Paper Info
Title
Raddacl: A Recursive Algorithm For Clustering And Density Discovery On Non-Linearly Separable Data
Abstract
In the realm of unsupervised classification, clustering algorithms have presented themselves as being inefficient in the detection of non-linearly separable or non-spherically shaped based clusters, without utilizing complex implementations or computational expense. In clustering algorithm advancements, excellent results are produced with complex concepts and implementations. Another current downfalls is that very advanced clustering algorithms proved to be computationally expensive methods. This paper presents a recursive clustering algorithm that is simple to conceptualize and implement, without sacrificing the ability to cluster the presented information. The algorithm has proven its worthiness in experiments to date. This paper explains the algorithm, named RADDACL (Recursive Algorithm for Density Discovery and CLustering), its origin, explanation and inspiration, along with preliminary testing results. With the simplicity of conceptualization and implementation, RADDACL is a prime algorithm for introduction to several topics in Computer Science education (and related fields) such as recursion, knowledge discovery, and clustering.
Year
DOI
Venue
2007
10.1109/IJCNN.2007.4371202
2007 IEEE INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS, VOLS 1-6
Keywords
Field
DocType
density,recursive algorithm,computer science education,clustering,recursion
Canopy clustering algorithm,CURE data clustering algorithm,Data stream clustering,Correlation clustering,Computer science,Determining the number of clusters in a data set,Theoretical computer science,Constrained clustering,Artificial intelligence,Cluster analysis,DBSCAN,Machine learning
Conference
ISSN
Citations 
PageRank 
1098-7576
3
0.49
References 
Authors
2
2
Name
Order
Citations
PageRank
Derek Beaton1475.52
Iren Valova213625.44