چکیده مقاله
High dimensional data clustering is challenging due to the curse of dimensionality, irrelevant features, and reduced accuracy of traditional algorithms such as k Means This study proposes a hybrid multi stage clustering approach that first applies sparse feature selection to remove irrelevant features and then combines k Means, MinMax k Means, and DBSCAN The final clustering results are produced through a voting based aggregation mechanism and meta analysis Experiments on the standard Digits dataset show that the proposed method outperforms Sparse MinMax k Means in most external metrics, including ARI, NMI, V measure, Homogeneity, and Completeness, as well as in selected internal metrics such as Silhouette and Calinski–Harabasz However, Sparse MinMax k Means achieves a lower Davies–Bouldin index and shorter runtime Overall, the results indicate that the proposed hybrid approach improves clustering accuracy while introducing a trade off in computational efficiency
کلیدواژهها
نویسندگان
MohammadReza Bahrani
Artificial Intelligence Laboratory at NEVISA Engineering Team, Shiraz, Iran
Maryam Mazarei
Artificial Intelligence Laboratory at NEVISA Engineering Team, Shiraz, Iran
شیوه ارجاع
Bahrani، MohammadReza و Mazarei، Maryam،1405،Enhancing the Clustering of High-Dimensional Data Using a Hybrid Multi-Algorithmic Approach and Sparse Feature Selection،سی امین کنفرانس ملی علوم و مهندسی کامپیوتر و فناوری اطلاعات،بابل
ارائهشده در
مجموعه مقالات سی امین کنفرانس ملی علوم و مهندسی کامپیوتر و فناوری اطلاعات29 مرداد 1405 · بابل