Scribd
Upload
19 views6 pages

Fuzzy C Mean

fuzzy c mean

Uploaded by

bensalemhalim19
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
19 views6 pages

Fuzzy C Mean

fuzzy c mean

Uploaded by

bensalemhalim19
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 6

Algerian Democratic and Popular Republic

Ministry of Higher Education and Scientific Research


Mustapha Stambouli University of Mascara
Faculty of Exact Sciences
Computer Science Department

Fuzzy C-Means

Presented by: Supervised by:

MerBELKHAHLA Mohamed Mustepha MerSalem.


MerBensalem Adbelhalim
Academic year: 2024 – 2025
1. Definition and When to Use It:
Definition:
Fuzzy C-Means is a clustering algorithm where data points can belong to multiple clusters with varying membership degrees. It’s commonly used in
scenarios where the boundaries between clusters are not well-defined or when soft clustering is preferable.

When to Use:
 When you need overlapping clusters.
 In fields like image segmentation, bioinformatics, and market segmentation.
 When data has ambiguous or fuzzy boundaries.

2. Principle:
 Assign each data point a membership value to every cluster.
 Cluster centroids are computed based on weighted averages of data points.
 Iteratively update memberships and centroids to minimize the objective function.
 Achieve optimal clusters by balancing the distance of points to centroids and their degrees of membership.

3. Mathematical Foundation:
Objective Function:
𝑛 𝑐
𝑚
𝐽𝑚 = ∑ ∑ 𝑢ⅈ𝑗 ‖𝑥ⅈ − 𝑐𝑗 ‖2
ⅈ=1 𝑗=1
Where:

 𝐽𝑚 : Objective function
 𝑢ⅈ𝑗 : Membership of point ⅈ in cluster 𝑗
 𝑥ⅈ : Data point
 𝑐𝑗 : Centroid of cluster 𝑗
 𝑗: Fuzziness parameter, 𝑚> 1 (controls cluster softness)
Membership Update:
1
𝑢ⅈ𝑗 = 2
‖𝑥ⅈ − 𝑐𝑗 ‖ 𝑚−1
𝑐
∑𝑘=1 ( )
‖𝑥ⅈ − 𝑐𝑘 ‖
Centroid Update:
∑𝑛ⅈ=1 𝑢ⅈ𝑗
𝑚
𝑥ⅈ
𝑐𝑗 = 𝑛 𝑚
∑ⅈ=1 𝑢ⅈ𝑗
4. Algorithm Description:
Initialize:
 Define𝑐 (number of clusters), 𝑚 (fuzziness parameter), and stopping criterion.
 Randomly initialize the membership matrix 𝑈.

Update Centroids:
 Compute 𝑐𝑗 for each cluster.
Update Memberships:
 Recalculate 𝑢ⅈ𝑗 for each point and cluster.
Check Convergence:
 If the changes in 𝑈 or 𝑐𝑗 are less than a threshold, stop. Otherwise, repeat.
5 .Hyperparameters and Their Values:
Number of Clusters (𝒄):
 Usually determined based on the application or using techniques like the elbow method.

Fuzziness Parameter (𝑚):


 Typical value: 1.5 ≤ 𝑚 ≤ 3.
 Higher 𝑚 increases overlap between clusters.
Stopping Criteria:
 10−5 ).
Threshold for convergence (e.g.,

 Maximum number of iterations (e.g., 100 − 300).

Distance Metric:
 Euclidean distance is most common.

6. Pros and Cons:


Pros:
 Handles overlapping clusters well.
 Flexible and interpretable membership values.
 Suitable for noisy data.

Cons:
 Computationally expensive for large datasets.
 Sensitive to initialization and noise.
 Requires predefined number of clusters.
Thank you

You might also like