms available in data mining. The data available from different data sources may be huge in instances, attributes, and in different formats. The clustering algorithms available are assessed based on how the algorithm clusters the given data and find its parametric values. The clustering of data may end in inappropriate results if the algorithm is not chosen wisely. This paper proposes a comparison between diverse clustering algorithms such as K Means clustering, Mini-Batch K Means clustering, Hierarchical clustering, Bagging, and Boosting by figuring out clustering strategies using high-dimensional datasets on each algorithm above. After the process of data cleaning in the dataset, we have clustered the datasets and compared the summary of each to showcase the comparability of the difference in their strategical values such as Clustering tendency, clustering quality, and data-driven approach for evaluating the number of clusters, Normalized Mutual Information (NMI) metric and provide an idea to choose the algorithm for clustering the data effectively. And as a result, Local Clustering Coefficient (LCC) with K-means clustering bunching method performs better than the other clustering algorithms and the results are reported.