Application of the Bees Algorithm to Fuzzy Clustering
Authors: D.T Pham, H AL-Jabbouli, M Mahmuddin, S Otri , Ahmed Haj Darwish
Abstract
This paper discusses the application of the Bees Algorithm to fuzzy clustering. The Bees Algorithm is used to optimise the performance of the fuzzy C-Means (FCM) algorithm and improve its clustering results. Computational experiments show that the Bees Algorithm gives a significant improvement over the FCM algorithm on its own and better results compared to the FCM algorithm combined with a Genetic Algorithm.
| Attachment | Size |
|---|---|
| iproms v3.ppt | 809.5 KB |
Hello authors,
A very basic question, why FCM is so popular? Is there any other researchers working on this problem and to compare various different optimisation techniques to improve the performance of FCM?
Do you think the Bees Algorithm generally outperform the genetic algorithm in mathematical functions (based on your work on the calculus based problem and also the work on function optimisations from the Bees Algorithm first paper)? Based on your experience using the Bees Algorithm, why it outperforms the genetic algorithm?
Is there any progress on the work to reduce the computing time with the Bees Algorithm? Is the genetic algorithm also suffering the same high computing cost, which one is taking more time than the other?
Thanks
ang
The usage of bees algorithm or GA with c-means algorithm improved the clustering results not the algorithm complexity and Both of them the bees algorithm and GA are suffering from high computing cost.
In the same time we can say that this combination of the traditional fuzzy c-means algorithm with GA or Bees can improve the clustering results in most cases because of avoiding trapping in local optimum problem and the good results of the bees algorithm depends on the nature of the local and random search used in it and the way approaching to the solution.
Dear Authors,
It is very interesting paper.
I am wondering about the values you have chosen for the algorithms, is there any way to automate them to suit all cases?
Dr Ziad Salem
Aleppo University
Aleppo - Syria
Dear Dr. Ziad
Thank you for your question
Actually, it is one of the Bees Algorithm' disadvantages that it has a lot of parameters to set.
Currently There is a nice paper (Fuzzy Selection of Local Search Sites in the Bees Algorithm) Proposed in IPROMS dealing with this issue using Fuzzy logic.
Best regards
Hasan
Dear all,
I noticed that all the classes used in the all data set were discrete classes, what about the continuous ones?
Dr Ziad Salem
Aleppo University
Aleppo - Syria
Dear Dr. Ziad,
Each data object in the datasets can belong to more than one cluster in the same time with different membership value
Regards
Hasan
Dear Hasan,
Thanks for your reply.
Actually what I meant is what will happen when you have let say data set with continuous values classes such as: (0.2, 0.44,1.2,1.21, 1.26, 1.299, 1.4, 1.47, 1.55, etc).
Do you think the same algorithm would work on such data set?
Dr Ziad Salem
Aleppo University
Aleppo - Syria
Hi, I am wondering why the FCM algorithm perform better only on the Control Chart data set? I think there should be some discussions about the results.
Dr Ziad Salem
Aleppo University
Aleppo - Syria
Dear Dr. Ziad,
That is because of the dataset characteristics, if you want more info about this dataset please find this paper "Control chart pattern recognition using neural networks"
Regards
Hello Prof. Pham and salaam Hassan and all the bees
A good paper but I haven’t’ finished reading it. I’ll come back tomorrow.
Best wishes,
Shah
Hello Shah,
It was good to hear from you.
We hope you are collecting plenty of nectar and manufacturing first-grade honey in sunny Malaysia!
The Cardiff Bay Bees are also working beesily although some seem more beesy than others.
We look forward to receiving your comments on our results.
Best wishes from all of us.
DTP.
Dear Hassan,
Please copy this paper and presentation to Intelligent Optimisation Techniques as well.
Thank you.
Zaidi
Bees-Algorithm.org