A new approach to present prototypes in clustering of time series
There are considerable advances in clustering time series data in data mining concept. However, most of which use traditional approaches and try to customize the algorithms to be compatible with time series data. One of the significant problems with traditional clustering is defining prototype spec...
Saved in:
| Main Authors: | , , , |
|---|---|
| Format: | Conference or Workshop Item |
| Language: | English |
| Published: |
2011
|
| Subjects: | |
| Online Access: | http://eprints.um.edu.my/13448/1/A_new_approach_to_present_prototypes.pdf http://eprints.um.edu.my/13448/ |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| Summary: | There are considerable advances in clustering time series data in data mining concept. However, most of which use traditional approaches and try to customize the algorithms to be compatible with time series data. One of the significant problems with traditional clustering is defining
prototype specially in partitional clustering where it needs
centroids as representative of each cluster. In this paper we present a novel effective approach to define the prototypes based on time series nature. The prototype is constructed based on fuzzy concept efficiently. Moreover, it is demonstrated how the prototypes are moved in iterations. We will present the benefits of the proposed prototype by implementing a real application: Customer transactions clustering. |
|---|