Scribd
Upload
7 views

K-means Clustering (1)

The document outlines the K-means clustering algorithm, detailing the steps for dataset preparation, training, and plotting results for different values of K. It emphasizes the importance of not using certain libraries and provides instructions for submission along with limitations and guidance on choosing the optimal K based on inertia. The report should include plots for K values of 2, 4, 6, and 7, along with the corresponding inertia values.

Uploaded by

zarifahmed180
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
7 views

K-means Clustering (1)

The document outlines the K-means clustering algorithm, detailing the steps for dataset preparation, training, and plotting results for different values of K. It emphasizes the importance of not using certain libraries and provides instructions for submission along with limitations and guidance on choosing the optimal K based on inertia. The report should include plots for K values of 2, 4, 6, and 7, along with the corresponding inertia values.

Uploaded by

zarifahmed180
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/ 4

Offline 3

4: K-means Clustering
Dataset preparation:
Use dataset ‘dataset.txt’ in the given folder.

Train:
1. K=4
2. Load dataset into 2D list "Data"
3. Randomly select K different data points from “Data” and store them into 2D list "Centers"
4. Initialize a 2D list named "Clusters" which contains K 1D lists for the K centers
5. for each sample/ data point "S" in "Data":
6. identify the center “C_i” that is the closest to “S”
7. Append "S" in "i"th list of "Clusters"
8. itr = 1, “Shift” = 0
9. while True:
10. for each 1D list "L" in "Clusters":
11. Determine the average of the data points. This is the new center of this list.
12. Update the center of this list in “Centers”
13. if itr > 1 and "Shift" < 50 break (convergence)
14. “Shift” = 0
15. Initialize a 2D list named "Temp_Clusters" which contains K 1D lists for the K centers
16. for each sample/ data point "S" in "Data":
17. identify the center “C_i” that is the closest to “S”
18. Append "S" in "i"th list of "Temp_Clusters"
19. if S belongs to different clusters in “Clusters” and “Temp_Clusters” then
20. “Shift” = “Shift” + 1
21. Now "Temp_Clusters" 2D list contains K 1D lists
22. Assign "Temp_Clusters" to "Clusters"
23. itr = itr + 1
24. "Clusters" will contain your desired clusters and "Centers" will contain your desired centers at
the end of loop
25. Plot them with appropriate color
26. “inertia” = 0
27. for each 1D list "L" in "Clusters":
28. “inertia” = “inertia” + sum of distances-square of data points of “L” from the center
Report:
Plot the data for K = 2, 4, 6, 7 and note down inertia.

Instruction
● Submit a .ipynb file and a report (report template) .pdf file.
● You must follow the given algorithm
● DO NOT USE LIBRARIES SUCH AS: "Sklearn", "Scikit learning" or "pandas" for this assignment
● Use your student id as seed
● Copying will result in -100% penalty
● Your marks will fully depend on your viva and understanding.
○ Full Algorithm: 16
○ Plotting: 4
Resources
k-means clustering

1. Select K random data points as the centers of K clusters


2. Assign each datapoint to the closest clusters (by calculating the distance from
centers).
3. While True:
4. Recalculate the center of the clusters (which is the mean of the data points)
5. Reassign each datapoint to the closest cluster
6. If no datapoint changes cluster then
7. break

Limitations:
- Need to know K in advance
- Depended on initial assignment of the centers

How to choose the K?


- Inertia measures how well a dataset was clustered by K-Means. It is calculated by
measuring the distance between each data point and its centroid, squaring this distance,
and summing these squares across one cluster. A good model is one with low
inertia AND a low number of clusters ( K ).

You might also like