Optimal Set of Overlapping Clusters Using Multi-objective Genetic Algorithm
| Autor: | Shreya Chaudhuri, Asit Kumar Das, Sunanda Das |
|---|---|
| Rok vydání: | 2017 |
| Předmět: |
Fuzzy clustering
Single-linkage clustering Correlation clustering 0211 other engineering and technologies 02 engineering and technology computer.software_genre Complete-linkage clustering Determining the number of clusters in a data set ComputingMethodologies_PATTERNRECOGNITION 0202 electrical engineering electronic engineering information engineering FLAME clustering 020201 artificial intelligence & image processing Data mining Cluster analysis computer k-medians clustering 021101 geological & geomatics engineering Mathematics |
| Zdroj: | ICMLC |
| DOI: | 10.1145/3055635.3056653 |
| Popis: | Clustering is an important unsupervised machine learning techniqueused in diverse fields to explore the inherent structure of the data. In most of the real life datasets, one object resides in many clusters with different membership values. Many clustering algorithms have been proposed for finding such overlapping clusters for knowledge extraction and future trend prediction. In the paper, multi-objective genetic algorithm based cluster analysis technique is proposed for finding the optimal set of overlapping clusters. As most of the real world search and optimization problems involve multiple objectives, multi-objective Genetic Algorithm is an obvious choice for capturing multiple optimal solutions. Thus the usefulness of applying the multi-objective Genetic Algorithm is to grouping the objects based on different objective functions for finding optimal set of overlapping clusters. The advantage of this algorithm is that it assigns a membership value only to the objects which are the members of several clusters, instead of assigning membership values for all clusters like fuzzy clustering algorithm. If any object positively belongs only to a single cluster, its membership value for this cluster is ' 1' and '0' for all other clusters. The overall performance of the method is investigated on some popular UCI and microarray datasets and the optimality of the clusters is measured by some important cluster validation indices. The experimental results show the effectiveness of the proposed method. |
| Databáze: | OpenAIRE |
| Externí odkaz: |
|