Year: 2024
Type: article
Abstract: Prediction is a developing domain that requires further advancement. In practical situations, one frequently encounters unsupervised datasets requiring analysis. Clustering techniques can efficiently resolve this difficulty. The current study designs the hybrid hierarchical density model to enhance clustering efficiency through a combination of hierarchical clustering and DBSCAN. Hierarchical clustering's initial partitioning quality, which lays a strong foundation for DBSCAN's refinement, is a key influencing factor for these enhancements in this model. DBSCAN's ability to adjust to local density variations further enhances clustering quality. This model is compared with conventional clustering models such as K-Mean Clustering, Grid-Based Clustering, Agglomerative Clustering, Gaussian Mixture Model, Constraint-Based Clustering, and Spectral Clustering using five different datasets named CIFAR-10, MNIST Handwritten Digits, Cats and Dogs, Plant Seedlings, and Oxford Pets Datasets. The goal of this study is to create more accurate and meaningful clusters by continuously improving parameters and using performance metrics that show average results, as the silhouette score value is 0.7, the Davies-Bouldin index is 0.43, the Rand index value is 0.89, the adjusted mutual information is 0.67, and Fowlkes-Mallow’s index is 0.57.
Author: Swati Gupta
Institution: Maharshi Dayanand University
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: bronze
OpenAlex ID: W4407040508