Comparison of Clustering Methods on Iris Dataset

Data mining has evolved into a dynamic field essential for uncovering valuable insights within vast and complex datasets. Among its array of techniques, clustering plays a fundamental role, aiding both as an independent method and a precursor to other data mining methodologies. Within clustering, th...

Full description

Saved in:
Bibliographic Details
Published in2023 5th International Conference on Frontiers Technology of Information and Computer (ICFTIC) pp. 86 - 92
Main Authors Ye, Mingrui, Zhou, Ziqian, Zhou, Yijie, Huang, Yiran
Format Conference Proceeding
LanguageEnglish
Published IEEE 17.11.2023
Subjects
Classification algorithms
Clustering algorithms
Clustering methods
Data mining
empirical experiments
Flowering plants
Heuristic algorithms
Iris
Iris dataset
k-means
mixed data
Partitioning algorithms
performance comparison
spectral clustering
Online AccessGet full text

Cover

More Information
Summary:Data mining has evolved into a dynamic field essential for uncovering valuable insights within vast and complex datasets. Among its array of techniques, clustering plays a fundamental role, aiding both as an independent method and a precursor to other data mining methodologies. Within clustering, the k-means algorithm is prominent, offering efficient partitioning of datasets. However, it has limitations, particularly in handling non-numerical data. This paper explores solutions to these limitations and introduces the enhanced K-Means++ algorithm. Furthermore, it delves into spectral clustering, a graph-theoretic approach known for handling complex data structures. We conduct empirical experiments using the Iris dataset to compare the performance of k-means, K-Means++, and spectral clustering. This work sheds light on the strengths of these algorithms and their suitability for real-world applications.
DOI:10.1109/ICFTIC59930.2023.10456161